Installer(s) Thread
@Zeglius @tulip @jade @Niklas ⚡ @antheas I would like to see if we can hack on getting a live ISO installer up and running. I know we are mostly targeting bluefin LTS for now as a proof of concept, but I definitely think it would be good to get the Fyra Labs folks involved since they are the maintainers of Katsu and we could investigate using ReadyMade as a way to deploy the actual image onto the disk.
https://github.com/Zeglius/bootc-livecd-katsu
@Zeglius I’ll take some time to look at the katsu documentation, but when you have some time, I would be curious to understand how this works a bit better. No pressure on anyone, just trying to brain dump while I’m on the plane.
9101 Replies
@jade add anyone from the Frya Labs team to this thread so that we can have a more direct discussion.
@thirtyseven, @Owen, @lea, @madomado
we’re ready to help with whatever you need on katsu <3
Pretty sure there was some work/testing done with installing bootc images with readymade. Shouldn’t be hard
Right so what got in that repo probably collides a bit with the way katsu is normally used.
For what I have seen in their images repo, is a more traditional approach of building a root filesystem with dnf.
We're as in mine, I literally just dumped the container root filesystem with a script
i noticed you had one config file yeah
we could add functionality to import a rootfs from an OCI root in Katsu maybe
technically you can expand it yourself before running katsu
One of the main things we want to avoid is having the image be on there twice. It's why our current ISOs are so dang big.
afaik composefs can mount an OCI archive
Offline deployment would still be best, if we can somehow make that happen.
we'll just need some initramfs magic
Composefs still isn't in the upstream images is it?
its on centos-bootc
using it right now
/sysroot/ostree/repo/objects as the datadirAhhh ok, so missing from fedora then?
yup
I know Timothee and friends are working hard on it, so I'm sure it will eventually be here.
actually i think
fedora-bootc has it?I'm not sure that Bazzite has composefs yet.
here are the composefs options on centos-bootc:
ro,relatime,seclabel,lowerdir+=/run/ostree/.private/cfsroot-lower,datadir+=/sysroot/ostree/repo/objects,redirect_dir=on,metacopy=onif you bind-mount the root to katsu/katsu-work/chroot it should theoretically work (tho it's untested because… well hidden "features")
we can avoid duplicate stuff if we use
rpm-ostree deploy for the installation
cuz then we can just have the /sysroot/ostree stuff
bootc install seems to not have host support...
we need to talk about this with colin actually im sure he'd have some insight to shareI can just add him to the thread.
he doesnt use discord much
better way to reach him is to talk to him on matrix
fedora-bootc room on the fedora space
@Zeglius I looked at it again
You will need to use rpm-ostree to stage the / instead
The way you did it will not work for derived images and cause a delta
Then you will also need to use rpm-ostree to create the ostree store
Probably the way you'd do this is first deploy the ostree store with rpm-ostree, that will not dupe the image to container storage
Then stage the image via hard links and place ostree in there
Then chroot, put your installer stuff
And squashfs the whole thing
Looking at the installer stuff again
Services — LVGL
LVGL is the most popular free and open source embedded graphics library targeting any MCU, MPU and display type to build beautiful UIs. We also do services like UI design, implementation and consulting.
This is the goat business model
Having an installer embedded into a separate partition sounds really nice. No need to bother with the iso stuff
Plus not running out of ram with the overlay, no space constraints
goated
Blue95 Live ISO + Installer? 👀
@tulip
omg 😭
this is awesome?
this is SO COOL
man we really need to figure out how to run bootc install on that media :(
itll be so awesome
Yep. "Then we'll be done!"
exactly.
this actually would be so goated
this looks so awesome
i cant get over this
i can see like retro youtubers making videos about this LOL
gotta get it to work first 😭
i know, is there any way to debug dracut?
maybe we can like get a shell inside the initramfs and try to get the
/var/lib/containers thing mounted into the system root as a bindmount from the initramfs instead of it being inside overlayfs
we just need the /var/lib/containers/storage folder to not be an overlay
and maybe /var/tmp / /tmpGive me some rust code or tell me what I need to do, then I'll wire it up in YAI
I added the installer prototype to the blue95 image. The ISO now boots to the desktop and I can launch the installer.
Time for some sleep 😴
I read something about dracut
rd.break=<step> kernel param
I feel that might not be an issue, maybe we just need the runroot there to point to a tmpfs insteadLike this(ignore the graphroot)
Doing something seems like a start point to overcomplicating the setup
oooo fancy
but you also need the persiste t storage tho
Persistent storage for what
sorry i was sleepy answering that
so um
we need the
/var/lib/containers/storage part to not be an overlayfs mountThat or using vfs
else it wont work for bootc install
we need it to be very fast
i dont kno exactly if vfs is good for that
but yeah still
we just need
podman run to run just like regular ol storageAlso we need to use squash merges to merge PRs, merges alone are too messy to undo
that
--rootfs finally found its use
No containerfiles
try doing
ln -s /usr/bin/dnf /usr/bin/dnf5 btw
or similar i dont know
cuz centos-bootc is gonna break LOL
but thats so awesomenah, then a bash function
we dont want to tamper the input (image) too much
@Zeglius i think i found the solution
to the podman run stuff
sudo podman image save -o myimage.oci --format oci-archive ghcr.io/ublue-os/bluefin:lts
podman run --rm -it oci-archive://$(realpath $PWD)/myimage.oci
nah didnt work wellwell, that was simple
if you get an OS image onto a livecd ping me, then we'll get the installer hooked up
We got it for a while, we just need to polish the edges
and get the container image inside it
yeah
its already working, we just need to make
bootc install work 😭ah
gotcha
i am considering just using rpm-ostree deploy now
oh using legacy software, ah huh (/s)
but if it works, why not?
Can't you just podman pull inside a Containerfile? Then create the live ISO from that image
Or https://gitlab.com/fedora/bootc/examples/-/tree/main/physically-bound-images?ref_type=heads
Yeah
Though the problem is when running it.
Because the root filesystem mounted by the iso is an overlayfs
its just too slow
root filesystem is overlayfs, we need to run
fuse-overlayfs to get podman run to work
and that takes waaaaaaaaaaaaaaay too long
we need to mount something on /var/lib/containers to make it happier, like a temporary ext4 partition or whatever i dont knowNope. You would run into variable behavior at runtime if you have a persistent storage
i dont mean that, like
maybe we can get like a separate partition or
.img file that gets mounted as :ro or something under /var/lib/containers/storage
we'd do that during the iso build, then make dracut mount it there
i was thinking about just dumping the entire container into ram via a tmpfs but probably not a good idea LOLMmm maybe. I will after lunch, to see if podman runs fine in a fedora workstation livecd
(akthough it would be pretty fast...)
My 8 gb ram will choke :v
yeah LOL
Plus if someone wants a kiosk setup, would be too demanding
i dont really care about that right now
but yeah it wouldnt work very well yet
new minimum req: 16gigs
Temporary workarounds end up lasting for the rest of the project lifespan :v
Let's avoid that, and see how others do
Boots in less than 30 seconds to the desktop to open the prototype installer
I love how you even included the og setup background in this 😭
NICE
this looks so awesome 😭
@j0rge
Now I should work on the Aurora one lol
Man this should be set in a museum or something
I love this, probably just buying a laptop run this for fun on heh
Oops just did
And the money goes to a refurbisher who employs people, can't get nicer :)
Alright now I have time to check how podman would do in a workstation LiveCD
the TLDR is getting podman working on the overlayfs-as-root environment
Oh
funny enough, podman seems to be preinstalled in the workstation livecd
oh nice LOL
you might need fuse-overlayfs for it to work btw
there are static binaries on their github repo
just plop that into
/usr/bin and podman is going to work, although quite slowlydnf install fuse-overlayfs
yup
dont think too much
wait
it works
like ootb
Note: this might be useful for us, at storage.conf
read/only image stores is interesting
:huh:
if you have fuse-overlayfs
Is this what Bootc uses for LBIs?
and its slower
Its commented
for whatever reason it still uses it
if you have it installed
So is like hardcoded?
i think so? LOL
if you uninstall it you'll see that podman wont work
Ah shit you are right
You know what? lets just use fuse-overlayfs :v
i tried
its way too slow for
bootc install :(
took about an hour on my machine to just partition some stuff
then it blew upit blew up?
GitHub
ERROR Linking tmp mounts to host: Exchanging /tmp <=> /tmp.tmp: Inv...
I have hit this error ERROR Linking tmp mounts to host: Exchanging /tmp <=> /tmp.tmp: Invalid cross-device link (os error 18) with example image in install.md file, i am currently using ubunt...
Supposedly fixed here
https://github.com/bootc-dev/bootc/pull/225
GitHub
install: Don't retarget /tmp or /var/tmp if they're not overlayfs b...
It may be that we're involved via a container flow where e.g. /tmp is already "properly" set up as a tmpfs.
In that case we don't need to do a dance in retargeting...
i ran into it...
on bootc 1.1.6
so i guess it didnt fix it 😭
Mmm, gonna try running bootc install from the workstation livecd in the vm
k, forgot i can run out of ram
Maybe if i enable composefs there...
true that might be interesting
for the container-storage
yeah
not sure itll work but its interesting
I mean that, or tearing apart the laptop to slap a mem stick :v
I mean still, the installer is meant to wipe the entire disk, right? We can create in the storage drive a partition that serves solely as a containers-storage
thats exactly what i was thinking about :)
we probably want something like that
but like
on the drive is probably not a good idea? i was hoping we could do a partition inside the ISO itself so we dont need to worry about the target before installation
Iso is read-only tho
yeah that sucks
Well let's see if I find the same error
Fuck. Ran out of ram again
its cloning the entire container into ram? 😭
Yup
Composefs won't save my ass it seems
Gonna try with our livecd and fuse-overlayfs
found out the mounting stuff
rd.break=pre-pivot rd.shell on the kernel command line lets you see what the fuck is going on before the initramfs pivots to the sysrootFedora Docs
How to debug Dracut problems
Foreword If you are experiencing a problem with system initialization due to Dracut, please see the common bugs document before filing a bug. Some easy configuration tweaks that fix a…
Play a little minesweeper while you're installing?
I put up a PR to remove all of the old manual livecd stuff which is not needed anymore since we're using livesys-scripts for that https://github.com/ublue-os/titanoboa/pull/11
Aurora takes a little bit longer but it automatically logs in now.
So in the main branch seems like we have support for GNOME, KDE and Xfce based images now
with the liveISO scripts stuffs?
NICE.
im trying to get containers storage going now
Yep, this is with the stuff Zeg added
man the splash screen
this is so cool 😭
let's gooooooooooooooooooooooooooooooo
@Niklas ⚡
even if we dont get the installer working for one reason or another
having these is very interesting
Let's goooooooooooooooooo
Alright that took me a while to setup lets see how the fuse-overlayfs does in an install
lets see how it goes
Ok it seem it got stuck or something?
yup, itll take forever
about 30 minutes and itll start partitioning
ok yeah fuse-overlayfs is not an option thenn
also.
say what
R/O true works with podman run?
30 mins?
yup it takes forever
ah yeah
we are done then
👍
we can just bindmount the containers directory straight from the initramfs root
and use it as r/o for installation w/o the overlayfs on top of
/var/lib/containers/storagethough i think you need to explicitly set the flag
--imagestorage= to be detected
that or modify storage.confnah we dont need to do anything
oh wait you mean directly in var/lib
its just the overlay driver we need to get working
i dumped the image in /usr/lib/containers/storage instead
we can just bindmount from
/containers to /sysroot/var/lib/containers on the initramfs via the fstab-sys dracut module
also worksseems like a solid plan
though, maybe /usr/lib/containers/storage would work better, just in case we need to pull images at runtime for something else
eh whatever
All it would need (bootc) is an additional volume mount to that directory
I love everything about what's going on.
I love the speed in this because it's been basically started yesterday
what's the tldr on secure boot?
we want it but it will take months to get
wym
oh on the image yeah custom kernel = no secure boot
vanilla kernel + akmods half broken test drive
I mean on the ISOs, we have it now with the existing setup
depends on the kernel the custom image has
i think that it should work
like our bluefin ones, I know LTS is missing
but what about stable and gts?
i dont even know if UEFI is working yet cuz this thing is super weird
so, unknown there
it probably would just be like that shim thing? we probably can add it yeah
YOYOOYOYOY
@ledif
if you put
rd.live.ram on the kernel command line during boot from that iso you'll make the entire thing run in RAM
its pretty interesting, ill probably add that as another grub option
oh my god
@Robert we can do a wolfi live ISOAwesome! Just need to package grub
grub is a bit of a pain
but you dont need to do it i believe
you dont need grub in the rootfs for this to work
its just a tool
grub-mkrescue that does the job afterwards@ledif @Zeglius @Niklas ⚡
bootc install /dev/vda --source-imgref containers-storage:ghcr.io/ublue-os/bluefin:lts
done
even with the overlayfs filesystem it doesnt matter
its a really fast install@j0rge
LES GOOOOOOOOOOOOOOOOO I T WORKS
IT WORKED
YAAAAAAAAAAAAAAAA
This looks awesome!
Is the OCI image stored on the ISO?
Is the OCI image stored on the ISO?
yup.
/var/lib/containers in the ISO
.
:)
we are so back
it literally just works
and works very fastDid the ISO end up being huge? I know there's things we can slim down if that's the case
AND we have a very conveninent rootfs mountpoint at
/run/bootc/storage
check this outinstalled it from here.
Wha... How?
What's the regular BIB ISO? I know there's technically no GUI installer, but it shouldn't take up too much more space
the BIB iso is bigger
dunno why dunno how
this works.
and its somehow smaller
we jsut solved our ISO problems like as simple as that
ill include flaptaks on-iso now and see how big it gets
i have no idea how this is so small
but im 100% sure the container image is in there
Does the ISO share the same files as the OCI image? Like some composefs symlinking magic or something?
Or is it two copies of Bluefin?
no idea
i feel like its overlayfs metacopy doing an amazing job
cuz most of the stuff in the image is the same as the host
like byte-for-byte
mount options on the ISO
A 4.83GB OCI PLUS whatever you need to make the live ISO in an ISO under 2.5GB
it must be some crazy compression
i am doing zstd compression on the initramfs and some other stuff
like i am not even kidding this is the actual ISO i installed from
Please push what you have! I need to go diving into how it works 🙂
its just the stuff we already have :)
just build the thing and ur doneFlatpaks then let's goooooooooooooooooo
just make sure to uncomment this line
Slap a minimal Gum TUI on it, with some flatpaks and ship it!
We can probably make YAI functional on bluefin over the weekend
yup
its already functional we just need to wire up some stuffs and bam done
:)
we have the disks
Yup
just a single program call in rust
this can be further optimized to be smaller too
do you have the yai repo?
Well please revise the disk fetching so it only fetches full disks and not their partitions 😭
Wait directly without podman?
How would we handle user creation? Would the installer handle that, and we switch from
bootc install to system-reinstall-bootc?yup.
Holy
this is so goated
yai sounds awesome doesnt it
i like the name very much
bwrap into useraddWe can chroot into the filesystem and useradd
ye
GNOME already handles that for us with gnome-initial-setup
but for kde and the others we can just do that
I probably need to go sleep now but take the code and go ham
Alright, lorax down with our live CD.
Next target: Anaconda
we need to figure out the god damn UEFI too tho
i have no idea how to get UEFI support on this
Shouldn't UEFI be as easy as dropping an uki in there and done?
no idea
we are using
grub-mkrescue to build the ISO, if that helps on anything
(although i wish we could use systemd-boot eventually)Yeah that would be neat
Make it ugly tulip and just dump the rust code in the repo I guess, we’ll make it beautiful tomorrow 🔥
For now let's keep hitting the iron while it is hot and keep going now with the installer itself
LiveCD in almost 2 days is crazy
Isn't there a grub-uefi package?
Or did that not work?
Mmm you are right, we didn't try that
Actually let me try, should be as easy as simply swapping the grub package with grub-uefi
im getting flatpaks goiong now
I think that's the package you use in arch to generate grub when doing a manual install.
hopefully
:P
guys
goals
- UEFI support
- flatpak in ISO
- simple TUI installer
if we can split this work it would be great
im working on the flatpaks on the ISO
if someone can get UEFI working or at least check possible stuff it would be awesome
cc @Noel @Zeglius
founbd it
For the tui, I think with
dialog can get something done
Ohfair
The CD boot one looks interesting
yup seems like exactly what we need.
That command isn't working for me
oh i forgot to mention
rm -rf /var/tmp
then ln -s /tmp /var/tmp
as root
then it just starts working
did it work? @ledif im curious
:pSeems to be
NICE
the install is super fast isnt it
We can add a tmpfiles.d rule to remove /var/tmp and add it back as a link to /tmp
sgtm!
flatpaks on the ISO
Already???
LMFAO yup.
Ran out of disk space :coolfingergunscry:
oh no LOL
i wonder if thats ram related
check out the ram on the vm
I demand that we use this for our installer.
its amazing isnt it
blue95 looks so awesome LOL
@M2
with the flatpaks the ISO is surprisingly not that much bigger
also added a few thingies
OLHY SHIT
HOLY HI
@j0rge
maybe we can do that in the Justfile instead
@Robert
donno
missed opportunity to add pinball 3d :p
we need to manually include the runtimes like the other ISO cuz it doesnt do dependency management for whatever reason
maybe is somewhat detecting the one in host? unless you are using podman run
prolly
im on the host
ahh there
This is way out of scope.... But you should add memtest and an entry in grub along with EFI shell
Very hype
not really tho
memtest i dont know
but the EFI shell is just like a single entry in grub
we just gotta figure out the thing we need to add there
Both require the EFI executable and can be stuffed into /boot + grub entries.
memtest is such a useful tool for making sure your RAM hasn't gone bad.
EFI shell is a bit more niche but not all UEFI ship an EFI shell.
If were making an ISO and already modifying the entries a little, it's a decent opportunity to add some tools that you may want around install time
we probably can do that ye
no idea how that would work but seems like it would be nice
we already have access over
/boot on the ISO so we can just shove in anything thereBut it's probably way out of scope right now and we have secureboot issues with them as well.
Everything gets stuffed into boot and boot/efi
secureboot i dont think so? we just need to figure out how the hel to get EFI support there
i wnated to find out how exactly
grub-mkrescue worksDo we have UEFI support?
i know it uses xorriso on the backend but dunno exactly what it does
It might be worth referencing the bootupd static grub config. Just to see what we can do
https://github.com/champtar/bootupd/tree/grub2-password/src/grub2
Might help, probably won't
not yet™️
but it seems to be a limitation with
grub-mkrescue which we can.. just not care about this if we dont use it
i was trying to make the ISO with syslinux before and apparently thats not necessary LOL
but im pretty sure we can shove in anything we want thereDo you have grub2-efi-x64-modules installed?
SysLinux is dead
i just installed
grub2-efi on the container
thats most likely what we need
grub2-efi-x64-cdboot seems important tooYeah
I will assume this is all what is needed?
really dont know how uefi works tbf
yopu dont need to install that inside the container
grub doenst need to be there
o
lemme get it
this
Oh okk
k i think will stay away from grub/uefi in general, cant go blind
This will do for bootc at runtime?
huh probably?
seems like it would
just do it lets test if it works :P
ya, probably later will hit bed
Im starting to malfunction
which timezone are you on
UTC+1?
I think +2?
whatever Spain drops in
LOL pog
oh no is +1
ok so
making the squashfs with the runtimes included
seems a lot bigger
it didnt work before LOL
we were under 2 GB :v
😭
Why does Podman never cooperate when I want to be involved in the cool stuff...
Networking in rootful containers is broken
There is still room
maybe with --networking=host ?
oh no, --network
Oh, that works.
But why does normal podman not?
But why does normal podman not?
¯\_(ツ)_/¯
computers
4gigs yall
with flatpaks
and the container
thats a fat one now
What's the old ISO + flatpaks size? The new one is still smaller, right?
and it works
flatpaked firefox
hell yeah, Kiosk iso
goated. with the sauce
If you mean the ones currently in the website seems like 5 gb
this is an improvement in every way
An we were worried of triplicating the size 😂
crazy right
this is awesome
.
ill test if uefi works now
and if it does then we just need the installer and we'll be done
gonna pr the tmp symlink and hit bed
thanks!
ill check it out
GitHub
chore(rootfs): Make /var/tmp a symlink pointing to /tmp by Zeglius ...
This is required in order to let bootc run at runtime without podman
Symlinks don't work on FAT32 file systems
we aint on a fat32 filesystem tho
that thing is just symlinking
/var/tmp to /tmp on the host
probably will work i dont knowI was testing out another way to do that
sgtm too
lets try both approaches
also @ledif pushed some important stuff, try pulling the changes
the flatpaks 1 billion percent just work
I made a bunch of changes to sudo stuff to get it to run in a GitHub action. Seems to work https://github.com/winblues/blue95/actions/runs/13981606917/job/39147809850
pr em!
we want to make titanoboa into an action too
we'll need to do a big call to action for testing these ISOs
im sure they'll break in weird ways
You are right. Iso9660 with rock ridge extensions support symbolic links
the image is formatted with squashfs tho
im pretty sure squashfs supports that..?
RAM usage seems very little
would someone mind making a
yai repo on ublue-os?
for the GUI installerDone!
POG
the installer just needs to:
- list the disks
- blast a single disk with bootc install running as root
- copy over the flatpaks to the right place on the bootc rootfs mount
- create a user or whatever i dont know
done
Did we decide we're simply not supporting dual boot for now, with the exception of multiple disks?
we can support it on the installer later
bootc install to-filesystem
:)
for now lets just blast the disk cuz its easierCool! Because as much as I want a nice installer, it would be best to not leave Bazzite on their own.
But yeah, another time would be fine
yup!!!
it should just work
and we can actually improve the god damn installer
instead of having to deal with cryptic issues in anaconda
we can actually like point users towards offline documentation in the iso itself
we can do so much stuff
If we can make this non-ublue specific, that would be awesome! I would love to use the new installer for my images
absolutely
lets do it.
titanoboa the best.
And even add it to BIB if RH likes it 🙂
its incredibly hacky right now tho LOL
It's two days old! Of course it's going to be a mess!
we can and will improve on it with time
two days old but years ahead of anaconda already
RH have some explaining to do!
we'd have to wait literally years till we got something as good as this coming from the guys
the ISO being as big as the ones we have right now with all this added functionality is just... awesome.
I guess we could easily adapt it to be a netinstaller too, but with proper retries
absolutely
we can do like a super minimal one if we wanted to
with like alpine as the base or some shit
:)
that just has that cool tui installer, done
offline bluefin installer in 2.0GB would actually be a reality LOL
we'll actually be close to being done now :P
Where was that gnome first run install wizard thing?
gnome-initial-setup
if you can be a huge gamer and figure out how we can wire up the installer to that it would be awesome.
Hmm. Does Bluefin have that?
Sorry, was at a meetup tonight.
yup, by default
Looks like we are making some serious progress though!
we are done with the ISO pretty much
just UEFI support is missing
cuz we need to figure that out
Sorry, I might be being stupid
oh that thing i think its on
/usr/libexec...?
i dont remember but its kinda hidden
only runs specifically when you dont have any users and gdm runs
the progress for today:
- ISO now is installable, you can run bootc install and itll work
- flatpaks in-iso work
- some fixes on the justfile
we just need the installer to glue it all together nowOh, that's easy.
LMFAO done
NICE
On a plane
Awesome!
Make it an action so we can drop a list of images and then it just generates and pushes to r2
im still impressed flatpak and brew works on this
So yeah, only thing needed is UEFI and a (G|T)UI.
Awesome work Tulip, Zeglius and ledif! (and others)
blue95 is gonna look so awesome with the live ISO
So if I wanted gnome to create the users as part of that first time setup, how could I do that?
I mean how do I just get those pages to appear?
Amazing
we'd just do it in the installer directly
we dont need to do the users step on GNOME cuz initial-setup is going to do it for us
on first boot
Yep
but kde and stuff is gonna need that
YO we can enable fsverity by default btw
Can we do a readonly root in the ISO?
yup
but like we can do fsverity on the target
via bootc install
This looks like quite a cool project
https://gitlab.gnome.org/p3732/os-installer
It allows us to drop in scripts for installation, and has pages for disk selection, internet and languages
Oh. It's 2am.
See y'all tomorrow
See y'all tomorrow
fancy!!!
We talked about this before
we could try using it
uefi support done
👍
boots up and works nicely
We're getting rid of like our worse external dep
Amazing
absolutely
Now we just gotta get rid of fedora lol
How did I fat finger a voice message on the train 😭
Centrora
this is making this easy
lets just wait till bootc supports the OCI backend and we can move to wolfi LOL
ok so we need fuse-overlayfs just so that bootc knows that the image exists in containers-storage
We can move... If we start packaging GNOME ourselves.
the install doesnt rely on it
we are gonna need to do that anyways
to backport GNOME 48 to Bluefin LTS
and wolfi has glibc and systemd, everything should compile nicely
had a worse time packaging GNOME 47 for freebsd
man this installation is lightning fast
bootc install done
ill just copy the flatpaks now and see what happens
ok getting flatpaks installed is a bit harder
Let's get fedora working
should work just the same
ill test it out
Uefi is working? You animal.
yup its working perfect now
didnt test secureboot :P
Everything committed?
yup
testing zeglius's PRs
How are you awake? Isn't it super late for you? Lol
its midnight
:P
Ahhh, not too bad then.
too late 4 you?
I'm old.
And also have kids that wake me up early.
So what is the RAM requirements for the live ISO to do an install? Do we duplicate the image or is it pulling from the image on the ISO itself?
i am testing it on a 4GB ram VM
it should work on any system
I just know you're loading the image for the install into RAM right?
no its a bit weird
launched a few flatpaks n stuff
and brew is getting installed onto ram
Whacky.
brew works
like the storage is all in ram
but bootc install gets it straight from disk
gotcha, so for a bigger image like Bazzite, more than 4GB of RAM is going to be needed to do an install.
about that
we need to test it
or to run everything unless we trim a lot of fat from the image.
yup we can do that too
there are a bunch of services that start.
@Kyle Gospo might be worth looking into what services we could trim or just mask for a live Bazzite ISO.
yea LOL
we can just make like a
bazzite-live image
that is trimmed down bazzitethat's my thinking.
still have the branding, but include like the bare minimum stuff for an install ISO.
yup!
I suppose we might run into the issue that it's not installing the actual image though.
it will tho
What the god damn fuck? Am I getting flatpaks right now? lol
the stuff in
/var/lib/containers isnt necessarily the same as /
yup.
flatpaks workonly 77 flatpaks to install for Bazzite 😢
its not the same as bazzite, im just getting GNOME flatpaks from bluefin LTS LOL
I mean, I expected this. As we all know, Bazzite is bloated 😉
alright so
after a full end-to-end test
it works
everything works
UEFI, BIOS, and installing works perfectly
even secure boot? 😉
flatpaks work, brew works
havent tested fedora yet
i can see it working
bazzite has a custom kernel, so I could see issues there.
we can just use the regular ol' fedora kernel on the ISO
we can do whatever we want LOL
right, we have so much freedom.
So I love the fact that we can have on screen keyboard too
we can have gamepad support too LOL
and so much more
OH FUCK
holy shit ok ok ok
accessibility stuff, like voice controls
so.
yeah we can do a lot
what if we could do the bootc install from game mode?
can be goated but i wont work on this right now LOL
i wonder if you could do something similar to that steamOS recovery image
hahaha I know I know.
yeah, there are ideas for this.
custom partitioning can just be done using gdisk or the gnome disks tool
we just need bootc install to target a partition.
for bazzite the resulting ISO is 6.8GB
should be a lot smaller now
that is a HUGE difference.
got this install via titanoboa
uh, which grub option should I be choosing?
first one
these are jsut for debugging purposes LOL
we are gonna remove that
soon™️
ok ok ok ok ok.
so
one thought.
powerwash feature
I saw a podman command you can run to nuke a whole disk and start from scratch.
fuck
system-reinstall-bootc
its already a thing :)did I have to create a user?
show a user?
oh i dont think you need to create it on the live ISO
for KDE image I do 🙂
it should just work...?
livecd should be the user i think
we are using the li-
OH I KNOW
it tried livecd: livecd
you forgor to add the damn user to the thing
just build (image) 1
livesys
@Noel
Coming along
just run these sequentially
HOW ARE YOU MAKING THIS THIS LOOKS AWESOME
RIP I started the build over with just build (image) 1
fucking lol
It's a Python script using GTK3 https://github.com/winblues/blue95/blob/main/files/system/usr/libexec/blue95-installer.py
remember, you gotta run
bootc install to-disk (disk) --source-imgref (image) to get it going
needs to be as run root
awesome!
you can run pkexec cuz it doesnt have a password anywaysit annoys me to no end that KDE doesn't have a first start wizard.
yup...
i hate that too
I'm going to tell Neal I'm going to switch to Gnome unless he makes a first boot install thingy
Damn this is amazing
also since it's late, look how fluffy my cat is:
I know, the ISO is 6.8GB now Kyle.
You madmen did it
how big was the ISO before tho
we can cut down on the ISO weight a lot
we have like lutris on the damn live ISO
lol don't look at me. @tulip🌷 you absolute madwoman
yeah, the main things we probably want to do Kyle is stop a lot of the auto starting services.
the thing boots, but last time I booted it, it filled my entire temporary disk lol
That should be trivial to do
yea it tries to get steam going LOL
I'm not sure what field your disc but immediate glance steam comes to mind as something that needs to not auto start
its steam
Or perhaps even just not be present
it downloads everything into ram LOL
yeah
Like just nuke the auto start and the desktop file
lets make it as shrimple as possible
yeah, we could probably do a really light image and then store the image somewhere else.
the output will be the same anyways
Just wanna make sure it's themed 😄
we can do grub themes on the ISO too
if you guys have any idea it would be really cool
dracut can also have a few themes
Things that come to mind immediately are:
1. on screen keyboard
2. powerwash (or recovery features)
3. custom partitioning will need a guide written for it for dual booting.
4. smaller base (maybe just generic installer image)
we'll also be able to get offline documentation in the ISO itself
Since we don't run steam, number one will need to be re-enabled only for the live ISO
On gnome this is as easy as removing one of the extensions we ship
only thing missing now is figuring out how to dump the flatpaks inside the image
no idea how to do that in a way that it wont blow up
Oh and kill yafti
right!
@tulip🌷 so what is the manual process to try and install?
minus the flatpaks, don't really care about that at the present moment.
just want to try and install bazzite using the bootc install command.
pkexec bootc install to-disk /dev/vda --source-imgref ghcr.io/ublue-os/bazzite:whatever this might be helpful for flatpaks btw: https://github.com/JasonN3/build-container-installer/blob/main/flatpaks/Makefile
GitHub
build-container-installer/flatpaks/Makefile at main · JasonN3/build...
Creates an ISO for installing a container image as an OS - JasonN3/build-container-installer
we just gotta copy them over to the bootc installed system
we already have it on the ISO at least LOL
the disk is still writeable after bootc install right?
yup!
its just that trying to copy it to var inside of a boot thing breaks it
OH I see.
it just thinks /var/lib exists and makes it so no service can edit that
then BAM everything breaks on boot
I wonder how anaconda is doing it then.
got an error
oh
typo
pkexec bootc install to-disk /dev/vda --source-imgref containers-storage:ghcr.io/ublue-os/bazzite:whatever
you need
containers-storage:
before the ghcr.
pkexec rm -rf /var/tmp
pkexec ln -s /tmp /var/tmp
then do the same thing againlol
do I need to specify btrfs?
maybe...?
--filesystem btrfsthere it goes
Live ISO is...
so fucking good
its awesome.
its so fast too
pretty slick.
just gotta figure out flatpaks, I'm sure another breakthrough is coming soon.
yup LOL
this project is literally just 2 days old too
bluefin:gts doean't work due to lack of dnf5 i think. Testing bluefin:stable nowprobably something on the livesys scripts
you can just skip that
I'll check
I have successfully installed bazzite, it's booting now.
lol uh oh
so
no user got created.
it did boot though!
Try the gnome image
that won't work either.
Bazzite disables the first startup gnome thingy.
I'm sure we can include something in the actual installer to create a user.
that shouldn't be hard.
I'm going to try and just create a user using recovery mode.
oh weird
i guess gts will just not work
we need dnf going
maybe you can force it to run dnf5?
But luckily we are going to shift up a fedora version in a couple weeks?? so maybe we dont care?
only units that failed were these 2:
Yeah gts will rev don't sweat it
the install succeeded with no flatpaks.
also we gotta figure out how to set a timezone.
interesting.
nah probably going to be fine
the fedora flatpaks repo is on the image.
hm i wonder why
lol uhhhhhhhhhh
oh
hold on.
I know probably why
Alright, I'm tapping out for the night.
Bazzite does install successfully.
List of things needed for KDE specific images:
1. Creation of User
2. Set Timezone
interesting:
Don't know why I recorded this but here's 20 minutes of me installing Blue95 from the live image
more importantly, when am I getting Blue95-DX?
or Blue95 Gaming Edition?
With Borland and Visual Basic pre-installed?
bluefin:stable is workingi was about to sleep
but i found out how to get flatpaks working
sudo umount /dev/vda3 (unmount the thing that bootc mounts)
then mount /dev/vda3 (whatever), mkdir -p (mountpoint)/ostree/deploy/default/var/lib && cp -f /var/lib/flatpak (mountpoint)/ostree/deploy/default/var/lib
nvm it brokeThat or disable the services (like brew setup) that we don't need at runtime
4AMMan the dedication is strong i love it
Maybe bind mounting to the livecd /var/lib/flatpak should do
Maybe in an unshare mount namespace? Probably I'm giving it too much thinking
https://bootc-dev.github.io/bootc//bootc-install.html
This recommends that install to-disk is only for demo/PoC
But it looks like there is a config file we can setup for install to-disk
But there are luks options already apart of install to-disk
Yeah I'm adding the suggested toml file to the live image
Mmm I think that's within the scope of the installer itself, not the liveCD
Besides, we were about to use
bootc install to-filesystem regardless, right?What is our plan for the installer? I'm working on a branch that adds a simple
titanoboa-install script that does a bootc install to-disk with the appropriate flags. Do we not want to include something like that?Also that won't work if it is bluefin-lts. They don't have btrfs in the kernel
Probably is better if we leave that to the installer, so we about cluttering the livecd repo with edge cases
to-disk is meant for demos
By installer do we mean yai?
I can see them not supporting it on the long term
Yes
Hmm okay. So I suppose I could instead move my installer stuff to my own repo I suppose
Also no, titanoboa should be meant to just create an iso
I don't want to end up it being too tied to the installer itself. Lorax made that mistake with anaconda
I would also want to avoid edge cases as much as possible. This should be something that benefits all bootc images from, as a template of sorts
Hell, if playtron wants to use it so be it, they really need a fancy installer as well :v
to-disk is just a wrapper for to-filesystem and to-filessystem is what bib uses. so if and when we get where we need want to support repartitioning we just unravel to-disk to be to-filesystem?
Why is the btrfs thing an issue? It should blast any fs no problem
We probably should ask Colin about what to-disk lifespan is about to be. It would be nice if we had it just for a sane default install, no configuration files needed
Is not. Just saying this is something the installer itself should specify
The upstream GitHub has discussions turned on ask there I know he reads those
Gonna drop there and ask later on
K I'm afk for a while still
I am in airport hell lol
No prob, it seems we are dealing with this quite smoothly
Yeah this is awesome, love how you executed your vision
You even didn't let noel explode the scope, well done. 😈
So from what I understand now, the scope of titanoboa is to just create live ISOs and we won't add any stuff related to actually installing to disk. But I suppose we should include hooks somehow to include an installer when running
just build? Or should the images that we are using as the source be augmented to include an installer? 🤔I thought the installer is yai
A separate program but included on the iso?
Ah, that's what I was missing. We'll include yai by default in the ISOs when it's ready
Should be as easy as a just recipe like
rootfs-include-installer and a podman run with dnf install installerMakes sense to me now!
Titanboa = livecd creator
Yai = Yet another Installer that uses bootc install under the hood.
Titanboa should be comparable to something like archiso. A tool/framework for building a live ISO.
Basically yeah
Lol so we got the names backwards
Oh well
You think I've given up? 😉
This was my understanding.
Basically. This should have been the lorax killer, not anaconda one
Titanoboa could still consume anaconda web UI 😈
The whole thing is stitched together without no semblance of separation, so might as well be called the same
Titanoboas we're huge let's just swallow everything
Also fun thing I learned is Red Hat is looking at deprecating squashfs in favor of erofs
Yeah same with snaps
Not surprising as they called out supporting Flatpaks in RHEL 10
Too late lol
Read a bit about erofs, can see why
Oh they are about to swap to erofs too with F42
https://fedoraproject.org/wiki/Changes/EROFSforLiveMedia
Maybe we should consider doing that as well
Strong push for erofs over squashfs
Might as well
Hopefully should just be drop in replacement for the commands?
I believe so.
Quick search and seems this is all. Gonna start from there and see
Well yeah that and mkfs.erofs
I got to say I'm enjoying the Justfile being used to organize all of the scripts.
It's basically a directory all tied together. And if we ever decide to use Python or another scripting language it can run it via the shebang.
Downsides are that you lose out on some of the basic linting editors can provide you since it's a justfile
Would say copy and paste in a separate file back and forth. Though jetbrains editors have this neat nested LSP thing
That nvm works half the times
erofs should be a super quick replacement
i was using that at the start too
You kept that branch?
nope, its just replacing the
mksquashfs part with mkfs.erofsWait that's it?
What about this one?
should work even without it afaik
Ooooh gonna try then
I'm giving a talk at a conference tomorrow on how to grow your involvment in FOSS. Decided to throw this project in as an example of how fun and exciting collaborative coding can be
im so glad you are liking all this stuff. this is awesome i hope it inspires more peeps to be on FOSS LOL
also
a big optimization i was thinking about
maybe we can literally just remove the entirety of
/sysroot/ostree*?
i believe there arent like symlinks there or anythingIf it's already erofs then there's no worry right?
probably not.......
i have no idea
Except you lose fs-verity or whatever
no like
we are doing
bootc install from the containers storage
the container itself already has /sysroot
the ISO / doesnt... need... that....
this would remove like 2gigs off of the ISO LOLAh right. Isn't sysroot and the rest of the rootfs deduped? Or it really is doubling the size?
to be completely honest i dont know
we arent like using the bootc features at all on the ISO except for bootc install
Yai should give and option do pull the latest image/a different image
tbh i think right now we've got to just make it as shrimple as possible
YEAH @ledif what about the cool installer
That's like a step 3 problem but should be easy enough
hows it looking
Prototype GUI for
bootc install in Blue95. It's mocked with sleep right now but I think once I hook in bootc install it should be rollingthis is so cool 😭
i wish bootc install had a json output....
we can PR bootc with that.
filing an issue
you can do a polkit rule to make
bootc install not ask for a password
we probably want that directly on titanoboaGitHub
JSON/parseable output for
bootc install * · Issue #1217 · bootc-d...This is more of an idea than anything else, but I believe it would be rather interesting to see something like that --progress-fd flag for bootc switch, but with the bootc install commands. That wo...
Nice! Gotta get back to work. I've been too distracted by this all morning 😆
Isn't sysroot the real root of the running system?
for this thing i dont think so
the ISO
/ is just... /When you open a terminal, does it say you are not booted running with ostree?
yup
not booted with ostree
Then yeah sysroot is not needed I believe and it's using the rootfs directly instead of hard links out of sysroot/ostree/
yup
ill test that out
Since you are already messing with the rootfs in the justfile.... Just rm -rf sysroot
yeah LOL
doing that
laptop is in its last legs, so slow to build the iso 🫠
all what needed is this right?
yup!!!
the compression is worse :(
i was trying to get high compression like zstd and stuff but it didnt help much
Maybe we just copy them and done
Eh whatever, gonna pr and start with the tui installer
Seems to boot at least
@j0rge something like this, right?
looks goated
just do whatever works tho
eh good ol tooling
do we have dialog on the host tho...
oh wait gum right?
dunno like
we can use gum, dialog, whatever
I mean can be installed with brew so :v
dialog will just most likely need a static binary
Yeah we can swap it out for gum
Gum is on the image already isn't it
gum choose disk -> bootc install blast it all
hell yeah
alright,
bootc install to-disk /whatever/drive shouldnt be too much of an issueHave any of you tried the luks option?
M2
You think zfs would work?
we'd need to the install process manually
but it probably would work
zfs isnt an option for to-disk --filesystem i think yeah
Heh
Yeah could be a possibility with to-filesystem. My reading is that to-disk has a few presets for blasting a disk. To filesystem let's you prepare a disk first
yup!
I think to filesystem would be good because then you could support dual booting pretty easily.
luks works you have to set the password after the install
we can do it on the installer
we can support both just fine
right the installer would call the to-filesystem option.
both works too
yup!!
we aren't marrying ourselves to bootc install to-disk right now
you can set it via cryptsetup i think
on the live ISO
I would hesitate to make a complicated disk formatter in the installer
so the install can do it
yeah I'll try it. I'm about to be on a 12 hour train ride so I got time lol
I would say the installer only provides an auto formatter or use existing layout options.
Then it's on the user to figure out how to partition the disk themselves
Advanced options for advanced usage. Make the user use tools like Gnome Disks to do their own partitions and stuff.
having the option for an advanced I know what I'm doing user would be helpful.
blasting the whole disk should be the default though, I do agree.
I'm saying the advanced option should be a bring your own.
Not a open and guide a user
right
right, I'm saying have the capability of a button to blast to a filesystem rather than blast the entire disk if that makes sense.
Yeah. I think we are saying the same thing
we can certainly have a written guide for how use gnome disks to create partitions or whatever on a case by case basis.
Bazzite would definitely need that since a lot of users try bazzite first on their handheld and then move fully over.
noel's making me cry again
I'm not saying it has to be done for Bluefin or anyone else lol
I don't think that is a priority. Anytime there is a guide from us, it is an implicit support requirement
@bsherman I need an assassination.
Can we have a MOSCOW issue for these things?
Then we can figure out priorities from there Like what MUST be on the installer for an MVP Then what SHOULD be there for rolling it out wider What COULD be added to increase support for other projects What WONT be the responsibility of the installer
Then we can figure out priorities from there Like what MUST be on the installer for an MVP Then what SHOULD be there for rolling it out wider What COULD be added to increase support for other projects What WONT be the responsibility of the installer
noel: remember the tale of ubuntu core?
for the MVP, we don't need to support dual booting.
I'm expressing it as a wishlist item.
hell, I don't dual boot myself, I'm asking on behalf of other nerds lol
check the readme. 😄
I think some of the issues need to be moved to YAI
it's very barebones
Titanoboa's responsibility is creating a live ISO and is a toolkit for running an installer on.
in my mind anyway
@tulip🌷 by any chance do you have an single-line for bootc install?
pkexec bootc install to-disk disk --source-imgref containers-storage:image
ghcr.io blah
Leg day. At gym
you'll also need --filesystem
depends if the base image has support for that
like
bluefin lts just works, fedora for some reason doesnt
interesting.,
does lts have btrfs support?
nope not yet
then xfs as default
we can just make like a lil file
on the container or something
that tells the installer what is supported or not
there is the bootc to-disk config file
put btrfs as default we can figure out the other stuff later
we probably want that :p
so should put bluefin:latest by default?
i dunno
Yeah ledif idea btw
though that file should be set in the base image
rather than us manually
ideally ¯\_(ツ)_/¯
We are modifying the rootfs so injecting the file as part of the livecd create is fine in my eye.
Could probably determine xfs vs Btrfs via what modules are available in /usr/lib/modules
Stable, we don't expose latest to people
i think we can straight up read the kernel config
I'm attempting a live boot from ventoy
NICR
NICE
did it work?
seems to not have worked LOL
its not finding the /LiveOS folder
Maybe I can try the grub2 mode I did normal boot
probably what you want
Mmm okay we need to figure out erofs compression
I think the iso went up to 15GB
Copy the flags from the change proposal.
Yep doing that rn, let's see how it goes
Erofs is definitely aimed not at compression by default
this is considerably slower than what we had before
its interesting tho
Yup
I think it froze
mine froze then BAM 3gb
YO
its much better compression wow
we can compress the grub files too
It's slower to compress. But it should have better performance when being used.
3.4G is tiny
absolutely
this is crazy cool
i dont care how long it takes when we are doing this in gh actions
we could put like lzma level 15 or some crazy shit like that
I would not go too crazy. Again performance is the main goal for erofs
true
And the meta data isn't compressed
having the binary data be compressed i think is more than enough LOL
im thinking we should do zstd compression
like we have by default on btrfs
lzma seems to be very cpu intensive at runtime
impressed about how small the ISO got tho
that one surely takes its time
Lzma should have low latency at run time
it works but disk performance is awfully slow
no idea why
Right. I'm here to help. Have we sorted flatpaks?
like it feels like im running from a USB stick, even off of my NVME ssd
we just need to figure out how to copy them to the installed system
copying it over to
/ostree/deploy/default/var breaks first-bootbake them into the image
Not a bad option. We would have a dedicated installer image created FROM the existing image
What is the current installer doing for flatpaks?
nothing 😭
nah, directly into the base image
No the one we currently ship
oooh
i dont know
we probably want to figure that out
I wouldn't be surprised if it's using the host as local remote and installing them from there
besides, now you would need to rebase at runtime, and make the logic more convoluted
wait how
probably systemd-tmpfiles gets mad that
/var/lib exists or something i don kno
need to debug thatThat shouldn't make it mad....
this is probably like a really dumb thing that can be fixed easily but i dont know
Stuff it in there
it just makes a lot of stuff not work
Well I have a test subject in shipment for the new isos on real hardware in the future
ill backup all my stuff and test it on mine too
Well I ordered a used laptop, you’re testing on your real machine :huh: 😭
2 based 4 you
:clueless:
zstd compression on erofs crashes QEMU
jsut straight up
So lzma
ill test all of them
lzma is just too slow at level 6 at least
I wonder if it’s accessing some cpu instructions that are not provided?
What compression does squashfs use by default?
im doing host CPU passthrough so it should be fine
none....
Oh then you’re cooked
Oh
no idea how squashfs gets such good compression
the
-Eall-fragments,fragdedupe=inode -C1048576 flags for erofs help a LOT thoSquashfs uses gzip by default
oh nice
didnt know
Otherwise your iso might be 20 gigs
yeah its like level9 gzip
That's why by default it's pretty good at compression. Gzip is king for small size
oh my god thats why lzma sucks
(maybe)
A
this is from the squashfs tools
--help
lvl9 lz4hc gives me a 4.6GB image from bluefin LTS
4.8GB iso and it seems to work
somewhat
its pretty fast, doesnt fuck up the disk performance like lzmaFor erofs?
Then let's use that
yup
erofs, lz4hc level 9
im testing level 12
mkfs.erofs --all-root -zlz4hc,12 -Eall-fragments,fragdedupe=inode -C1048576 /app/{{ workdir }}/squashfs.img /rootfs
maybe we can just use gzip on erofs?
Mm does it work? I don't have gzip listed in my --help
no idea
higher level for lz4 doesnt affect file size
9 and 12 are 4.8GB (bluefin LTS)
It works! (I walked away and the screen went to sleep but when I came back it was installed and booted up)
NICE
this is so awesome 😭
we literally just need to figure out flatpaks
heh
Gotta put 90s sitcom references in a 90s bootc image
i wonder how much time it actually took to install
Last time I did it it was around 20ish minutes I think
Mostly stuck on that fetching layers part
is the blue95 image that big?
It was closer to 4.5GB before we added flatpaks
you include a lot of stuff in it then wow
but thats cool LOL
Flatpak do be beefy
Yeah, it should be "blue95-dx" but I wanted the dx to be default
Basically same packages as aurora-dx for the most part
noice
Yeah blue95 is something I want to install first
It looks so cool
so
pretty damn quick install
at least on NVME storage
feels just like an update
Pr that thing👀
its just lz4hc compression
pr yours :P im gonna be away for a bit
This right?
exactly that
set it to level 9
cuz 12 doesnt improve it
i wish we could use zstd
K down to 5gb
Solved
for which image ?
Bluefin LTS probably?
if you said bazzite id ship it tomorrow
I'll check Bazzite.
It was 6GB last I checked.
Erofs might make it smaller.
deck image is 10gb now
I'll try the deck image. It likely won't boot in a VM very well though.
dont need to boot it, jsut tell me the iso size
bluefin-lts
We broke through his shell, boys :clueless:
@j0rge
I realized the last image I built didn't include bazzite's flatpaks, it included the default list of flatpaks for bluefin.
We got just the iso, not the installer itself btw
I don't care if you can do 6gb I'm sold
We can even work through the secure boot issue
But you need to include flatpaks
Noted @Zeglius
all that needs to be figured out with flatpaks is getting them copied onto the actual installed image.
bootc install grabs the image it's built with and deploys it.
@antheas you have any ideas around that, happy to take them.
For some context regarding flatpaks
No, it needs to be 6gb with flatpaks
Otherwise no dice
checking now.
For secureboot, sign grub with the mok key and it will throw you to mok manager
Problem solved, and one less thing to worry about
I'll see where the base bazzite desktop image ends up.
check the deck one after.
Alright script finished installing. Rebooting 🤞
Yep
Yep, seems like a fully functional install
what does it say when you do
rpm-ostree status?I'm trying to dig through the original installer to see how it handles flatpaks... And I still have no idea...
This seems like the commit introducing the functionality (there may be a few fixups afterwards)
https://github.com/JasonN3/build-container-installer/commit/bfa150ceebf096e1e23c652a1e437a98ed4f6085
Does it just hand things off to lorax? Is that why I can't find anything that makes sense?
when I looked last, it doesn't show the install as installing a signed image.
or ostree verified
hands things off to anaconda.
anaconda does the actual installation, lorax makes the packages available on the ISO.
I don't think that's a huge deal. Things are slowly moving away from rpm-ostree anyway. And it's probably something we can fixup in bootc if we want it sooner
Probably need to add the sigpolicy flag
Like with bootc switch
holy shit this output is verbose.
any way we can quiet this down?
2>/dev/null
Or maybe has a quiet flag
I mean, I don't want to eliminate all output lol
Nope, no quiet flag in sight
Bazzite Live ISO with all flatpaks is 8.1GB
we should try the sigpolicy flag
If you bootc update --enforce-container-sigpolicy, does it show verified after reboot?
Lemme try
Oh. Sorry, it's a switch flag. Probably won't help
is there a sigpolicy flag for install?
Yes. It's the same flag as switch
Ah yeah there is
Oh actually it seems doesn't even need to download the image, will simply reuse the current one if it is new enough
On the live ISO. Is /sysroot still present?
Oh right
We can purge that
After all we don't even use ostree
Or rather, the contents there
sysroot is still there.
what about the size
is the compression for squashfs better?
Noel in the justfile when making the rootfs, remove sysroot. It will probably halve the size of the iso
squashfs ootb is better, with erofs needs to be specified, but is close
wait
Though unsure if it will actually save space if it's all hardlinks
OH BOY
thats were all the fat is at
so wait, what should I be removing?
It's mostly hardlinks in there....
So it's fake savings.
On the container I'm messing with it still saved over 10%
Also this image I'm messing with may not of been rechunked.....
Yeah if the image is rechunked it won't be anything
Yepp. Literally neglibale savings for the rechunked image
Are you on zeglius pr? With the erofs compression?
I'm on the master branch.
erofs changes got merged.
somehow the image is bigger than our current ISO for Bazzite 😢
Well yeah
The flatpaks
Wait anaconda does include the flatpaks, right?
Or are these only installed at runtime
Anaconda installs our flatpaks
Alright so got an idea
@antheas do we really need it to be a bazzite liveCD, or can be a dummy image with an installer baked in?
IMO, it doesn't need to be a Bazzite LiveCD image.
Right, so all we need is a vanilla image and that's it
currently what we are doing now with anaconda is a very minimal startup image that runs anaconda and has the OCI dir and flatpaks deployed through using Anaconda.
Reason we would want a live CD is so we can support onscreen keyboards.
since anaconda doesn't do that.
Right, and lorax is not as maleable as what we have here
not particuarly no. There is a way to make a live ISO using Lorax, but we haven't tested it and it would make the ISO even bigger while still having anaconda.
Alright, so what we can do is, once this is mature enough (which should be after we manage to install flatpaks) we can make a fork of this for bazzite were we modify the recipe to install it's kernel and stuff in the initramfs with a dummy image and whatever Antheas wants to use as installer
and include the OCI Archive (or whatever bootc install takes as a payload) and flatpaks?
So you mean like anaconda?
Nah, we just do as rn, we dump the bazzite image in the rootfs container storage of the dummy image
But yeah other than that, I can't see how else we can fullfil that <6GB iso requirement
Wait
These hard links at sysroot are pointing to the files at the rootfs, correct?
Wait nvm should be
Yeah.
Agh
So futher deduplication isn't possible?
Yeah. Need a smaller image
Or maybe we can just take the flatpaks and brute force compress them in a tar
Nah. Probably ain't possible, besides that's beyond the scope, would need to be it's entire separate thing
Asking for a 6gb iso for bazzite ain't loable with the current setup without ending up mimicking anaconda
working > size for mvp
Yeah besides we really need to start giving Niklas a hand with the installer, once we figure out flatpaks
@antheas idk ask Kyle to cut down some fat there or ain't possible :v
if you gimme some rust code, I'll do the rest wiring it up in the UI
Not learning rust has come to bite me back uh
well yeah, i know nothing 💀
In my current stste I feel I can only oversee tulip so she doesn't go crazy with DBus and crazy overengineering shenanigans 🫠
Aka not much
Shouldn't be too difficult? Just maybe some
Command::new().arg() calls and done maybe even someone as me can help somehowyeah dont worry, i think will tulip help out
agh, called it
Why is there a limit on 6 GB?
Idk ask him, that was his requirement
Oh I don't give a shit
Literally I give less of a fuck about that than the poop I just flushed down my toilet
Don't even worry about it, just focus on cooking
Whatever we do, he probably will be more comfortable with the embedded initramfs he was planning
Yes, that will be a great option for handhelds and for pre-installation on something like the framework desktop
But the live ISO is its own thing with its own benefits
How big are the live ISOs now?
For bazzite-gnome/gnome-deck
Noel — 19:49 Bazzite Live ISO with all flatpaks is 8.1GBDon't know if that's gnome/deck
@Noel which bazzite?
The current gnome deck ISO is 8.8gb
So that means we took 700mb off and gained a live environment?
Eh, you won't have a 1:1 functionality, we need to disable a lot of stuff at runtime like steam autostart and the hardware script thing
But pretty much yeah
Yeah that I'm aware of
That is fucking awesome
Nice job all
Let's see if we can get a fancy installer now
And this one at least will tell what errors are there :v
That and the flatpak installation
I'll be honest here, I don't quite understand why a live ISO of our final images is important for us...
I understand the want for a DIY installer - to resolve all the install issues.
What would we lose out on if we created a live ISO from silverblue-main for example, and embedding the Bazzite image OCIs onto that smaller base? Would this increase or reduce the ISO size?
I understand the want for a DIY installer - to resolve all the install issues.
What would we lose out on if we created a live ISO from silverblue-main for example, and embedding the Bazzite image OCIs onto that smaller base? Would this increase or reduce the ISO size?
KDE
It's actually bigger.
Live ISO can support on screen keyboards and the like.
Also be useful for recovery purposes.
We don't lose anything per se. But I'm aiming for this to be multipurpose
Ah, understood
So 9.5gb to 8.1
So we can have dumb stuff like kiosk live Cds
That's insane
Not the deck image.
Yeah the desktop one is 9.5
Bazzite KDE
Oh hold on, that seems sus
Are you sure? The deck one is that big.
Bazzite now on your grocery store kiosk
That's 7.9
It grew in that case
Right.
It should have done the opposite
You sure?
I'm gonna guess it's erofs
Yeah instead of windows blue screens, it will be systemd ones
Squash made it smaller if I recall correctly.
Though I had the wrong Flatpak list, but it was more Flatpaks rather than less.
, would love a scientific retest later on
But I'll handle that
@Robert we can do usb sticks now, for like conferences etc, so people can try it
Besides fedora is switching to erofs for their livecds
Should just work, I can give you the command I ran.
Right.
So probably we will want to stick to that
though ya'll had the live thing working before we could even have a discussion on which one we wanted so might as well roll with it.
Plus it's nice for users to be able to test their hardware ootb, especially nvidia
Gotta warn about that, behavior is not 1:1. Per example, you can use dnf in the livecd, as the entire root filesystem is writable with the overlay
By now
nuts
but also still fine
And Nvidia probably wouldn't work in the live ISO, would it? The UBlue SB signing keys won't be enrolled, so the kernel modules won't load?
Yeah just make sure of having that as a disclaimer so ppl doesn't end up with deceiving expectations of how an installation behaves
I'm not too picky, it needs to be more reliable than anaconda, that's all we need
If the errors would be more descriptive, I will take that in a heartbeat.
Antheas mentioned something about secure boot lemme check
Colin made a PR to expose errors in the installer, but it had stalled
https://github.com/rhinstaller/anaconda/pull/6019
GitHub
payload/rpm-ostree: Copy stderr to our output by cgwalters · Pull R...
The rpm-ostree container deployment path can fail for many reasons from networking to details in mount setup.
What we really want is a proper API with progress out of bootc/ostree; I will work on t...
This
We just need to run mokutil as part of first boot. Not sure how anaconda does that.
Theoretically since it runs the full image in live mode, can't we just call the ujust to enroll the key?
Yes, though that means booting into live ISO, running ujust, then re-booting into live ISO to get the drivers to run in live mode.
It works, but not ideal 🙂
But I don't think there's any other way
Ah my bad
Eh
I would also be fine with a non live CD and tui installer - but I think we're past that lol
We can call efibootmgr set the livecd to be the next reboot target
But yeah, maybe the scary blue screen needs to show
I mean the tui script is there as a fallback
We can use it if if we take too long in he installer
yeah I think once we get flatpaks it'll be good enough for lts
gts has to wait untill next version of fedora anyway right?
We could make a simple TUI that the GUI wraps.
Best of both worlds.
Kinda like how Devpod has a Tauri UI wrapping the main Go TUI. All the heavy work is done via the Go CLI Not saying those tools are the things we use. But the concept
Best of both worlds.
Kinda like how Devpod has a Tauri UI wrapping the main Go TUI. All the heavy work is done via the Go CLI Not saying those tools are the things we use. But the concept
Eh
Smells like overengineering there
since lts is beta I'm fine with shipping the tui and then adding the gui as it comes along. We'd just cron a new weekly build and it'd be automated
then as the gui improves the ISOs just get better over time
The GUI will be easy enough to improve
It's html css after all
And js
Mostly just JS since it's react 😂
But the point still stands
And many people know that
The trouble with tauri is none of us know rust for the back end lol
Tulip
But shouldn't be too hard
Command::new() and whatever follows for argsyeah that's why I'm thinking we just automate slurping in the installer as niklas works on it
You might as well just run the installer directly :v
no I mean having it in r2 that way, just regen the ISO every week
but whatever, we can cross that bridge whe we get there
Otherwise the entire logic might end up completely rewritten and not progressively. After all, the script is so dumb simple that one could simply reimplement it to rust in one day
Maybe
Speaking out of my ass given how little rust I know
I mean, it's a justfile with some inputs. Could we not just have it be an action?
As far as generating the ISO?
Each project will probably have different requirements.
We can add a recipe that acts as a placeholder for whatever you want to do with the livecd root filesystem, from installing the installer rpm, to copy it from a local directory
But that's it, we won't put the titanoboa Justfile inside the iso
That's room for lunatic logic
As for that recipe, think of it as in the build.sh of the image-template
The jank is part of the experience
omg can I toot this?
lol sure, go ahead
Man wtf this looks so good:dispair:
Is there an action for this yet?
Or is it local build only?
Oop
I have a PR up for an action and I believe the plan is to make it reusable
Local build only for now.
Yeah but that only works for the current repo (titanoboa) we need to make a composite one of it
Creating a reusable action is easy enough. But not much point until we get an installer
Yeah
For now I'm gonna take a break till tomorrow, gotta get dirty with rust
❤️
GitHub
Use Mariner as the target · KyleGospo/titanoboa-mariner@6aac127
bootc installer - eats anacondas. Contribute to KyleGospo/titanoboa-mariner development by creating an account on GitHub.
🤨
Shouldn't need sudo in a container?
I guess that's for @ledif
Are you working on submariner at Microsoft @Kyle Gospo ?
That's awesome!
This is just a playground
GitHub
GitHub - KyleGospo/mariner-fedora-bootc-base
Contribute to KyleGospo/mariner-fedora-bootc-base development by creating an account on GitHub.
having some fun
for now I'll test locally
Oh.
Submariner branding is gone...
Submariner branding is gone...
quick -- any fun way I could prove I'm in a live ISO from the command line?
rm -rf /usr/bin
sudo bootc status should show error
You did ask for fun 🙂I think this is a me problem
no dracut-live because no fedora repos
and also the kernel syntax is totally different
just fixed both
cat /proc/cmdline and check it rd.live is set
got some issues to fix, but it's doing what it should
So azurelinux Bazzite in wsl when?
mariner linux desktop WHEN
you just need dracut-live cuz of the
dmsquash-live and dmsquash-live-autooverlay dracut modules
they make the live ISO possible pretty much
@ledif can i toot that ur thing is working?
i wanted to toot a cropped version of thistulili :v18: (@tulilirockz@ursal.zone)
The Universal Blue team was tired of waiting for upstream and we got it to ourselves and made a Live ISO builder for `bootc` images: Titanoboa! It's currently in a pretty alpha-ish state since we've started developing this like 3 days ago, but its already somewhat functional and Blue95 (a custom windows95-like image made by Adam Ledif) already has a functioning live installer for the image if you wanna try it out :)
github.com/winblues/blue95
blue95.neocities.org/
github.com/ublue-os/titanoboa
#bootc #fedora #linux #bluefin #blue95 #ublue #universalblue
Ursalzona
we did such an amazing job with this
i am so impressed
like, the only thing missing now is the god damn flatpaks and they can be done soon-ish if we figure out all the quirks
this is being awesome
more efficient, accessible, faster and smaller ISOs
in just a few days
i have an idea on how to get flatpaks working
maybe we can copy the flatpaks not to
/var/lib but to /var/flatpaks
then we have a unit that copies it over to /var/lib LOL after systemd-tmpfiles does it job
alrgith folks i am dumb. we can literally just fucking copy the flatpaks over to the installed system and itll work
it literally just works
my VM storage was just too small and i got out of space errorsi installed calendar straight from the live ISO
we are just done, everyone
we just need installers now
titanoboa MVP is completed, about 3 days after starting development
the installers are the big thing now :)
ill be working on yai nonstop now
for the ones writing installers: you need to figure out the mount for
/run/bootc/mounted, mount the root of that mount as /mnt/(tmpname), then copy over the flatpaks from /var/lib/flatpak into (mount)/ostree/deploy/default/var/lib/flatpak
itll literally just workGitHub
Real hardware seems to not find the LABEL=bluefin_boot label on the...
I dont know exactly what is going on but apparently dracut cannot mount LiveOS/squashfs.img since the label/location specified by the kernel command line on our grub configuration seems to just not...
ill need ur guys help with figuring this out
Mmm maybe we should try /usr/share/factory and a tmpfile at that point, moving the flatpak directory there at livecd runtime instead of bae image buildtime
So we can avoid straying from the convention
Nvm seen you solved it now
@Niklas ⚡ so I cloned yai repo. Which should I use, npm or yarn?
Oh wait there is also a bun lock file
heh i used bun with it
but you can use whatever it doesnt matter
Oh ok, bun then
theres just a tauri thing
like on the config it uses (x) package manager
just change it to ur favorite one
im a big bun fan
hey so idea/thoughts
Yeah is fast
Bun is fine
so you know how I was like, blast, reboot, then yai
but if we're in a live ISO environment
you'd be online
we could do it in one step, blast + yai before the reboot?
or am I overthinking it
mmm?
yai is the blaster tho
Yeah
OH.
YE
ok hold on
let me back up
yai + reboot + yafti
Yeah we do bootc install to-disk --wipe
Yeah
That makes more sense
couldn't it be yai + reboot but yai also does what yafti does but in one step?
You would want to include yafti in the base image though
Mmm
right
nm, yafti does a bunch of other shit
That would mess first boot I think?
right right
nevermind
Or at least services that want to use the ConditionFirstBoot
oh that can also happen
Donno
nod
like, we can have literally anything on the live ISO so we can just... do whatever at some point
ok nevermind, was just showerthinking if we needed one app or two
you can just specify the
flatpak installation to the installed system then BAM yafti but coolerok so using bluefin as an example
i personally think that having yafti be a post-install step is better, but dunno its good to know its possible
I boot into live CD
all the flatpaks are there
Oh btw, probably we should move the flatpaks installation to the installer
I click yai on the dock
select my disk, blasts, reboot
on reboot gnome first run runs
Or it just pops up
I make my user, and I go into the finished desktop
then yafti shows up
Yeah should be like that
not on bluefin, but still
well we don't have yafti now but I would add a tauri one since that seems to solve everything
yea LOL
I think we should because if we can get our users to install a ton of flatpaks on first boot then that would be great for flathub
We just spam
Command::new and we should be good, easywe can literally like embed stuff from the flathub api, if we wanted to
like if we showcased good flatpaks right up front
Btw, no DBus shenanigans with the disks :v
why doe
Lsblk should do
its gonna be kinda nasty parsing the lsblk stuff
Plus DBus sounds like a pain in the ass
its kinda yeah
its a lot more reliable tho
Well the output format won't change for a while
Plus it has a --json flag? I think
it just feels like the wrong way of doing it :P
I mean you end up pulling DBus lib just for fetching the disks
Nothing else
Seems redundant
dunno lets see
Plus adds more maintenance load
Yeah let's see
if it works out it works out LOL
Just saying cuz idk about DBus personally and don't want to put all the weight on your shoulders
Alright gonna see if I can setup the dependencies for this thing, bootc usroverlay time
you can build it on a distrobox i think
i've been using that
and a symlink to the system dbus session
Yeah
That's what I did
Because layering the dependencies would've been a nightmare
@Kyle Rankin is the second command there mounting the disk from a flashed USB drive?
Don't need layering when got a big ass overlay in /usr :v
True
im really like at a loss i dont know how to make this work, the format seems to break when i flash the ISO on my USB drive
I'm not gonna be home today but tomorrow will be YAI programming day I guess
No, from local disk
it jsut dumps the entire partition stuff from the ISO directly to the drive so it doesnt even know how to mount the embedded hfsplus thing
maybe we can just specify the drive itself as the root instead of the label
@James i believe you ran into this too
Take it easy
indeed
and modern bootable USB disks are fine with GPT?
i believe so?
ive been using GPT tables since forever afaik
i managed to replicate my issue somewhat with the iso directly
dd if=output.iso of=output.raw -> losetup -P /dev/loop10 ./output.raw -> ls -lah /dev/loop10*You say that and we managed the live iso in like 3 days 😭
same partitions as the USB drive
It's a candle, not a fire thrower lol
Take a spray and it'll throw fire
So your desired behavior isn't for the 3rd partition to have the bluefin_boot label?
Nah but you're right
i wanted the
bluefin_boot to be /dev/sd* (the USB drive itself)
because thats the behavior i've seen working on the QEMU vm
its just pointing to the ISO root itself, not the embedded partition in it on QEMU and that seems to workcan you replicate your issue by dding the iso to your usb drive instead of using Impression?
ill try it out
also i just figured out that we are re-inventing kiwi
https://osinside.github.io/kiwi/
yeah I just wonder whether Impression is trying to do something clever
maybe :p
kiwi us what timothee used to make the first live prototype
dd is the honey badger of imaging tools
all of these tools do too much
we just need to blast the disk
I used ventoy
ye but it does like the same thing
i saw you couldnt mount the root cuz of the
/dev/root logs
true tho
kiwi is a bit muchends up distroless can apply to installers. 😈
we are slowly inching towards wolfi bootc.
alright finally, managed to install the deps
this is fancy
nice!
i was gpnna add erofs to katsu but idk how to uh
someone needs to make a dracut module
I think it works ootb
ok but dracut
the docs mentioned nothing about erofs
how are they calling dracut modules to mount the erofs
iI think it detects it automatically
GitHub
dracut/man/dracut.cmdline.7.asc at master · dracutdevs/dracut
dracut the event driven initramfs infrastructure. Contribute to dracutdevs/dracut development by creating an account on GitHub.
This works ootb, no need to pass any arg (to dracut)
literally zero mention of erofs
>squashfs.img
wtf
ye, ey, but if it works
¯\_(ツ)_/¯
imagine naming erofs as squashfs.img
We can check how fedora does it, by dissecting their isos. They will switch to erofs with F42
needs to be that name
dmsquash-live hardcodes it to that name
if you dont want to specify a different name on the kernel command line
yes i know
but
what the fuck
the docs make 0 mention of this
it literally only says "squashfs"
The system will boot with a read-only filesystem from the SquashFS and apply a writable Device-mapper snapshot or an OverlayFS overlay mount for the read-only base filesystem. This method ensures a relatively fast boot and lower RAM usage. Users must be careful to avoid writing too many blocks to a snapshot volume. Once the blocks of the snapshot overlay are exhausted, the root filesystem becomes read-only and may cause application failures. The snapshot overlay file is marked Overflow, and a difficult recovery is required to repair and enlarge the overlay offline. Non-persistent overlays are sparse files in RAM that only consume content space as required blocks are allocated. They default to an apparent size of 32 GiB in RAM. The size can be adjusted with the rd.live.overlay.size= kernel command line option.
same mount options with erofs and squashfs
if you have support on the kernel itll just work
docs simply didnt account to that
indeed is confusing
youre telling me they straight up call
mount squashfs.img
and not mount -t squashfs squashfs.img
@madomado damn we can actually use erofs ootb
It just works
how the fuck is it doing that
GitHub
dracut/modules.d/90dmsquash-live/dmsquash-live-root.sh at 5d2bda46f...
dracut the event driven initramfs infrastructure. Contribute to dracutdevs/dracut development by creating an account on GitHub.
So literally a drop-in replacement
found this.
GRUB2 How To (2) : Make boot-able ISO with xorriso
There are many tools to make boot-able ISO already. Some tools are good at some aspects, but lack supports for other features. Some tools o...
https://github.com/FyraLabs/katsu/blob/1e26ecf74164c90bc24299a66f8495eb2aef4845/src/builder.rs#L771-L832
for reference
GitHub
katsu/src/builder.rs at 1e26ecf74164c90bc24299a66f8495eb2aef4845 · ...
The vicious image builder. Contribute to FyraLabs/katsu development by creating an account on GitHub.
noice
cc @Zeglius
For UEFI I assume?
lets only support that
BIOS is always a pain and we dont even support it anyways
also
noice.
use that then!
told you was that simple :v
looks perfect to me
Sigma stuff
Super easy to parse
Okay you take that part, somehow Im struggling more with javascript than rust
😭
Yeah don't worry that's my part
based sigmas
I just need you for the rust code 😭
Now that sounds kinda wrong
But I guess you know what I mean
@Zeglius HES USING YOU
do not. help him.
:glorp:
yours dont give the disk size :v
its cuz im too cool for it
Fine, write the UI yourself 💅
only LOSERS (like @Zeglius) need that
Zeglius disk parsing is much easier for me to display than some dbus mess
Who needs UIs, a TUI is more than enough, these kids today tsk
anyways, jokes aside, your call, you handle like 99% of the workload with the ui
If you write me the bootc call then I'll handle the rest
Not sure what I need after that
Probably some logs would be nice to display but no clue how one does that
theres this
but we need realtime logs
so portable_pty is what we want
mmm some sort of stream I assume?
portable_pty - Rust
This crate provides a cross platform API for working with the psuedo terminal (pty) interfaces provided by the system. Unlike other crates in this space, this crate provides a set of traits that allow selecting from different implementations at runtime. This crate is part of wezterm.
Huh
Sounds like what we need
right but how do we tie that to javascript
jsut kidding its ur problemo
i dont know actually
there was a terminal written in tauri
GitHub
GitHub - marc2332/tauri-terminal
Contribute to marc2332/tauri-terminal development by creating an account on GitHub.
Yeah good question
i guess we can literally just use that thing
That will probably be the hardest thing to figure out not gonna lie 😭
Man, UI developemnt sucks :v
@Niklas ⚡ heres your part https://github.com/marc2332/tauri-terminal/blob/main/src/main.ts
GitHub
tauri-terminal/src/main.ts at main · marc2332/tauri-terminal
Contribute to marc2332/tauri-terminal development by creating an account on GitHub.
alrigh alright, before we dip our toes, lets take a look at tauri docs, there must be something regarding streaming data to json
Actually nvm @tulip🌷
https://v2.tauri.app/plugin/shell/
NICE
Wait
We could have used this instead of the Command::new
🤦♂️
Cursed
alright
This is waaaay easier
love the iconography
I'm gonna separate that out in it's own little react hook later but this looks fire!!
yeah. gonna pr
wait nvm had it already, gonna open it for reviews
Then we need to get it into a live iso somehow to battle test it
yeah now we need to figure out how to stream with that shell plugin into react components at real time
NICE this is exactly what we want.
Yeah so idk about react, but I assume we can make a function that takes as params an html block that has a text parameter while is installing, aka the log in realtime
And another param for an html block when finishes
thats just javascript, @Niklas ⚡ probably knows how to make a component re-render as a variable changes
Actually I donno if react let's you do that
i know you can do it with vue
solid too
Hard habits die hard eh
LOL why tho
i personally dont super like the react syntax but its all the same shit
Nah I mean vue, it is still ingrained in my head
i think you can do it via the react useState hook
vue the goat tho
Indeed, love that thing
That's easy
i don rember how to do ir
it
Wait how does one pass parameters to the next page
Need to pass the disk
check out the router's docs
i think you also have react state there
Either via the router or we store into some store
i think i found it
navigate({to: "/", params: {}})url params?
looks like it
Yep
afaik you can also fetch that w/ raw aah javascript
No please don't
The router has type safe constructs and all that built in
didnt know
nice!
https://tanstack.com/router/latest/docs/framework/react/overview
Tanstack stuff is some of the most ergonomic tooling for every framework in the web dev world
okay, i have no clue about routers, but I think this should be enough to somewhat hook the bootc install
https://github.com/ublue-os/yai/pull/2
https://github.com/ublue-os/yai/pull/3
Sick
I'll take a look when I get home
Much appreciated!!
anything I can do to help?
Not really I think
I just need to get home because I don't have a computer with me right now :Harold:
I PR'ed the fun qemu-docker run to the titaniboa justfile. So I can use it to do some remote live iso testing from the train
Yeah my Linux laptop is just not too reliable right now, I'm waiting for my framework
@imbev in here
I think PopOS does something similar with the recovery partition
@tulip🌷 @Zeglius oh I got it
we just say yai is a component of titanoboa
that way we can keep the cool name
PRs incoming
Another prehistoric creature?
I'm sensing a trend
I'm just waiting for a project called "Velociraptor"
too common
Maybe an image focused around clustering.
Next up pick something from the prehistoric water era, those giant water monsters 😄
if we're going to kill an anaconda we need to bring the big guns.
actually everything will be fine the people who need all that advanced stuff can just use anaconda
Or just include something like gnome disks on the ISO
you know that's not a bad idea
it's a flatpak we can just add it to the list lol
Useful for troubleshooting too!
then we just make sure everything included has an updated runtime in flathub, and BOOM, lots of savings
noice alright
sounds good to me!
Btw heard GTS didn't have an iso for a while. Maybe this will do for an use case?
probably
there might be some funkiness but lets see
GTS is always a pain point
for whatever reason
well it'll be what stable is now in a month.
so we are good to go then LOL
yeah I would ignore GTS
this is more important since it basically unblocks LTS from having a real release
and stable's will be better too
just an overall improvement overall
we can also ship ptyxis as a flatpak now :)
The ISOs haven't been on the website for a while now it's all good
the heresy
im thinking we should get the live ISOs going, some testing, then just straight up revamp the picker w/ ARM/LTS and stuffs
yeah
oh
this will also mean ARM and GDX will have good installers
true LOL
are we prepped for arm?
for this I mean
i believe so
we'd just need to run this on arm runners, theres nothing really arch specific i think
ok so this fixes every bluefin except GTS, and in a month it won't matter.
SOLD.
yeah LOL
this is so awesome tho
like tbh we are a single issue off of fixing all ur problems
GitHub
Real hardware seems to not find the LABEL=bluefin_boot label on the...
I dont know exactly what is going on but apparently dracut cannot mount LiveOS/squashfs.img since the label/location specified by the kernel command line on our grub configuration seems to just not...
this
and this can be fixed, i already found the fix
how is the image on the iso compressed, for the OCI image I mean, the
bluefin:stable one. We'd want to optimize for max right? to get the iso size down? Or is there a sweet spot?the ISO itself mostly depends on the compression for the rootfs
squashfs has the best compression but worse read times
we are using erofs with lz4hc compression, it trims down the iso by like half over not having compression
zstd would be best but QEMU apparently crashes with it
lzma is way too CPU intensive but it makes super small ISOs
i believe we can make the OCI image on the ISO itself a lot smaller, like, i dont know how we can compress the image in containers-storage itself, but it would be awesome if we could
bootc install supports any transport mechanism that bootc switch supports, check out their --help pages
we should probably compare the sizes and install times with erofs and squashfs when we get the installer done and see if its worth it or not
grub itself also makes the ISO slightly smaller, about 300megs on x86_64 with x64 and x86 modules. we can compress them but probsbly the best idea would be to trim some out
right, so like after we finish and we need to spend more builder time to squeeze it, I think it might be worth looking into
absolutely, we can get some crazy gains
LOL
Erofs is compressed so it doesn't matter
If you compress the image too you will just eat more CPU cycles and prevent block deduplication between the image files and the installer files
So is this thing also a free net-installer since you can specify a remote image?
Yepp. I just installed bluefin:latest with this.....
It had fedora flatpak repository instead of flathub so... No flatpaks
you just need to copy the flatpaks over from the ISO
it uses the stuff on the container by default
i think we could also just add flathub from the ISO
interesting, where would the remote have come from tho?
they get defined in /var/lib/flatpak/repo/config
we can also specify the installation target's flatpak installation if we need to do some overrides or whatever
yeah but mine only has flathub
I wonder where fedora is coming from
it's the default that gets created i guess by the tmpfiles.... and we override it somehow?
well installing my custom image now with this
i feel like its the default that gets put in by fedora on /etc/flatpak(blash)
we did some overrides on bluefin LTS and i didnt know thar we could have flathub directly on the image
fedora probably does the same
also.... do not, i repeat do not attempt to mount the same ISO in the live mode at the same time to multiple VMs.... they all crashed
oh LOL
LES GO i think i figured out how to get the containers straight from the initramfs instead of with the overlay
just gotta bind mount
/run/initramfs/live/(arbitrary_folder_that_has_containers) to /var/lib/containers
/run/initramfs has the contents of the entire ISO in itlike boot!
for whatever reason
/boot doesnt get mounted by defaultis it perhaps looking for a label for
/boot ?i mean, you dont need
/boot on a live image
i think they dont mount it by default just cuz theres no need for it
i think grub-mkrescue was defaulting to making a hfsplus partition because the squashfs.img was more than 4GB in size
(max size for FAT32)
this is being an amazing learning experience at least for me
hopefully it is also being one for you guys
i've never extensively messed around with initramfses and stuff like thatBoot often isn't actually mounted unless needed
yup!
i was just mostly testing out if it works.
although this means we cant do online installs if we dont wrap podman with a different containers storage
/run/initramfs/live/containers /var/lib/containers/ none defaults,bind 0 0Okay so I'm noticing things not working right with erofs. Some boot failures.
Switching back to squashfs
interesting
we should make it an option
but its not erofs thats making it funky tho
where are you seeing the boot failures?
With my images when using erofs, dracut cannot find the root fs
Haven't noticed with bluefin-lts
are you booting from the (RAM) option?
Both
i wonder if the kernel that your image has just... doesnt have erofs?
no idea
I'm using a fedora kernel
you should try manually mounting the
/sysroot thing on the initramfsLazy
run the (Debug) option
LOL get you
I swapped to squashfs
lets see if it works
would be goated with the sauce if you PRd like a "if squashfs then mksquashfs" thing.
cosmic
WOW NICE
man a live image for cosmic is something id not expect like ever
with .... zfs
ooooooooo
fancy!
i wonder if you could actually just straight up do
root-on-zfs with this
on the targetno comment
hard part would probably be getting the initramfs to like it
https://github.com/ublue-os/titanoboa/pull/29/files
@Zeglius THANK YOU for working on this
me no like github actions
What did actions did to you 😭 is just yaml with embedded bash
weird.... this vm doesn't have a disk
its badly documented, hard-to-secure, hard to test YAML with arbitrary embedded languages
oh the qemu action doesn't have a disk
oh LOL
i wonder if you can add one
im just using the
just vm recipe and using virt-manager directly when i want a diskWdym badly documented? GitHub have docs :v
Anyways
yeah but its so hard to discover like anything there
its not the good stuff
Its a matter of using the good set of tools
i wonder if theres a gh actions LSP
GitHub
languageservices/languageserver at main · actions/languageservices
Language services for GitHub Actions workflows and expressions. - actions/languageservices
? wdym lsp?
like for the
${{ expresions }} ?for the completions and stuff
i dont have anything on helix
ah yeah, jetbrains has that
"yeah, jetbrains has that"
Jetbrains is the goat
Why else do you think they can have a subscription for their editors?
Add the yaml-language-server
oooo
@Zeglius Created this PR to test your action.
https://github.com/ublue-os/titanoboa/pull/30
NICE.
i think we should have like, bluefin, bazzite, all the cool images as the check
so its comprehensive
Planning on also running the ISO and capturing the screen once booted. Would be some integration testing
openQA maybe?
Na. Just
just vm (plus a couple tweaks to make it work headless)they use it for GNOME OS i think
OpenQA needs a server IIRC
ooo
didnt know
although isnt the github actions runner enviroment a VM?
nested virtualizaton on the free runners seems like a bad idea
man we also need to test aarch64
It should allow nested virtualisation already. It's used in the original ISO action tests
nice!
aarch64 is just a matrix too. Not a big deal
i personally dont have any idea if aarch64 even works with this
@Zeglius based
so bootc install to-filesystem is a bit of a mess to read through right now.... but from anaconda here is how things get installed
https://github.com/rhinstaller/anaconda/blob/main/pyanaconda/modules/payloads/payload/rpm_ostree/installation.py
mmm for what I see, it handles bootloader installation
damn, bootc really cut down some fat, eh
Well bootc is also doing a lot of this stuff as well.
Both are wiring things together. Bootc uses bootupd to install the bootloader
It looks like the main thing is setting up the rootfs, creating bind mounts, setting up the ostree repo, doing ostree container image deploy .....
@madomado https://github.com/FyraLabs/katsu/blob/1e26ecf74164c90bc24299a66f8495eb2aef4845/src/builder.rs#L245
is this
efiboot.img getting replaced on the mkefiboot step?
like, im trying to replicate what you guys did here with katsu but in bash LOL
that mkrescue thing seems to be getting overwritten when you make the EFIboot partition on https://github.com/FyraLabs/katsu/blob/1e26ecf74164c90bc24299a66f8495eb2aef4845/src/builder.rs#L151ok so
@M2 @Zeglius we will have to somewhat trim down the single
squashfs.img file
i managed to trim it a lot by just moving the containers off of it, into a separate directory on the ISO rootlike this
maybe we can do something similar with flatpaks?
or some other part of the image that can be slightly trimmed
we cannot get this working without fitting on the arbitrary aah FAT32 file limit
hey also tulip
mmm?
if the yafti workflow is fixed, we could slum it by going browser-only on the image if need be and then post install the flatpaks
yup!
we can literally just do the same thing we are doing with yai
tauri yafti 2.0 coming soon ™️
i wonder if we can use exfat instead of fat32 for the partition
it worked with hfsplus..
ill just trim down the image rn and see if it works
that would be ideal
probably not going to happen btw
we will most likely need to figure out some way to remove the flatpaks off of the squashfs.img and have it side-by-side with it
we can/will iterate on this further anyways so its gonna be ok
or we can just use squashfs instead :P
mmm then lets do that
it booted
seems the least complicated option
IT BOOTED
with the super complicated script
ill pr this
now
with the what
you'll see
Im scared lol
its... terrible
promise you
GitHub
fix: use fat32 as the main storage partition by tulilirockz · Pull ...
This allows us to boot straight from USB drives and stuff
tf
its bizzare but it works
got all this code from katsu
ouch, it really is difficult to read
we can/should refactor it a lot
its just a pain to understand at first
alrigh, also if you do that, remove the sudos within the podman heredocs
otherwise github actions screams
yup!
yes it's getting overwritten
this should actually be bootable now like on a USB drive and stuff
nice alright
kinda
wait
it worked
what it does is it replaces it with the same content but a larger file
we might not need to dump the container image in container-storage after all
yup
mkefiboot() takes the image and grows it until it reaches 25MiBwe just gotta get bootc install to recognize something bootable at all
why specifically 25MiB?
you need to grow it because you need to copy recursively
new_root/boot/efi/EFI/BOOT into mountpoint_of_efiboot_img/EFI/BOOTyeye
15 might be enough, but yeah we made sure it's really enough so we made it 25
@Zeglius ISO is bigger
dang
why is that
no compression on containers-storage
cuz its on the FAT32 partition
weirdly /boot is being mounted during install
i specified it on the fstab just to make sure its working
okay, sound it will be easier by using other than fat32
check out the
fstab.sysno during the install, it makes the partition and doesn't mount it
i think we can make like another erofs image that gets mounted on boot
or squashfs
whatever
okay, so an erofs of the container storage?
heres the fstab, it gets mounted after switching to the image
yup
then we mount it on
/var/lib/containers
itll be read-only tho
cuz its going to mount over the overlayfswell, /usr/lib/containers instead then
based
i wonder if
bootc install can find that location
oooooooo
as long you set that in the storage.conf, should be all good
but i think like centos doesnt have that by default
we might want to just patch that LOL
yup, default on c10s
probably upstream too
i gotta also test if the shim works, this might have enabled secureboot for us
i mean the official fedora shim
a
nvm then
no need to do that if we have that
:P
@tulip🌷 Did you figure out copying over the Flatpaks to the install?
I remember seeing something about it yesterday, but can't find it again
I remember seeing something about it yesterday, but can't find it again
yeah, I think if you copy over all of the shim stuff it will work
cp -f /var/lib/flatpak (mounted_target_filesystem_root)/ostree/deploy/default/var/lib/flatpak
findmnt /dev/(target_device) then mount its rootOh. That's it? Awesome!
YE
man
this thing boots on UEFI
just fine.
aight read only filesystem error
gotta mount it to
/usr/lib/containers/storage and patch storage.conf :)
just like you (@Zeglius) saidWhich step are all these logs coming from?
Is it the
flatpak install commands?mkfs.erofs
its super verbose
its on the just squash stepYeah. Makes the GitHub Actions logs unusable 🙂
LMFAO oh no....
maybe we can do actions grouping?
https://github.com/archlinux/archiso/blob/d4bfaffdadb3c50045e9bf292667c973919792a2/archiso/mkarchiso#L760
Might be helpful for putting together grub for uefi
maybe we can do actions grouping?I don't think that would help. GitHub Actions refuses to display anything, including inside groups, for logs over a certain number of lines
a
its working tho
-d1 still does those super huge logsthen 0
nice reference tho
sorta... we have very little control over what's making it bootable
now we do tho
we have much more intricate control over what is on the EFI partition now
(this also means we can eventually use sdboot and UKIs)
more spagetti
absolutely
also mkfs.erofs logs killed my laptop
brave froze
LMFAO oh no....
mine also freezes
then it just... works
for whatever reason
yeah 0
mucho texto
PPFFFF
It errored out with no space left on device. Added the maximise build space action in the recent commit
LOK
Hope logs aren't filling up the drive 😆
i think it was the logs in ram
they might be saving thrm somewhere
how much ram does a runner have
no idea, but even then, maybe the place where the logs are at has a smaller tmpfs size
16 GB
lets just use squashfs right now. It works "fine"
sgtm
yeah we can fix later, I think close the loop asap
we are super close
and not toot about it tulip. 😄
NOOOO it was so cool 😭
its still undercooked tho :P
yeah, it also doesn't work yet
BRUH, even with -d0
the 1 billion lines
LMFAO oh no...
lets literally just
>/dev/null it all
its impressive how much stuff gets on the erofs output
i just set it as --quietCan we do that while keeping stderr? It's probably obvious
yup, its already doing that
the
>/dev/null
fuck it
based LOL
pr it
M2 should be pulling those changes in his pr
he has a PR?
so whenever he triggers again we will see how it goes
yeah for testing
yeh second one, is using my PR branch for the action
OH
wait
its from @Robert
im blind
@p5
pfff
:clueless:
Triggered a new run. Feel free to trigger it yourselves too - don't wait around for me 🙂
working on getting the flatpaks working on erofs
gonna do
erofs for them + overlayfs for the flatpak storagecant we really switch off of fat32?
seems would solve all that issue and would simplify
honestly good luck figuring it out cuz xorriso is the worst thing in existence
we probably can do it, its just... very difficult to do it with our xorriso tooling rn
and nothing else replaces it
we are both one letter and one number
true LOL
like t3 guy
uefi can only read fat32....
thats only for the UEFI partition, we have that and a big F32 partition for the ISO contents
before this rework it was
HFS+ so it had support for the stuff we are doing directlymake an ext4 container
yeah thats the idea
but..
heres the xorriso command that generates this
(this is awful since we'll also need to manually change some stuff for other architectures)
that's why I sent you the arch one....
can we use it for that?
its bash
calling xorrisofs
i mean, their stuff is different tho
the literal functions have configurations for grub/systemd and so forth
ill test it out :P hopefully its better than this
It all goes back to making the iso itself, damn
this is working tho
i am thinking we should just stick with this right now and change the iso building later when we are sure everything is going well
I mean yeah is working but we dont know how :v
anyways
oh @Zeglius dont remove the sudos, just use the sudoif PR that @ledif was working on
the action simply calls just with sudo already so
we gotta have that for local testing i think
dunno, shrug
So I've been pretty off today but then i read this and remember yall using erofs...
AI-Approved Live ISOs!
almost as if using all that server tech for the desktop is a good idea.
yeah LOL
exactly
theres also going to have a rust implementation of erofs
A Re-Implementation Of The EROFS File-System In Rust Has Started
The EROFS read-only open-source file-system has seen initial patches posted today for beginning to re-implement the C code within the Rust programming language for better safety guarantees and the possibility of more performance optimizations.
:v
Fire
bruh
awesome tho
I mean
why reinventing the wheel
telling you you need to stop thinking like that
😭
KISS and not-reinventing the wheel are huge fallacies
(ofc nice sometimes)
I mean, we are here writing an installer
who are we to throw stones lol
true
we are pushing the bar up
also someone wants to merge the things in the YAI repo? or so i can do it myself, i want to get it upgraded with more bazinggga
bazzinga
wow nice 3 prs
yeah
zeglius cooked
2
other is renovate
you have write now niklas, go nuts
sick, thank you
Okay finally it did build
almost :v
nice LOL
hell yeah
im fixing the flatpaks situation
they are
2.3GB standalone
aight hopefully this should be done
lets see if the flatpaks overlay workswe can cut that down
like if something pulls in a runtime we don't want, we can nuke it
i think we have two GNOME runtimes there
thats it i think
yeah we can fix that later, maybe PR updates, etc to slim it down
Wouldn't break then? It pulls only the needed runtimes?
but for now comment out things that make it slow
no I mean, we should only have things on the ISO that pull in the latest gnome and fdo, if an app is behind and wants to pull in an older runtime we can just not ship that on the iso and punt to yafti
Yeah it will break regardless. Some apps use older runtimes intentionally because of regressions in the new runtimes
Wait nvm missed the last part
yeah I just mean the default ones on the ISO, for bluefin it's like GNOME apps, which should all be on the same runtime
@tulip🌷 from bluefin lts, you can call bootc-install i guess right? just so i can test a vm and test the disk blasting :p
yup
bootc install works for all our stuff
oh great
time to get a vm going
jsut make sure you are on an older iso, the new one isnt working 100% yet
yeah about that
where do i get one
just build ghcr.io/ublue-os/bluefin:lts on titanoboa's repo
i recommend using zeglius's fork right nowyeah i'm just wanting to install the thing onto a disk and then test blasting another
GitHub
feat: add iso building action with rootfs hook by Zeglius · Pull Re...
Implement reusable action with hook-post-rootfs input, allowing users to do customized changes to the rootfs without needing to apply them to the base image (such as placing their bootc installer o...
because i think setting up the tauri app on a livecd is gonna be a pain
nah its just a RPM
not going to be a pain at all
even without the RPM its still not going to be annoying
:huh:
i can sense sarcasm
no sarcasm
its really fine LOL
oh lol
but isnt layering disabled? 😭
eh whatever, ill figure it out
There is a little script I made old up
As reference
nah we have access to the entire rootfs
nothing to worry about
sudo dnf5 install-disk-blaster
ok fine
yup
i was bit outta the loop on that one, good to know
ty!
Oh fuck I already merged main
mm?
no worries we can just fix the conflicts
Oh is just that?
I fixed them already
nono
im getting the flatpaks thing working
A
B
ok so. we are kinda screwed with the flatpaks 😭
no dedupe with the container image
add a little more and then we have 10 gigs
magic number
hell yeah
Ok so the root filesystem files are supposed to be the same than the container image ones right?
we can just have the flatpaks n everything on the same
.img file but stuff like bazzite would blow up
i think soStupid idea, what if we abuse hardlinks
The root filesystem is already protected by the overlay
We use something like rmlint to create hardlinks, and we slim down the thing
you know
we can do the same erofs shit with the containers storage too
man i forgot about that
that trims down like 2 gigs
?
Wait
Didn't you make an erofs of the container-storage already?
no....
i forgor
💀
:huh:
Bruh
and theres the next fat trimming opportunity
Also mount already handles erofs ootb
All you would need to do is to make it and done
With the same path
wut how does that work
/var/lib/containers.erofs?Yeah
Something like that
No flags needed, mount is smart enough to handle it
Wait
amazing
Actually we have a problem there
We need to somehow stream the contents of storage and removing the moved files all at once
I think mkfs.erofs has some flag to handle tar streams, and tar itself can remove files already read and passed to the stream
Otherwise we end up hitting the max size. I'm already doing in my custom image
Anyways I'm hungry going to prepare dinner :v
--ungzip
i love the question marks during flatpak install
aaaand action broke, great
gonna restore to the sane version, good that i keept it in a branch
Will try to fix it later
almost getting the container stuff working
that made the ISO a lot smaller
without any more optimizations the bluefin LTS ISO with flatpaks n everything is about 7.3GB rn
Damn
That's pretty good
Now put Bazzite through it
Will it explode?
It will be roughly the same size without the flatpaks
bazzite is huge
we should have like a small image just for the ISO cuz its just too much
Nah :v, USB sticks are pretty cheap nowadays
Besides, Kyle gave us green light regardless the size
we already have gigantic ISOs so i guess its fine right
LOL
Also want to stress test that thing later on with other than our images
Heard of playtron?
yeah bluefin stable is 5.7GB
Let's see if can keep up with an image that is out of our control
add it to the checks
bam done
but like, happy to trim down the flatpaks
we can do lots of stuff to get this better optimized
Trimming down fonts would free up like 1GB
yeah there's probably a kde runtime in there we can ditch (bluefin)
noice
i got the flatpak list straight from bluefin LTS
the fonts are dx only though right?
yeah we just need to audit that
i think... so?
but we can move the fonts to yafti whatever
Fedora media writer
Thunderbird looks pretty big and not too useful in a live ISO. Uncompressed, 275MB
No.... We went with the nuclear option for tofu
oh I thought we switched to that gnome image maker thing
and yeah tbird can go in post inst
or once yafti can work go "nothing on the ISO except firefox" and then post install the rest
Did we decide whether the container contents are deduped with the live ISO contents?
Because I think wallpapers are quite large too
Because I think wallpapers are quite large too
i dont think so
the ISO wasnt much smaller than this when we were bundling the wallpapers w/ the ISO
the jxl's really made a difference
true
I know they are working on shared deduped storage
but we probably won't have that for a while
i wish dracut wasnt so locked down
@tulip🌷 something like this?
pog
just check out the latest commit
try rebasing off of it
its working good now
a
managed to get it working :P
so erofs in the container storage isnt needed?
we can just mount the erofs thing as
/usr/lib/containers/storage
no need to like do anything
@Zeglius i got... a few conflicts 4 you
i forgot that i didnt merge the fix when i told you that
now its merged and you have a few conflics
ill try to fix em
man
this live ISO is working kinda crazy good
loving it
installed a system with flatpaks and it just worked flawlesslyoh i see you already did the erofs thing thats why conflicted
yeee
also got confused
the fstab can just be replaced by the one on
mainoh k will drop mine and done
the other stuff should be ez to fix
so @Zeglius do i fix the conflicts or you
i saw that you commited something
if you wanna do it ill just try working on p5's pr
no need, fixed em
nice
My PR should be good to go? I don't think there's too much needed.
It's just to prevent regressions
Over time, we can build out the matrix. But it's probably not needed at this time
aarch64 is broken right now
but it should be easily fixable
katsu has some code there we can just steal from them
this is a banger idea
i mean, how else do you get the installer there?
How I expect this thing to work when you all are done cooking on it: https://youtube.com/shorts/pLLDpAxqWKY?si=mUgyps0dcrHJnqKy
CodeVenture
YouTube
relatively clockwise #coding #programming #javascript #python
this video is a fun take when in programming things work out somehow but now how it was supposed
follow form more programming content specially in javascript , python , c++ , java and become skillful coder and developer
Don't forget to LIKE, COMMENT, and SUBSCRIBE for more tech comedy and insightful developer content!
#programming #coding #De...
first:
then,
then BAM congrats you have the installer
its surprisingly less funky than you'd expect LOL
well, that ignoring the xorriso thing :v
yeah
that one is a fat line
we'll probably find some replacement at some point
now you see why no sudo inside podman heredocs
didnt test it on gh actions so i had no idea LOL
sudo is such a shit program
i mean you are root in a container already
true
Why does yai have bun, npm and yarn?
Which are we wanting to use? (I vote yarn 🙂 )
Which are we wanting to use? (I vote yarn 🙂 )
i dunno we just have all slightly different ways of doing stuff
i feel like bun is pretty dang fast when installing the deps
yarn is also nice
never npm tho....
yarn is slow
bun was instant
zig supremacy
bun is the goat
also yeah kinda messy so many lockfiles
Bun it is then. Will clean things up and add devcontainers
it doesnt really matter in the end :P
@Niklas ⚡ do you support the bun decision or not
I just don't want 3 lockfiles in the repo. Makes it impossible to get the same deps
just
rm -f yarn.lock package.lock bun.lockfile every time you work on stuff :clueless:
also
man we should make like a distrobox assemble for this thingI'm making a devcontainer
me no likey devcontainer cuz i cant use helix on it 😭
just kidding tho its ok
I made this
idk how did that npm sneak there
probably the template
I think we should use the methods we recommend to people. Or at least try to, then change our recommendations based on that
i just dont wanna use vscode in my case
helix is just too goated
dont get me wrong, this is just a subtle hint that i cannot be left alone with the installer :v
have you seen the bootc install hook in there?
you can just use distrobox doe
you probably can bindmount the system dbus session on the thing too
yeah but then I would need to install rustover in there
rustup works with like both i think
its like brew
oh wait true home directory
ye
im so tired of working on ISO shit today
The ISO is done though, right? You've solved all problems.
We just need an installer and to clean things up
We can bikeshed on making the ISO smaller once we have something usable
we need aarch64 support and we need to merge the actions stuffs
other than that i think we are done
like for real
We have a tiny aarch64 userbase. They can wait 🙂
yeah LOL
for real tho
the scope is almost entirely covered
im writing the live ISO to my USB drive
ill send a message here if it works
after booting into it
I learned to not believe in such fantasy
it never ends 🔥 🗣️
per example forgot of the Grub menu templating :v
Yeah make it Default
ugh true
gotta do templating
should be relatively quick
ledif was already working on it
is there a pr?
also some mkfs.erofs call is not being quietened
GitHub
chore: add automated tests · ublue-os/titanoboa@4e5cc35
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
but the entire thing worked on gh actions just fine
aaaaaaa
heh it failed cuz clean removes the damn iso
told you, it never ends :v
its so close tho
honestly @Zeglius just make clean not remove the ISO
its super dumb that i did that
now now
i saw it on the github webui :clueless:
goated
GitHub
chore: add automated tests · ublue-os/titanoboa@4e5cc35
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
this should be it gamers
lets see
if we are gonna get the cool action going
I meaaaaan
its done bru
we probably run out of space with bazzite images
ur pr is good
ugh true
lets see what happens
as a gentlemen once said
clear throat
isnt that from stanley parable?
yep
also this is the most goated wallpaper ever
i wonder how good it would look on an oled screen
too bad oled monitors are extremely expensive
ugh oh no
why is this damn thing failing
You know what, let's take a break and fix it tomorrow
yeah
see... thats why i dont like gh actiosn....
its always this mess...
😭
Yeah, I need a VM or something that runs these things locally
Act ain't enough
i mean we are doing something wrong its just usually a mess to debug like that
yeah
YO
usb drive seems to be working
imma boot into it
@j0rge @Zeglius @Niklas ⚡ @Robert @ledif this was an astounding success
im on the live ISO right now
running off of my crummy USB drive
it runs really fast too
we just need to disable the
brew-setup.service cuz it tanks the performance right at the start
erofs + overlayfs in ram is just too goated
its not laggy at all even on really bad disk speeds
although i got some errors due to probably missing setuid binaries that we need to manually add
heh gnome software doesnt like it
yup it was just fusermount missing setuid perms
Please can I have a review
https://github.com/ublue-os/yai/pull/4
Standardises on Bun, gitignores the rest. And adds a reference devcontainer config
oh ofc
acked
Should I get an action going that will build an rpm out of the code? Might be useful for testing in a VM
There is already the packages repo
I think building RPMs in GHA during PRs would be helpful though. Use the packages repo for "prod" builds, then GHA for dev/test builds
Tauri also builds native RPMs, so we might decide to just use GitHub Releases to distribute them. Saves us maintaining spec files
my idea was only to get it going in GHA because tauri has native support for rpms and that stuff so its easy to implement
Agreed.
Let's try not over engineering before we've built something usable 🙂 These decisions can all be made at a later date. We need something that works now
Let's try not over engineering before we've built something usable 🙂 These decisions can all be made at a later date. We need something that works now
sounds perfect
i think the ISO itself is like done now
also @Niklas ⚡ the shadcn guy is the only maintainer for shadcn
did you know that?
i thought that shadcn was made by like a lot of people
I would be surprised if he gets no prs
But yeah
I get your point
It’s interesting
When’s tulip UI coming out?
daisyUI — Tailwind CSS Components ( version 5 update is here )
Best Tailwind Components Library - Free UI components for Tailwind CSS
Hm
tbh big fan of this one
Good ol reliable
disk != usb drive
usb drive is rip
i was running off of a usb drive
model?
super cheap multilaser 64gb USB drive that i bought off of amazon
Pen Drive Twist 64GB USB Leitura 10MB/s e Gravação 3MB/s Preto Mult...
Pen Drive Twist 64GB USB Leitura 10MB/s e Gravação 3MB/s Preto Multilaser - PD590
exactly this one
its a piece of shit
@Zeglius we did it
GitHub
chore: add automated tests · ublue-os/titanoboa@4e5cc35
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
And the log is readable!
yes 😭
honestly like i wanted to not require
just build to be run with sudo
i guess we can do it with sudoif or somethingwhy did you not buy a sandisk one or smth 😭
dunno
it was cheap
yeah because it works one time and then dies
tbf you can buy a few of them for the price of an actually good one
also this means the minimum specs for the Bluefin LTS live ISO in specific are like
- 16GB flashdrive or something (at least 5MB/s read speeds)
- 4GB of RAM
- any cpu that has released in the past 10 years
this is not a lot
to ask for
yeah
I placed the aurora "system requirements" at 8gigs of RAM
Why would you care about write speeds though?
because less is just shit
you gotta care about read speeds sorry mistyped
read speeds cuz you'll need to read the containers storage and flatpak storage off of the live ISO
I mean, it’s great to have offline installer and all but like…
ISOs are not updated that often
bootc updateDoing that only makes the OOB experience being updating packages that was just installed because it’s out of date
and we rebuild ISOs every 2 weeks or something like that
Still, you could also make the live image a netinstaller and not bother to update it until the next LTS release
yeah, we had netinstallers before the current offline ISOs and now LiveCDs
it wasnt great
It would be a little less shitty having an actual live system for users
GitHub
feat: make github actions tests more comprehensive by tulilirockz ·...
This just adds a few more images so we can test this better!
looking fancy
hopefully everything should work
omg what have I missed today
github actions for titanoboa :)
now im just making sure it works multi-arch
GitHub
titanoboa/.github/workflows/ci.yml at main · ublue-os/titanoboa
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
we can pretty much already just replace the Bluefin LTS build ISO workflow
but im trying to make it work multi-arch so we can actually do it just fine
can I get a test iso somewhere?
i can send it to you via warp i think LOL
@j0rge
3-hideaway-glucose-stethoscope-dreadful
no installer on it yet
itll take a bit, my network connection is trashhttps://github.com/ublue-os/titanoboa/blob/main/action.yml
this thing too is very nice
GitHub
titanoboa/action.yml at main · ublue-os/titanoboa
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
oh ok
do we want to start pushing test ISOs to cloudflare?
like just get that out of the way?
sgtm
then people can test en masse
yup!!
Oh ISOs would be fire
I'm trying out titanoboa for LTS on ARM64
not gonna work right now
im patching up the justfile right now cuz of that
i think its like a single argument on the
iso: part that is broken on aarch64lol I'm pathcing the justfile too macOS didn;t like
:Zand the packages too
i just built an ISO but it doesnt boot lol 😭
which commit?
and image
dunno just cloned main 30 minutes ago
and bluefin-lts
where isnt it booting?
qemu?
yeah
linux, too
right?
maybe something failed?
find work on the repo root
also LMFAO find work
i mean the iso has 7,8 gigs
wow what
so i guess it worked but somehow the boot broke?
did you um
ctrl+c it before it finished?
on the
iso step?no
😭
everything probably worked, its just the ISO step that got mad
ill wait for your ISOs
try running that one again
its very close so it should be ok
yeah
i can wait heh
what does the RAM mean in the boot menu?
beautiful
it just loads up the entire thing to RAM LOL
i dont think it works anymore
oh wow this is so good tho
YES
its amazing
om let's have it not ask and just do a live session and announce this
the performance is crazy good too
can we remove the toram option?
then we'd be good
yup
manually added that for debugging
o
the to ram option works!
bluefin-live.iso and then we'll say for f42 we'll have one for each.
oh if it works then leave it I guess hahah
best would be just removing everything except the regular ol install option
should be able to reuse the push to R2 action
yup im doing that right now
jsut fixing an error
well for now let's just do the live cd because the installer isn't done
it'll just be a live CD for now
like remove the dialog so they don't have to care for now. We'll add it when it's ready EZPZ
wiring up bluefin LTS to it
plus we should tell people we're doing this it's too good
YES
its awesome
like let's get some wins
you'd be surprised how fast the ISO is on a flash drive
theres really like no lag on the UI at all
its miles ahead of most live ISOs tbh
dang
so we win a double
better OS and live iso experience
😭
live ISOs are getting built
on the bluefin LTS repo
removed checksums cuz we didnt implement that yet
that should be fairly easy.
ye
YO they are actually building
wow
@tulip🌷 do we have an issue on what is left for the Installer?
yai or titanoboa?
yai just needs to get the installation screen wired up i think
Yai. Titanoboa seems like it's pretty much done.
ye LOL
yeah so yai just needs that
i think
cc @Niklas ⚡ @Zeglius
There's still no user creation screen or backend, and we're not using Rust at all - all commands are originating from the frontend.
Needs a bit of cleaning up
user creation would be important for KDE specific images.
mm?
Fortunately Gnome doesn't need that.
we are using rust tho
We have a rust function, but from what I could tell, we aren't calling it.
Disks and bootc are being called directly from the frontend
oh yup
lemme see actually
yup we are straight up using the thing there
i think its fine tho
dunno
yeah its called directly from the frontend via tauri adapters
its just a shell call, in rust or in the shell tauri module
it doesnt really make a difference
does it work with a bazzite image?
too big
:(
wait what?
I tested a Bazzite image the other day.
the container is way too big for fat32
we moved to it cuz else it wouldnt be able to boot from a USB drive
why are we using fat32?
mostly because thats the default implementation on katsu
but we can try to move off of it
as for that, maybe i found a solution
yoo?
may
I'm confused. It was working fine the other day?
I even tested a successful install minus flatpaks.
because was using hfs+
it was using hfs for the backing storage, that thing breaks when you flash it to an USB drive cuz its not being treated as a single partition like the QEMU thing thinks it is
we might be able to literally just append a partition with xorriso like we are already doing
and plop the ext4 storage there
EXT4 should work I believe.
@tulip🌷 looks like your upload step to cloudflare failed
LOL what a thing
gonna have to refactor it
It never ends
hard part is making the ISO volume ID point to that
yeah, FAT32 is a huge limitation.
could you use EXFAT?
no
X_X
nothing except the dreaded hfsplus
ok if you dont want to work
Change tauri.conf.json beforeBuildCommand from "bun build" to "bun run build"
yeah i did that heh
it fails at the rpm bundling step
yeah i know why
@tulip🌷 try with that
This hurts to look at.
-iso-level 3 -allow-limited-size -udf was addedooo
i thimk we cant do that cuz of fat32...
but lets see
🥹 finally
Okay so, fat 32 is just because of qemu, right?
its because of real hardware
ok announcement ready
someone get me some saucy screenshots
@tulip🌷 so where is the Fat32 portion?
I'm a bit confused on that.
xorriso just automatically gets the entire contents of
iso-root and puts it into a big FAT32 partitioneven for Fedora? or do they have theirs use HFS?
f32 on all the ISOs ive tested
sudo losetup -P /dev/loop(number) ./thing.iso
sudo fdisk -l
ill show you all the partitions on the ISO
yeah, I'm looking at the workstation ISO.
I wonder what we did for Build Container Installer.
on that one theres a folder in the ISO root
called
/images
that has an OCI image with the stuffbecause we store an entire Bazzite image + the default anaconda runner.
if we do that on ours the ISO gets so much bigger :(
no idea how to get this going
at least for bazzite
GitHub
Fails on iso image size more than 4GB. · Issue #27 · Distroshare/di...
Creating the iso xorriso 1.4.8 : RockRidge filesystem manipulator, libburnia project. Drive current: -outdev 'stdio:/home/distroshare/live-cd.iso' Media current: stdio file, overwriteable M...
Yep with what I added in theory should work now
NICE
ill add it on my PR
Mf half an hour and Im still pulling the container image
Actually, it makes sense this would work because of the DVD ISOs.
for RHEL.
do you have a branch? I have a recent Bazzite image to test.
Gonna push one give me a sec
Oh god the laptop on its last legs
I have a hefty rig, I can build images pretty quick
Okay pushed one called simpler-xorriso
Can't send link cuz laptop dies then
I'll give that branch a shot.
In my fork
oh, a github fork?
Yeah
looks like you are 4 commits behind 😛
I'll give it a shot though.,
@Zeglius building now.
for Bazzite Stable
I'm at the flatpak step
@Zeglius is there going to be some way to disable services on the LiveCD, but have them be enabled on actual running hardware or VM?
we don't need Steam to autostart for example.
Yeah I added a Hoon for that
In the action you can embed bash to do whatever inside the root filesystem
OH
so we could disable stuff on the root filesystem and that would leave the actual image being deployed untouched?
Yep
that would be sick.
We could probably trim a lot of fat
Including installing the installer itself inside
I like how Bazzite has less flatpaks in the KDE version than Bluefin does.
Yes, Kyle is the biggest gnome simp of all us
Actually probably will need to add an input to the action for flatpak list
that would be really convenient.
@Zeglius
What I ran:
just build ghcr.io/ublue-os/bazzite:stable 1 1 flatpaks.txt
wut
also guys
ill be out now
seeing my gf
if someone could review this it would be awesome
GitHub
feat: make github actions tests more comprehensive by tulilirockz ·...
This just adds a few more images so we can test this better!
like see if im crazy or something
this adds aarch64 support to the thing
and some other small things
how dare you. (have fun 😉 )
Ah. Probably because I'm missing some commits, gonna rebase
yeah, was wondering about that.
Okay, now should work
Well should let's see if the fat32 doesn't screen with this one
doing a build
@Zeglius
did your changes get overwritten?
Nope
Wait
I don't see your new options added lol
Oh did you pull the changes?
I did!
Is there in my log
oh
I"m dumb
nevermind
disregard.
wrong branch.
makes sense they wouldn't be there lol
I was scared for a second there thinking didn't work lol
sorry to give you a heart attack
trying again. This time with the right branch!
really hoping bazzite will work
me too!
oh my god
the aarch64 improvemrnts n stuff are working
@Zeglius you can rebase ur thing if you want to
unknown option
just removr that -allowed-limited-size arg and run just iso again
AAAAAA didn't even reach the flatpak step
@j0rge the live ISOs are on cloudflare
LMFAO 😭
@Kyle Gospo why this thing is so fat?
its the zzite
LINKKKKKKKK
-udf unsupported option
wut
need jorge to publish them :)
it should be just bluefin-live-blah.iso
My god
bazzite worked on CI
it workded
It was udf
Not udp
AAAAA
Okay who cares it works now
14GB ISO
jesus
rebase ur thing its good to go for multiarch and bazsite now
Jeeeeesus
half of it is the live OS
The flatpaks
Holy
we don't have any deduplication?
when it was HFS+, it was not this big.
yup
I mean
It does boot:
If we simply don't include them
Ok good finally for fucks sake
Wait does the runner have enough free space for this?
flatpaks is 2GB
the container itself is like 4GB?
dunno
we can make the compression better n stuff
also maybe now we can merge all the stuff into a single .img
Well 4GB * 2 cuz of the embedded image
Oh yeah pls take out all that eldritch xorriso line
we'll need it
like forever LOL
o
K as long does work
i just mean the squash flatpak and container steps
we can put em on the rootfs now maybe
Ah yeah please
here is your problem.
Yeah that was done as a workaround when the squashfs was too large
@Kyle Gospo what if we made the ISO even bigger 😉
Now we can revert it, doesn't matter
https://download.projectbluefin.io/bluefin-live-lts.iso
only one right?
Should be as easy as reverting commits right?
A plague of gigabytes upon ye, woe
that shouldn't matter because we are using EXFAT?
We are not though?
Are we
I dunno, lemme check.
jesus, do I even want to burn this to a disk?
or should be try to deduplicate? @Zeglius @tulip🌷 ?
I mean deduplicating how?
would this get rid of the doubled up image size?
Bluefin is 7.3GB
It would simply it that for sure
For deduplication, not sure
dedupe yeah
it gets a LOT smaller
especially if we do squashfs
what would we need to revert?
to put everything back on the rootfs?
since we have the
-iso-level-3 switch.The use fat3w as main storage partition commit from tulip. Well, part of it
just remove the erofs steps from the thingies
dont revert thst one
I think would be better you take care that one, is quite messy and don't know what to discard
im not doing it right now cuz im out w/ mah gf right now
waiting on the mcdonalds line
Oh ok forgot
It doesn't have to happen immediately.
the tldr is just moving the flatpaks and containers folders to the rootfs and removing the erofs steps
yeye
The good news is we have a path forward now.
yup!!!!
awesome stuff
all that needs to get added is
-iso-level 3Just that? what about all of this
dont revert that commit 😭
this is essential so that we can get real hardware booting
lright
yeah I better dont touch this thing, probably you will wake up and this damned thing will still be pulling the image
but hey, at least we dont have the 4gb limitation
exactly.
so ideally
roughly 8GB ISO
for KDE non deck image.
yeah all we need to do is get rid of the containers.img
how big are our IDOs?
ISOs
normal ones?
yup
I think bazzite reached 10GB?
or something
nope
being at least close to those is the best target wr can have
7.8GB for the Bazzite KDE desktop image.
oh
8.3GB. take it or leave it
time to try out those new ISOs
LES GOU
aaah now i see why antheas was asking for 6Gb
not even the current ones are that size
we can make em that small tho
nah nah, relax
is fine
7.8GB
if we deduplicate the image.
pretty much i think
rmlint ./work
booted first try and was soooo fast
ea
current deck image is 9.3GB
Bazzite KDE Nvidia is 8.1GB
Nvidia adds about 700MB
@Niklas ⚡ now we just need that beautiful frontend 🙂
now that i have a machine and ISO destroy and test on, you bet im working on it
does it count cheating if I attempt to make one with vue :v
Couple feature requests
1. Ability to create User account (for KDE based images)
2. Ability to use install filesystem to install to a specific partition.
Bazzite will ship something like Gnome Disks to make dual booting possible.
1. installer is going to implement it
eh
2. installer too
right, those are feature requests for YAI
with the current setup, probably only will wipe the entire disk
ye
right now yeah
that's fine for MVP.
probably can be fixed eventually
but dual booting is a thing that Bazzite has to contend with.
time to unleash zeglius code on this VM i guess
wont work, I can tell by now
Idk why tho
gut feeling
yeah there is no way getting the nvme from the screen before but that will change now 😛
would be particularly funny if it worked
lets see
@Niklas ⚡ did it work
im curious
hold on
i needed to wire it up to the code
give me a second 😭
man true we can live test it
cuz the ISO is writeable with ram
I can't believe you turned my immutable OS mutable. Have you no shame?
Ublue tails-os when? /sarcarsm
its like bootc usroverlay. its still immutable :)
i mean, someone could do it for sure
Install all the security tools, good to go.
Is there a place that lists the difference between btrfs/ostree immutable from bootc immutable? Are they even different?
same stuff right now
btrfs does not play a part in this
Ok i need to get some kind of dev-setup going, uploading an rpm and then trying it is a nightmare
and it does not work
make the RPM available in COPR?
yeah but then i still cant debug it properly, like read JS errors
thats my issue right now
oh damn
so its gonna be local testing time
onto a flash drive
:clueless:
based
you can also just install to a disk imagr
with the --via-loopback option on bootc install to-*
:clueless:
not particularly sure about what this means
the thing that allow shell commands to be executed has a wrong regex
i am trying to fix it rn
time to run it outside a distrobox
how do i make this some kind of root
eh setuid bin
So idk if this is supposed to work but it doesnt get any images back
there is nothing to install from on the live disk @tulip🌷
(theoretically, the installer works and asks for a password on pkexec, but it obv has no image to install from :P)
So the installer should be running as root or something. Prompting for the password is odd for an installer
Also, the install-to-disk option definitely leaves a lot to be desired. I actually don't think the default ESP size it creates would support one of our Nvidia images.
It creates a 512 MB partition. Meanwhile we use a 618 MB ESP and 1GB Boot partition (formatted as ext4) for our default installs
Yeah that's just a matter of rerunning it, I wasn't in that case
There is, but is in /usr/lib/containers
Hmmm
But yeah, since I wrote the bootc function before tulip moved it there, well
add that to containers storage.conf
as an additional thing
well actually was me that told to be moved there
it shohld work
just fine
we also dont need to ask for a password we can just add polkit rules on the iso itself
Yeah I need to try that tomorrow, I have to sleep now :dispair:
Whenever the separate container storage squash gets undone should be fixed
letting bootc run fine without password
Also how would one build an rpm of a tauri app? Can we add these rules to yai?
tauri build --bundle rpm
something like that
Ah so no spec file?
the spec should be different, no idea
gotta search they did something similar with cinny
Check out my new branch
Just bun run tauri build
It spits out an rpm
Or i didn't push it
F
ok now
now.
There is a branch now pushed :huh:
also we should test umm
secureboot
we have all the signed shims and everything
also man i wonder how fast the new installer method will be
Try it
I have to sleep now
someone video it!
😭
ill do like a side-by-side comparison thing
oh so its already functional?
aight!
EATS ANACONDAS
Secure boot? Dunno:p
what about being functional
is it
also, ill try making the cool workflow to get rpms for the installer
then ill write em up to the ISO
Oh neat
If you get it working ping me as usual
im just not particularly sure on how to like
how to put the RPM there...
ill un-squashfs the flatpaks and container
to see if it works
Hm
niklas
you here
what if i made a dev livestream
Barely
or a video
Why not
Tulip dev streams
Now on twitch
Wire that into an action for lts
That way we have something ready and we don't hardcode yai into the thing, two birds in one stone
yeah
Universal Blue
Rebasing Fails Due to Small /boot Partition
Hi everyone, I’m trying to rebase from Bluefin (default) to Bluefin-DX, but my /boot partition is too small. The process fails due to insufficient space in bootfs. I tried to resize the /boot partition, but I couldn’t extend it. Is there any way to proceed with the rebase without reinstalling the entire system? Any help would be greatl...
can we make sure we up fedora's defaults?
i mean, bootc install should do the job for us
i dont know tbh
yeah we should ask colin for a review of things
like when you feel it's ready enough
that makes me curious, maybe they pinned a few deployments?
/boot is usually enough for more than like 4 deployments i think
their boot partition is 332 MB.... They did something absolute bizarre
yeah LOL
manual partitioning maybe?
very much so
yo. is it just me or ...
who wants to do systemd-boot while we're at it?
i want to do it.
its just complicated to do it on the ISO
i made an issue tracking that
Yes, but no
You are going to fight shim's idiosyncrasies here
@Vitor There's an new graphical OS installer that you might be able to help with here.
welcome!
ok I was hoping since this thing is so magical we could have everything
and zfs root
and there would be paved roads of gold ...
and we'd be done, we'd be so bored
they mentioned specifically not really being too much into programming and stuff
i do think it would be awesome to have like a cool design for the installer tho.... or website or whatever
don't worry.... i'm still thinking about this
and then we'd have some magical uki things
we can tho 😭
UKIs are particularly close to being a thing, and zfs on root
just needs some work
for 2.0
this is already better and every day without good ISOs is a wasted one
true LOL
Looks like I'm getting some bootloader problems
o
wha happened
I accidently diffed with the wrong commit
Discord supports highlighting for diff, nice
yeee
its awesome
honestly sgtm
looks good
I might be doing something wrong, but qemu isn't finding a bootable device
mmmm
Doesn't seem to be a change that would break booting
oh you cant put that on the ISO root
also try building it without those changes just to see if it boots
maybe i broke something somehow
I'll do a sanity check
I'm pretty sure it was working on the same commit
mmm
Perhaps it is a permissions issue?
There were so many sudo prompts that I ran it as root
yeah LOL
you need to run it as root i think
Why sddm conf in the iso-organize recipe?
yea
you want to put it on the rootfs-include-blah stuff
or the livesys one
Or maybe not even that, there is a hook to embed a bash script in the action
I'm just experimenting with it
ye!!
also you gotta put the sddm config not on the iso root but on the rootfs at
work/rootfsthis target?
rootfs $IMAGE: init-workJust point me in the right direction and I'll do it 🫡
One way yeah
GitHub
GitHub - ublue-os/yai: WIP: Installer for Bootc Images
WIP: Installer for Bootc Images. Contribute to ublue-os/yai development by creating an account on GitHub.
@j0rge maybe you have more of an idea where to point em to :P
@tulip🌷 do we have any mockups or other screenshots?
i've got to run it
lemme just get some
You can also do this
yeah not sure frontend isn't my bag I think we'd wait for nikolas
(this is actually looking so nice)
There are a few placeholders in the last one
ye
But yeah, look nice
@Kyle Gospo
@Kyle Gospo we can also theme it however we want btw
Link me to this text so I can copyedit it nicer
bazzite branded installer would be the goat
I have ideas
Mmm for bazzite would really need some touches
Kyle is scared of CSS. He historically only likes GTK styling.
oh got it
GitHub
yai/src/routes/disks.tsx at 59c90afa7eec080e2f93eed60dd0a7f0b97ed47...
WIP: Installer for Bootc Images. Contribute to ublue-os/yai development by creating an account on GitHub.
all of the stuff in
routes/ have the user-viewable text
man we can do i18n with this too
multi language installer....ok take a new screenshot of the welcome screen pls.
I added some changes. 😄
Aurora too :)
absolutely
yo we need to get weblate going
yeah its a pain
but we can do it
not it
heh
man why didnt we think about using tauri for our stuff before
this is so much easier to mess around with
😭
and it looks better imo
It wouldn't follow gnome/plasma theming, so that's a factor
eh
But it is the same cross-desktop and is more flexible
its so much easier to maintain
Maybe we can find a way to read accent colors from the desktop
oh man
@Zeglius just found out why you couldnt erofs the entire blob
mkfs.erofs puts the entire thing in ram
and the rootfs with everything is like 9gigs.
we had the container separate from the root before so it wasnt blowing up as much
:P
man
squashfs has just so much better compression
whoa
what's in that?
what image is that
everything
bluefin LTS
with flatpaks?
yup
what!
and the other container
is that before we optimize it?
yeah thats the difference
the list of flatpaks I mean
yup.
ok so
are you saying
before any optimization
the final ISO size is 4.1
yeah
let's push that now LOL
check this out
raw rootfs is 14G
wow
final ISO rootfs is 4G
crazy
😭
what about bluefin:stable?
or would that take a while to run
just curious
didnt test it yet but its surely smaller than what we have now
wow
its pretty fast on NVME, need to test on my crummy USB drive
this is gonna be awesome
YEAH
we should still trim the flatpaks
these are the kinds of gains we can get :)
and maybe the fonts too?
yup!
also MAN
@Noel NOEL WE SHOULD USE SQUASHFS FOR BAZZITE
its gonna be such a huge gain over what we have right now...
I mean.
its older tech but man
its just so much smaller....
@Noel
we probably should benchmark all this stuff btw
so we can actually know what we are gaining from each approach
Just because it's older doesn't mean it's worse.
i know its just like
better support for erofs in general
Not necessarily. It's newer. It had a different design goal than squashfs.
If squashfs meets our requirements better than erofs we should use squashfs.
sgtm
ill just have them both as options there on the actions and stuff
i can see smaller images w/ less flatpaks working wonders with erofs
also @M2 do you know if we can do something like
just ghcr.io/ublue-os/bluefin:lts argument=value?
instead of specifying all the argumentsNo. Order things from most likely to change to least.
Unspecified arguments will use the default
You can work around this by having them be environment variables
GitHub
Add flags and options · Issue #476 · casey/just
Recipes should be able to define options: foo --bar BAZ: echo {{BAZ}} Maybe flags too: foo --BAZ: echo {{BAZ}} # is it "true" or "false"? Or "" or BAZ? Large changes t...
https://github.com/casey/just/issues/476#issuecomment-2697973509
Another neat method
GitHub
Add flags and options · Issue #476 · casey/just
Recipes should be able to define options: foo --bar BAZ: echo {{BAZ}} Maybe flags too: foo --BAZ: echo {{BAZ}} # is it "true" or "false"? Or "" or BAZ? Large changes t...
@tulip🌷 do you have the reverts in place for everything on the rootfs?
I can test Bazzite again.
everything is done pretty much
just use my branch
GitHub
fix: embed flatpaks and containers again into the ISO by tulilirock...
This should allow for much better compression, although I have been getting a few local errors with this for whatever reason
Falkon can't run as root
Is there a way to create a live user yet?
that appears to be a chromium problem
the liveuser scripts
just build (image) 1 1
1 for the liveuser
@tulip🌷 flatpaks appear to be missing?
yeah, no flatpaks on the ISO in your branch now.
did they get fetched?
try using squashfs as the compression, maybe ur system failed to compress the huge erofs image and it cut off on the flatpaks
Is there a way to appy this?
sudo systemctl mask hibernate.target hybrid-sleep.target
A file to delete?add it on the gh actions input
theres an input for arbitrary commands on the ISO pre-rootfs squash
Fairly certain you can use --root flag with systemctl to treat the target directory as / and attempt to make the links directly
Not sure if it works with mask though since that tries to link to /dev/null
for both
that command generates a hybrid image
I'm not sure if I mentioned this, but the livesys-scripts package allows you to run arbitrary commands on boot of the live environment by creating a file in
/var/lib/livesys/livesys-session-extra. This is how I did it in Blue95 to change the hostname to blue95 and rename the liveuser: https://github.com/winblues/blue95/commit/6479bc60c5d3ca208f6dd859ecebd2e2c81d5e98funny rock ridge
@tulip🌷 this is what I ran:
just build ghcr.io/ublue-os/bazzite:stable 1 1 flatpaks.txt
yeah, it didn't install any flatpaks.
didn't even tryNah, just make it all squashfs and done
already did it
If it works better, why even being a param
just finishing up the PR :)
erofs is lighting fast if we get away with it
Did you undo the flatpak/containers-storage individual squash?
yup
GitHub
fix: embed flatpaks and containers again into the ISO by tulilirock...
This should allow for much better compression, although I have been getting a few local errors with this for whatever reason
just checking out if the actions like it
so i guess if you ship this then we can try YAI on bluefin lts
we can already do it
we just need the RPM on gh actions for yai
should be "easy"
I think he meant without the /usr/containers/storage shenanigans
oh
i mean
yeah
even with them its just a
sed linei dunno what i need to do here
it should just work on fedora too
Nah nah nah no sed, should work ootb
nothing, we just gotta get the RPM going
its just c10s that doesnt work ootb
fedora already has that
/usr/lib/containers store by defaultso only centos is goofy
I mean yeah, but always will work if is in /var/lib
/var/lib/containers is expected to be read-write by podman
this is more ideomatic than anything tbhWasn't already back then?
also that little bootc tui script didnt work when i tried it :> but i guess that will be fixed now.
LOL
podman doesnt have any images in its list
Ah yeah because the image isn't at /var/lib
donezo
That's why I'm saying
Back then the image was at /var/lib without issues
its specifically because the storage.conf didnt have /usr/lib/containrs there
but doesnt matter its the same anyways :P
Ah alright
if podman works then i dont care where its stored
CI should get happy soon™️
ill make the workflows for yai
do you wanna merge in my bad code?
theoretically it could work
sgtm
its still super experimental anyways
i don care if it doesnt work yet
yeah
do it. resolve the conflict
a big code cleanup is imminent anyway, especially when making it brandable
oh oof
please no cursed gifs
having a conflict on a lockfile is cursed
m,erged
wow
what did you accept now
wut
wdym
well how did you resolve the merge conflict i mean
i just got everything from ur branch and nothing from main
ah
good
too much of a big gamer for these conflicts
react v19 is still in there
its good
thats all i wanted to know
too fast for me old gamer
yo
we can use wolfi for the rpm build LOL
but why
i mean yeah sounds nice
its just better
big fan of wolfi
but does it have a tangible benefit over just running the thing in the runner and dumping out a rpm?
oh we just want to run in a container cuz we dont want a separate env from the gh actions
like this kinda
I'm fine with it as long as you dont overengineer it and you have the webkit2gtk dependencies :P
also
bun run tauri build is enough, its only outputting rpmsi forgor about that :clueless:
its configured that way
gonna be fedora fuck it
sudo dnf check-update
sudo dnf install webkit2gtk4.1-devel \
openssl-devel \
curl \
wget \
file \
libappindicator-gtk3-devel \
librsvg2-devel
sudo dnf group install "c-development"
theres the stuff tauri lists on their docs
i dont think you need allat
idk
i installed that and it works so no clue
@Niklas ⚡ CI worked
GitHub
fix: embed flatpaks and containers again into the ISO · ublue-os/t...
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
hell yeah
nice
and where can i get those isos now
:p
those are just checks
but im adding a thing to get the output
ok sick
time to blast cloudflare with our isos
big fan of this stuff
you mean the end of life runtime?
(i know you mean aarch64) :clueless:
all the flatpaks n stuff
the output is neat
true
so much output, so wow
maybe we can get YAI working this week
i think that will be awesome
yeah pretty sure
then we'll just need to wire up some stuff on GNOME/KDE and thats it
i dread the code cleanup in yai but it has to be done
then translation, and then we make it brandable
hell yeah
now that i think about it, branding and code cleanup will go hand in hand
i18n too
we can set up weblate right away
i really dont wanna do this later
:clueless:
sounds painful
but yeah
eh not really
most i18n tools have nice support for this stuff now
we can use hosted.weblate for our stuff
i am more thinking about the branding stuff, do we just build in a bunch of flags for aurorafin and bazzite and then add some build args?
no..
i personally am going to explode if i need to do branding
for more than 1 thing
just let like the CSS be replaceable or something i dont kno
lets see what we can do
yeah
issues for later alligator
weblate thing
seems easy enough i guess
hard part is wiring it up
should be fine tho
im dreading that part yeah
oh @Niklas ⚡
i think tha thing is broken
maybe
what is broken
getting typescript errors
maybe i broke the bun.lock?
or its my local thing i dont know
probably
lemme try
javascript bringing you to knees
well we will fix it
yo we need to add this
and we want this
there is already a desktop file somewhere
i guess
because it does install one
huh
if you install the rpm it generates, there is a desktop file that gets added as well
its just a default one
ok so
im particularly lost now
i remember there was a way to add yai to the gnome-initial-setup thing
but i cant seem to find it?
yeah you broke the lockfile tulip
ehrm what the sigma
yes
it works now that ive regenerated it
clone a fresh project and try again @tulip
i just refreshed the cargo and bun locks
too slow i merged my PR now
maybe you shouldve looked at it
oh well
hell yeah
fire in the house
now we just need to publish the RPM
done
since youre the workflow god, ill let you do the honors
ooo it worked!
lookin good
Ofc it did
heh
Pls design good logo bye
omg we should ask that guy for a good logo or something i dont know
or adelie?
not like this is going to matter at all tho, we probably will just replace it with whatever
Does not matter right now at all
ye
GitHub
release-please · ublue-os/yai@cd9f4db
WIP: Installer for Bootc Images. Contribute to ublue-os/yai development by creating an account on GitHub.
pog
ill merge the "1.0.0" release tahts not really 1.0.0
pog
very sigma
jesus christ i think its actually working
wait youre running yai and an installation?
oh the release
nvm
ye
😭
we also get aarch64 releases too
POGGERS
OMG NIKCLAS
IT WORKERD
GitHub
Release v1.1.1 · ublue-os/yai
1.1.1 (2025-03-25)
Bug Fixes
actually upload release artifacts (#18) (51818b5)
@Niklas ⚡
FIREEEE
now that is awesome
yo now
last step is integrating the stuff into gnome-initial-setup and the plasma autostart
@Robert @imbev @Niklas ⚡ any tips? 😭
literally last thing for built-in yai integration
I'm on mobile, but I sent the config file a day or two ago.
mmmmm
do you remember what you said?
so i can search it
Can you search from myself. It includes a video
i definitely dont know what i would need to do lol
The message probably said "easy" somewhere in it
there
FOUND IT
NICE
thank you!
i figured out the plasma welcomerc thing
So replace nautilus with the yai desktop file
fireeeeeeeeeeee
doing that right now
GitHub
fix: embed flatpaks and containers again into the ISO by tulilirock...
This should allow for much better compression, although I have been getting a few local errors with this for whatever reason
goated with the sauce
have my review
which is not relevant
if this works first try ill be very happy
i mean who wouldnt
literally building with fire to make fire
ugh i think i forgot a flag on jq
unacceptable.
lets see
yup. i forgot it and its gonna fail on actions
click me
HADHWAHDWAH
@Niklas ⚡
:clueless:
Wait what with that log
mm?
thats my host
dnf install --transient $YAI_LATESTThe part with the overlay in /usr
yeah!
--transient
its just bootc usroverlayOh did they implement that in dnf already?
yeah the command fails because non imageref :clueless:
dnf4 ye
erhm what the sigma!!!
Oh dnf4
we should do some cool fancy transitions
with the router
first it needs to install something 😭
me wnat fancy.....
tho yeah LOL
oh @Niklas ⚡ we can just use
podman images to get the stuffDoesn't do that already?
no...
i dont think so
Oh wait nvm did that in the script, not there
it does
oh it does
Oh nvm then
OH i dont have the image on my host
got it
yeah
we should handle that gracefully tho
yeah
erhm what the sigma
podman images needs to also be run with pkexec
yep
yeah
add it, pr it, ship it
enjoy it
hell yeah
the 4 way cycle
with the building now done in github, its easy heh
absolutely
done :clueless:
so fancy
Mmm probably will get confused if there are multiple images, but bleh, another thing for the future
lets see what happens
oh yeah, the comment above is the code i commented out for the time being
theoretically it shouldnt run on your system since its not a livecd
i mean
it works
It will
If Colin implemented the run on host thing it will
oh
fire
then i can wipe my disk from inside my OS
That's why I added the systemd-analyze
self destruction 100
we are literally doing that tho arent we?
GitHub
fix: use pkexec for podman images checking by tulilirockz · Pull R...
The installer cant find the images if we dont run podman images as root :P
then we merge the release and rebuild the ISOs
Yeah, and more reason to not comment out the systemd-analyze
yeah tbh
what systemd-analyze are you talking about tho
see the commend in the disks.tsx file
ooo
put that back in
before the bootc command
yeah
Wait
Maybe needs some tweaks
it works btw
Ok no nvm
when i ran it on my laptop it detected the non-live env
and told me to f off
Yep then it works
just put it back in the multi line block before bootc install @tulip
we should have a variable for debugging tho
YAI_DEBUG_BREAK_MY_SYSTEMAnd env var yeah
Idk how you get to read these in JavaScript, can't remember
just detect the process.env.NODE_ENV
or wait
hm
not sure if that works
but there is definitely some way, since the devtools also detect a dev bundle
sigma momento
?
im just happy
That won't work
why tho
whats wrong
You are missing $
LMFAO wow i didnt notice that
need some glasses
... im wearing glasses right now....
miopia is so shit
Also, don't know how if env vars get inherited like that in tauri
Probably does
probably yeah
else $PATH wouldnt work right?
no idea
the pkexec podman images didnt work
no idea why its not being called?
does it output anything in the command line?
i dont think so
nothin
might need to debug this later
Wait so it kills the entire app?
no i just killed it
Ah ok
Needs to add console.logs or something if an error happens then
yup
glad we can actually do that
big anaconda error hater
Mmm let's see if bootc is more verbose with the errors
I hope
even then
we can manually identify the errors
and stuff
its just going to be better in every way
YO
YO
GitHub
fix: embed flatpaks and containers again into the ISO · ublue-os/t...
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
ISOS WORKED
Why wouldn't? You got it right :v
add aurora in there
then we have full coverage
its already there
just missing aurora-lts now :P
Add bazzite
also there
oh yeah dont mind me
O
i didnt see it
and bluefin:stable / bluefin:lts
everything just worked
i need some new glasses
🔜
Oh wait but without the flatpaks right?
with the flatpaks
and container storage
But bazzite flatpaks I mean
squashfs
the checks dont do that
i added tons of options for the
actions.yaml file
so we can manually specify the flatpaks file from the repo
and compression n stuffdo you want a bucket to dump some isos in?
ill merge this and re-spin the bluefin LTS ISOs
@Zeglius another ack? https://github.com/ublue-os/titanoboa/pull/36
GitHub
fix: embed flatpaks and containers again into the ISO by tulilirock...
This should allow for much better compression, although I have been getting a few local errors with this for whatever reason
ACKNOWLEDGE IT
or else...
/s
pretty damn happy this worked
i can see why
Checking rn
you shouldnt be sad when something works
:clueless:
released another yai version with the fix
it should get pulled in by the ISOs
Oh nice we got rid of the nasty fstab.sys
yeah LOL
is anyone against me merging this?
There is a wording issue with the flatpak part on the action
oh
just write what you feel is best and ill paste it there
Check the PR comments
done
yai release done
so we just need to merge this and rebuild ISOs
done too
GitHub
Build ISOs · ublue-os/bluefin-lts@1169772
Bluefin LTS, built on CentOS with bootc. Contribute to ublue-os/bluefin-lts development by creating an account on GitHub.
At how many reviews does this auto merge?. Can't see in phone
i just merged
there isnt a lock yet cuz the project is so new
We need to add one then, we are close to be stable
true LOL
And also because i don't want to wake up and see all of sudden something like the container and flatpak squash separately getting merged :v
That was really nasty workaround
is it new?
or the old one
old right now
when https://github.com/ublue-os/bluefin-lts/actions/runs/14059813380/job/39367579376 is done its gonna be the one with yai
GitHub
Build ISOs · ublue-os/bluefin-lts@1169772
Bluefin LTS, built on CentOS with bootc. Contribute to ublue-os/bluefin-lts development by creating an account on GitHub.
and its gonna be smaller cuz squashfs
well
Good call with the squashfs btw. Read an article, and seems only is smaller when fine tweaked
Gains are marginal
@Zeglius maybe theres something funky with the way we did the hooks
?
the
YAI_LATEST variable is expanding before it gets to the containerOh
Lemme check
i think its just straight up getting expanded here
O
Forgot at $?
Oh yeah definitely missed one, weird
Is not that is getting expanded, it seems just doesn't get passed to the hook
mind PRing it?
or just telling me how to do it
Yeah just booted the laptop
aight
Wait rebooting, the greenlight got me to a previous image
Done
Let's wait and see if first builds the image
GitHub
Build ISOs · ublue-os/bluefin-lts@6305ea7
Bluefin LTS, built on CentOS with bootc. Contribute to ublue-os/bluefin-lts development by creating an account on GitHub.
Oh I meant in the pr
It's not in main still
FIREEEE ME UP BABY
Another set of workflows to setup for Aurora Helium :)
Oh wait I see why is failing, at the var check
Is not even the same name
ISOs are updated btw
4.1GB
based
@imbev you can just copy the bluefin LTS workflow i think
for ISOs
its still being weird for whatever reason
?
Ooooh I see
Yeah the heredoc wasn't quoted in the action, lemme finish lunch and will fix it in the PR
i might have merged it prematurely
Nah is fine
Wait why did you do that
https://github.com/ublue-os/titanoboa/commit/d4a830a19f205e529721ea77f1062f2364659933
GitHub
Revert "fix: faulty hook-post-rootfs checks (#37)" · ublue-os/tita...
This reverts commit 0b8cdd5e7acbbb1db41700f53f82e281b9262abf.
i dont think the script ran :(
WHAT
Bruh
i clicked it then i closed the tab
LMFAO
Yeah we need to setup auto merge asap
alright so
oh yeah @imbev another ping, I'm gonna add the tokens to the cloudflare bucket
gimme a second
the script seems to not have run at all
I mean
for whatever reason
Wdym
it shows up here
then we dont get like any dnf or anything output
ill try to reproduce this locally
Yeah, revert that revert first
it didnt merge
why file descriptor 3 tho
i think you can just use the stdin
Was planning of adding more of these, but yeah, let's change it to 1
we might want to just dump those into files and read them from there
myabe just changing /dev/stdin to
< 3
or something
would workI mean one or other, you just said to do two different things
You know what, yeah let's make it read a script file instead, lemme see
dunno
just braindumping
aight @Zeglius cant use the 3d file descriptor
maybe
Yeah tweaking the recipe to pass a script file instead
Less shenanigans
also
what the hell
why is the final ISO so much different
ok booted into the new ISO
how do I run the installer app?
did you see how different it is
it doesnt have the installer yet :(
you can install it tho
lemme get the URL
np
add it tulip :devilpanda:
it's so magical!
dnf install -y https://github.com/ublue-os/yai/releases/download/v1.1.2/yai-0.1.0-1.x86_64.rpm
on the terminal
then launch yaiit might break
it looks so fancy for whatever reason
but if it works
lol
that string is so linux
i wanna change that logo....
YUP 😭
yeah make all that shit our artwork
but whatever let's close the loop first
because even if it's ugly it's working
i think it got the DX version as the base for whatever reason
yeah
this is so awesome tho
looking really good
this is gonna be so good!
the ISO is just so much smaller, too
with the squashfs compression
oh I know
if this is dx!!!!
looking like it :)
sec, these stupid dnf mirrors keep timing off
vscode on a live ISO is kinda crazy LOL
it is DX
oh man
the qemu guest drivers are working
LES GO
OMG IT WORKS
HA NICE
oh I don't know what to put in as my password
what
its installing?
@Niklas ⚡ @Zeglius be proud you guys did an amazing job
OMG
IT IS INSTALLING
does it give more output?
like more lines
because it should, theoretically
what password do I put in it's asking for a sudo password to install
same password that you put at the start
i have no idea why that is showing up
I NEED ANSWERS @tulip
it writes a few things
then it gets stuck there
cuz its installing
yeah okay
fucking sick
heh i did something wrong
your fault
we gotta have logic for if the install gets canceled
unacceptable.
yeah
omg it's doing it
oh its absolutely
-dx
thats the container thereoh it's accepting my password it's just adking a lot
it has to ask at least like two times
i think
we'll make this invisible with polkit rules tho
more like... polkit runes :clueless:
we set up the user account on first boot anyway
can't we make this just be root or something?
i dont think so
it shouldnt even have any users on the ISO
we should've gotten the same thing as before
what happens when you do "sudo yai"
that "try or install"
fails
well, looking great so far!
oh well
I never prompted try or install, I'm sure we'll figure it out
great progress let's goooooooooooooooooooooooo
LES GO!!!!!
I say wire it up and just get the functionality down
now we make it beautiful
even if it's ugly
yeah
error handling and the like is needed
yeah I wedged mine
but for a first prototype that its working somewhat is funny
this is awesome progress
absolutely
in less than a week :)
the reason we need the loop closed is then we can tell people to get it every day, run through an install, and file bugs
we now have smaller live ISOs for literally everything. hell yeah!
we only need a handful of people to start but once we get things going in github it'll make adding to all the images and stuff easier
MAN @Niklas ⚡ we can do something like a little terminal window thing
that expands or not like the ubuntu ubiquity installer
that was my plan
perfect LOL
a scrollable container that contains the shell output that comes back
absolutely yeah
lets get this going
this is so cool tho 😭
installer written in tauri
i can see the old linux guys foaming at the mouth
oh i want to see that
i would be laughing my a** off
:p
lets make this actually really good
then when these dumbasses test it they'll see its awesome
nothing less than perfect is enough for me
thats them
oh what
man the install was super fast
the artwork is going look so good
thats also why i rewrote the aurora website, i hated it
huh
this is going to look so good with the image artwork in place
is it actually done?
yeah
it really did finish
LOL
wow.
what the frick
rebooting
sir, we have achieved installer internally
yup this 1 billion% worked
does it boot?
LOL
we effectively closed the loop
just needs a bit of pruning
i really wanna test this on a USB drive
to see how fast the install is
so this means i can use my test laptop to install with my own installer now
what the frick is this timeline
yeah LOL
me with the disk
these effects look like youtube poops from 2010
youtube poops were great
better effects than todays tiktoks
tiktok good too
no
you just arent on the cool kids club
i am
they have the big funnies there
who wrote the installer
:hmmm:
nah JK
edwin
i have no clue what that means
not on the cool kids club
im too old for that
as shrimple as that
ALSO
this live ISO stuff means we have nvidia drivers on the ISO too
kernel modules and everything
yeah
this is super handy for testing
people wont have a shitty OSS driver when they use a livecd
hell yeah!!!!
oh @Niklas ⚡ we also need to copy the flatpaks to the target system
we might want to do that straight from rust cuz then we can implement some cool progress bar for it
yeah
if you write the rust code
im all for it
fine bah me
i wanted to do the dbus stuff so that we wouldnt shell out...
but whatever right
ok so the install actually works. 😄
yup!!!!!!!!
Eh, whatever is easier for you
not using lsblk is much harder
couldnt care less
lets get this going
make it work, make it right, make it fast
make it beautiful
make it awesome
make it SHINE
man we can include the cool lil games on the ISO...
doom while we blast your disk
how about that
as shrimple as that
maybe i need to start using tulip slang when i want to be in the cool kids club
Seems like a game of snakes is appropriate given the project name...
@j0rge mentioned wanting to put the snake game there LOL
yeah, a nice libadwaita snake game would be perfect
don't be obnoxiouis though, add a button that's like "Kill some time"
like don't auto open a game
Universal Blue
Feedback on the initial user experience
Hi 👋 I have been a Mac user for over twenty years, but moved over to a Framework 13 AMD laptop around a year ago. Since then I have been dabbling in different Linux distributions, and I was surprised and amazed at how much better the experience around ease-of-use and stability has become over the last few years. I love the work that is bein...
yup!
oh, moving to bluefin-dev
Peggle. We need to compete with Nobara 😉
just put straight up doom on the bazzite ISOs
freedoom
@tulip Btw you probably want to update the bluefin-lts iso workflow when I merge this PR
yup
no prob
is going to use a path to a script for the hook
ill need to update it anyways
@Kyle Gospo @Niklas ⚡ @Noel what if: https://silentspacemarine.com/
Multiplayer WebSockets Doom
Multiplayer Doom on Cloudflare Workers
@tulip any reason why you made these exported?
dunno
just seen that on other thingies
i also wanted to not use interpolation
cuz its easier to parse
?
you are doing it already anyways
{{ livesys }} -> ${livesys}
yeah i didnt refactor it allk will do
though maybe is better just to nnot export them
why tho
idk, polluting env scope, though we dont use env vars
na will just refactor to ${}
also will change this to squashfs
yup
sgtm
@Zeglius :dispair:
jesus christ this is painful
?
script didnt run at all
nothing on the logs :(
maybe its just some path we got wrong?
ah i see
its -n, not -z
:p
Push it, build it, ship it
OH
@Zeglius i know what is going on
we need to specify
${{ github.workspace }}?
where
sent a suggestion
on the pr
https://ubuntu.com/download/desktop
for bluefin we're like a gig under this, we're gonna slay
🔥 🔥 🔥 🔥
@Zeglius LES GOOOOOOOOOOO
alright now is working
mind if i merge it?
or is there something else you wanna do there still
too late already merged it 😎
😭
bah, its working anyways
Tulip is too fast
At least only one time something broke with a lockfile
I say "yai yai!" like Ice Cube does in We Be Clubbin'
And grub templating done https://github.com/ublue-os/titanoboa/pull/39
GitHub
feat: Add grub templating by Zeglius · Pull Request #39 · ublue-o...
#28 But with automatic templating based in the os-release of the image using PRETTY_NAME
so dumb question @tulip I saw you added some more params. How would I build an ISO for Bazzite now with Flatpaks?
It completely missed the flatpaks when I ran it last night.
or @Zeglius
There is a new input,
flatpaks-list, you pass there the path to a file with the list of flatpaksIs the Flatpak file on the host or the image? Would it make sense to read the Flatpaks somewhere from the image (
/etc/install.d/flatpaks.txt or something)should be in the repo where you build the iso from
Maybe somewhere under /usr/?, doesn't make much sense to have it in a mutable path
Yeah usr would be better
I know bluebuild places their flatpak file in
/etc/bluebuild/default-flatpaks/system/install but /usr would make more sense hereIf we can wire up our own implementation of Flatpak preinstall.d, that would be best IMO. Using the files and directories they are using.
We then match upstreams
A link to the file spec would come handy if we want parity
It's still a WIP.
Issue: https://github.com/flatpak/flatpak/issues/5579
PR: https://github.com/flatpak/flatpak/pull/6116
Anaconda PR: https://github.com/rhinstaller/anaconda/pull/6056
I see the issue.
Mmm. I would wait for the flatpak pr to get merged. That way, we can simply chroot into the root filesystem and run
flatpak preinstallthe new flatpak install just script won't install using fully qualified flatpak names:
No need to reinvent the wheel
Ah yeah my bad, should follow this format
https://github.com/ublue-os/titanoboa/blob/main/src%2Fflatpaks.example.txt
GitHub
titanoboa/src/flatpaks.example.txt at main · ublue-os/titanoboa
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
how do you differentiate runtimes?
I don't think we can with the current setup(?) lemme check
Yeah seems is a bit hardcoded
https://github.com/ublue-os/titanoboa/blob/main/Justfile#L111
flatpak should take the full info rather than hardcoding anything.
or I can change them.
I see it's doing a lookup.
lemme make sure it parses the correct flatpaks
for Bazzite
yeah, it missed several flatpaks when I switched from the less verbose form.
GitHub
Make include-flatpaks step less messy · Issue #40 · ublue-os/tita...
The logic to parse the flatpaks list file is a bit overengineered, and won't work with more verbose lines like: app/org.mozilla.firefox/x86_64/stable I think we can achieve this by simplifying ...
we need this else aarch64 breaks
cuz it wont find stuff like firefox
the current one just makes sure the thing exists before installing
no idea why it fails so hard doe
Mm okay, feel free to close that one if you see it unnecessary
naaah
its good cuz i dont know why its breaking
like, it should jsut work with the
app/chungus/amogus formatWe can simply append the arch right?
Is a little more verbose but eh
mostly cuz of necessity
but its nice to figure out something better than this
For now I guess that will do
The flatpak preinstall would potentially solve that issue
would be a banger
What's the main limitations on filesystem choice for the live ISO?
Would love to experiment with Composefs, since it seems like the perfect candidate with all the deduplication
Would love to experiment with Composefs, since it seems like the perfect candidate with all the deduplication
No clue and probably we don't want to mess with the filesystem in the iso anymore
Has been already a pain in the ass just dealing with fat32 and the 4GB limit per file
right, but it should hard fail if it doesn't find an entry.
Besides we already have rechunker, we cant achieve more deduplication as far I know
ZFS TIME
In an iso
It's ok it's a meme
Just saying, tulip and intrusive thoughts are a bad combination, don't wanna wake up and find another wheel reinvented :dispair:
No offense meant
hahahaa
it's ok m2 already looked, it won't work
(not for the iso)
oh for the rootfs for bluefin?
Yeah. Was just thinking it's a way to learn both live ISOs and Composefs.
Composefs could maybe be used for both the container storage and rootfs, using the same backing store.
Not suggesting to change how we're doing things (unless it happens to be a huge improvement)
I'm seeing timothee next week, I can ask a bunch of questions
I mean learn is an understatement
Mostly because xorriso is esoteric enough
We are lucky enough to have it working somehow
the curse of el torrito
Yeha, torrito in the middle of salamanca
Anyways, we probably need to setup a rule so a single person cant merge PRs blindly, we are already reaching stability with this thing
jesus christ i hate xorriso
Eh. I guess ISOs in general
not really like theres not a single reason why this thing should be so bad :(
I mean
i guess no one cared about it for years now so we are cooked
Isos have been already out there for a while so, tooling do be building rust over time
[...] tools do be building rust [...]
I can do this
what's the test matrix thing look like?
like what do I need to gate on?
same kind of thing as the images
Gate on the "Check all builds successful"
you just gate on the check at the end
yup
k
ok all set
now we are good to go
ill fix up the bluefin LTS workflows now
hopefully thats all we need to do on both projects LOL
now its just making yai amazing
Tbh I just making my own installer in Godot for the funnies
Can't touch JavaScript would having an headache :dispair:
do it.
its not like it would matter how its written LOL
There's already an installer in Godot get in line
Or use python like a normal person
it would be so based to have an installer in godot tho
i wanna see it
Hum. You are assuming too many things with that statement
https://github.com/ShadowBlip/gamepad-os-installer marvel at it
man they dont have a screenshot...
Rust underneath
🫠
so based....
oh man
@Zeglius why dont you have a github sponsors profile
you 1 billion percent should
dunno if i already asked you that
Dunno it feels like it would be a pain in the ass to setup
its super simple
and you are doing amazing work
you should get money for it
Trafotin
YouTube
All FOSS Devs Need to Start E-Begging.
It's hard for open source devs to make money, but it's even harder to communicate. Devs becoming content creators might fix both of these problems.
Website: https://trafotin.com
Donate:
✨ Patreon: https://patreon.com/trafotin
▶️ YouTube Membership: https://www.youtube.com/channel/UC5U7mHlhP6s6478wd7ZvnhA/join
Links:
▶️ YouTube: http...
@j0rge
bluefin LTS and everything with yai for all the flavors should be on the R2 bucket now
hell yeah.
NICE!!!!!
snagging it
its 100MiB smaller :clueless:
somehow
alright so
the ISO works, the installer is going strong too
i just have no idea why the "Try Or Install?" thing isnt showing up
and i cant seem to find it on the gnome-initial-setup source code
if someone could help me out there it would be super awesome
I get stuck in the password prompts
afaik it should work just like a regular ol install
maybe theres something wrong with ur keyboard input source
no it's just there's too many lol
oh yup
they'll go away
it didn't default to english though I hadn't seen that before
we just need a few pkexec rules
same
thats different
Is it doing install-to-disk or are you doing a custom partition scheme?
install to-diskProbably we can setup something like systemd-repart config for the partitioning?
Yeah, language support is something we get free with anaconda.
There's gotta be a standard for that I would think for GUI apps.
Also for YAI, we would probably want a way to pick language the image installs with.
yuuup
it worked just fine tho
the gnome-initial-setup is a bit bugged out
there's no way we can skip that rpmfusion asking question is there?
there is! i PRd it to the bluefin package
im just trying to figure out why initial setup isnt thinking we are on a live image
oh man
ill rebase it locally, github webui hates this :P
also
we did this in 6 days
@j0rge
this is very impressive tbh
indeed! We're almost done!
every time you say that you curse us for more 3 months of development.
let's blame kyle for this one
True
alright ill sleep now
last thing we need to figure out for titanoboa is how to get that damn "Try or Install?" session again
no idea why it isnt showing up anymore on bluefin ISOs
everything else is done, we are good to go and we just need to work on yai now
The way titanoboa does the extraction of the image it will only work with rechunked images
i am all for it
Wait what. So without it we are screwed?
The "extraction of the image" being dumping the contents of the image into the rootfs?
If so, I'm sure there's other ways we can do it
I'm guessing he meant something else
Like the file tree structure or something? Idk
Doom with everyone else currently installing the OS would be epic
I told you before
You need to use rpm-ostree to pull the image
Into an ostree store
And stage it
We're using bootc, pointed at containers-storage.
Why do we want to use rpm-ostree?
Okay, let's say we don't use rpm-ostree.
What would be the consequences
Because the feature is not ported into bootc
Colin still uses rpm ostree for this stuff
So we lose a bit of deduplication or something?
No
If the image does not use rechunk
You pickup the original commit in the iso
So for bazzite you would end up installing kinoite
o
You need to use ostree to deserualize the image into an ostree repo
Stage it into a root with hard links
Do some moving
And then do what you do now
Doesn't bootc handle non-ostree images now though? I.E. it does the conversion on the fly?
The plan here is to use ostree for deduplication no?
Bootc only supports containers
I don't think we are deduplicating? (looking at you, Zeglius 🙂 )
I mean, tbf, if comes with deduplication, it would nice
How is the iso 6gb
But yeah, no. We want something that boots up
While being offline
I guess ostree deduplication yeah
Well to do it properly, rpm ostree, stage, put commit in there
Then do some funnies with etc
Nah, out of the scope
I'm building a silverblue-main:41 ISO. So from what I understand, we shouldn't have anything ublue has added once booted?
We are aiming for MVP
Iso will work
Installed image will be kinoite
Silver blue
Ok, so after i
bootc install it to a disk?What gets bootc installed
Did I miss something?
bootc is installed by default
We run
bootc install to-disk <image> currentlyWhat image
we pull the container image twice
Once for the root file system
And another into /var/lib/containers/storage
Of that same root filesystem
bluefin-lts was installed yesterday by tulip and jorge successfully
with bootcinstall
The image we want to install.
So if I wanted to install silverblue-main, CI pulls silverblue-main, injects it into rootfs and we run bootc install pointed at that local containers-storage
In theory you can use a completely separate image just for the root filesystem and other for the containers storage
So we are not benefiting of deduplication anyways
Yeah. If "bootc install" works, then I don't see why we would end up with a different image
Otherwise that would be a bug
So yeah, we are not using ostree in theory
but i think no one has tested installing bazzite or non bluefin-lts yet
with the new installer
We could probably remove /sysroot and /ostree from the root and would probably work just fine
Rechunking (both hhd-dev and upstream's implementation) moves everything to /sysroot if I understand correctly.
Yeah, as hardlinks
You can remove these and still have the files in the expanded root filesystem
Oh, cool. I didn't know that!
So yeah, no need for rpm-ostree
We have been using only podman from the beginning and wasn't an issue
And squashfs compression is good enough to deal with the deduplication overhead when comes to size
I tested installing Bazzite with bootc install and it worked fine.
its squashfs deduping the shit out of the rootfs + containers storage
we are doing this the right way
worst case we can rechunk the images with the rpm-ostree rechunker insteas
theres nothing wrong with our approach
Maybe, anyways regardless if we are doing it right or wrong, the maintenance is sustainable
We can keep this thing alive compared to lorax and kickstart recipes
it does
it just works
i tried it out
Yeah, did as well
No changes in the iso size
@Zeglius i think we'll need to do that podman pull instead of push
cuz the gh actions runner hates it when we have multiple of the same image
maybe we can check on the container storage if the image is already pulled
dunno
！
@tulip @j0rge what branding or looks would you guys like to see in YAI, as in images displayed during install, button colors, something else that might come up
tbh i feel like we should make it as clean and simple as possible
lets just get it working right now then we can style it nicely
I don't care about art right now, I want to ship a working ISO as fast as possible
so as long as the buttons line up everything else we can do later
im trying to fix some stuff on the yai installer rn
yeah I just need next->next->done. mostly
True
Works for me
@Niklas ⚡
we should use this.
for the terminal.
Drawer
A drawer component for React.
but eventually (for bluefin) it should look like the website, just the wallpapers, etc.
doesn't need to be over the top, I'm sure bazzite will do the bling
are we doing FDE with a password setup in there?
not yet
but can be done
cuz bootc install has options for it
nod
I wanted to rearchitect the UI a bit and adding dark/light mode support in the process for the future
so
Sounds good
the light/dark with the wallpapers will be baller
our objective right now:
- clean that disks part a bit cuz its kinda bugged out
- make a lil terminal part on the bootc installation with that cool drawer component or something else that works for it
- add a page (before installation) for user creation if configured somewhere (necessary for KDE)
@Niklas ⚡ ill try to experiment a bit with the lil terminal thing
okay?
Yeah
Do it
We should get these listed out but: (in no particular order)
1. User creation
2. Setting Time zone
3. FDE Support
4. Setting Language
Setting language should be the first
Cuz of keyboard layouts
true. Keyboard layouts.
Hypothetical question
Do you think we could
bootc install to-disk --via-loopback to the squashfs.img file on the ISO? This way, we'd be using bootc/ostree on the live ISOyou dont want that
otherwise you end up including Livecd dependencies
and end up mixing
Plus removes our capability of being able to separate the livecd runtime image from the one to be installed
I think I found the goodies
😳
oh even better, we get to reuse the keymap set at runtime
nice.
this is really good LOL
what does not have systemd at this point? lol
not a bad thing 😈
systemd THE GOAT
systemd-repart for configuring partitions for dual booting too? 🙂
yup
Ooh! systemd-homed for the users!
we dont even need to mount the partition
selinux is still a tiny bit funky with this
we'd need to wait just a bit for it to actually work well
Kde isnt compatible as far I know
and probably matching upstream would be the best
afaik anything that uses accountservice is compatible with systemd-homed, the problem is when you dont have any users except systemd-homed on the system
GDM gets mad and boots you into gnome-initial-setup
SDDM i dont know
you also need
authselect enable-feature with-systemd-homed and restoreconf /var/lib/some-systemd-homed-directory-hereMaybe for the next installer rewrite in F44...
Or whenever someone ditches anaconda
not really, its gonna be super easy once those are done by default on GNOME/KDE
:)
I cant imagine anyone wants to maintain that
What was switching from podman push -> podman pull to rootfs supposed to fix?
https://github.com/ublue-os/titanoboa/pull/50
noooo dont revert it
podman push is faster if you already have the image in your host
Don't redeem it
Oh. I thought you said we wanted to switch to pull?
podman pull can be used if we dont have the target image to install to the rootfs
bazzite:stable for iso root, bazzite-deck:stable for target installation for example
if we do pull then push we have about 3 simultaneous images on the runner
well the thing failed anyways
but it mostly worked
i think we can clear the container image from root storage before we create the ISOs
at least for the gh runners
from container storage?
where? in the gh runner storage?
containers-storage, /var/lib/containers
we can make it like a hidden option just for gh actionsif [[ -n "${CI:-}" ]]; thenfine bah me
@tulip broke templating
LMFAO I FORGOR
Thats why you dont push to main directly
I Did Not Push To Main Directly.
Neither PR and instantly merge :v
FR THO I DUDNT DO THAT
Oh ok was Jorge giving greenlight nvm
my bad I won't do that.
but it was green tho
I mean yeah, there are no checks for grub entries
how do we do that tho
we cant
unless you want to go and run qemu for that
@Zeglius we should clear the image around here
cuz then there wont be any more steps using the container
i think, right
after dumping into containers-storage
basically like this
wait
ok that one shouldnt be
now now
also random niche trick to set default values to empty vars
noice
do you need the
:?without it
: , wont act if the var is set but empty
basically is "act even if the var is an empty string" of sortsnoice
ok so
@Zeglius do i merge mine?
seems that you just fixed what mine mas missing
Which one?
th
o
GitHub
fix: add polkit rules for podman and bootc privileged without passw...
Will make our work a lot easier for installers! Just lets pkexec bootc and pkexec podman run without a password on the live environment
Fixes: #43
the polkit one?
yee
I mean yours seems to be failing for some reason
its literally what you are fixing right now
we can just merge this one then i merge yours and bam done
Alright
bam merged
BAM
lets see if builds now
for the KDE images
fireeee
ill make the disk encryption prompt later too
shadcn is the GOAT
dump it in there and well adapt it later to dark/light mode :)
yupyup
this thing is looking so fancy oh my god
I WANNA SEE IT
you'll see it.
Another tool for the installer, this is from
useradd, we can simply run sudo systemd-dissect --with /dev/rootpart useradd --root . <whatever> to create the account
localectl to list/set the keymap at runtime alsomy disks.tsx literally just reset
@Niklas ⚡ what the sigma did you know that could happen?
reset?
it literally just wrote "hello /disks"
😭
uh
that should not happen
yeah thats really weird
well im rewriting what i did
now i know how to do it so its gonna be faster
heh
the router normally adds some boilerplate code to the file if you create a new route
automatically
but not when youre writing code in it
oooooh
prettier -w made the router think the file was cleanohhhhh
ooof
dang
Ok now I understood why around there
heh there you go
gonna delete also that image around there
ran out of space in the bazzite one
@Niklas ⚡ it fucking did it again
i managed to undo it
but we might want to disable this
dont use prettier
heh
prettier is just the GOAT tho
not formatting is CRINGE
oh LOL
ye
are you formatting the entire project or just one file??
@tulip
i might have formatted the project
You let tulip loose, first mistake
yeah, so if the routetree gets formatted it might get regenerated
🤣
and thats when your new code shits the bed
bruh, thats a new one
add a .prettierignore file and then add
src/routeTree.gen.tsLMFAO GOT RATE LIMITED
I mean
They ratelimited themselves
¯\_(ツ)_/¯
and try your formatting
wdym?
i just use the ide with webstorm for this, but well you cant really do that with inferior helix
and precisely in the bazzite one...
well try prettier -w and see if it fucks up the code again
i sure hope not
oooh
you can tho
LOL
can i retrigger manually the runs? i guess first need to cancel the workflow
lemme retrigger them
this in the root of the project obv
lets see when i regret it /s
Whenever she switches to Rust for the entire frontend :v
she can do that if she wants
but then I will just be moving back to a mac and continue on with life
/s
okay restarted the runs
no worries ill do that later
on my cool tulip installer 2 thousand
leptos as the frontend framework so we can have 100% RUST
RAAAAH
have fun maintaining that
is Html at the end :v
2 html tags then a LOT of wasm :P
Is just 0 and 1 and the end of the day
I mean rust for html sounds cursed already
not if i write the rust on paper
RAAAH
Oh god, taking a test in paper in rust
New medieval level torture
LMFAO TRUE THO
i had to write java code on paper in school
same
so whats your excuse
@Niklas ⚡ how did you separate the buttons on the install section
the
go back and install onesturbofish
im trying to do it my thing is not liking it
A flex container
a flex container is the answer to everything
i wanted to not give you a spoiler but fuck it
@Niklas ⚡ not working...
there might be some custom styling for the dialog footer section
yeah dont set the height in the footer
wha
eh width
it blows up if i do that LMFAO
maybe i just need to not style it?
maybe its also webkit fucking you over
dunno
can you make a branch
and gimme the code?
so i can try
no i dont want to give you MY code.
(PRing it)
Is CSS really the best answer to styling stuff? Why is centering a div so hard?
yes
centering in CSS is cursed to this day
its actually pretty good after flexbox started existing
Also did you guys try with the zed sharing code thing?
actually nvm, it will get rust stuck recompiling all the time with you two writting at the same time
oh yup i did
its pretty nice
we should do it actually damn
It’s not hard
Once you know it
same can be said about rust
:clueless:
Don’t compare rust to css
Or I’ll remove your shadcn rights
UI is done for user creation and FDE
timezone can be set automatically from the ISO with
systemd-firstbootwhere my branch
maybe the language stuff too
BRANCH
gonna commit and pr right now
GIMME BRANCH lol
ok am I merging this?
Also, did we simplify the flatpak install script?
@Niklas ⚡ https://github.com/ublue-os/yai/pull/24
GitHub
fix: format all this and add user/FDE pages by tulilirockz · Pull ...
I forgor and formatted literally everything with prettier
not yet, you can try fixing it if you want to rn LOL
i have no idea how we are gonna fix that
"900 lines changed"
😭
i rewrote the entire project
:why:
this is so fancy 😭
undercooked but its going strong
true
very fancy
OOO
we just need route transitions...
i must admit
so do we just need a checkbox for FDE?
check out the video
but the spacing on the last dialog with the buttons in a position which i cannot condone
skip if you dont want the encryption
Oh forgot, gonna add a list of TODOs to yai
hmmmmmm....
i cant f ucking fix it 😭
FDE by default >>>>>>>>>>
I don't know if I like skip because it's not clear what you're doing there.
"skip encryption" then
i dont know
encrypted by default isn't a bad way to exist I suppose...
its very nice...
we probably will need to change this when we do manual partitioning
unacceptable
just kidding
imma fix it
maybe
i managed to maybe kinda sort of fix it
yes please before noel keeps adding things
oh wait can someone enable subissues in yai?
I probably won't have time tonight, but I might be able to poke at it a bit tomorrow.
on it
check your inbox and then check here again
why are the buttons empty @tulip 😭
erm what the sigma
i am kidding, thats because i have light mode in my local cache somehow
lool
Yep, enabled
also the skip button doesnt work
you may ... indulge yourself
did all ya mess with lsblk :v
none work yet
yes
actually i dont think so
yup
needs wiring up
i expected a fully working installer
i just wanted you to fix the button issue for me then i wont need you anymore.
Looks just like how I left it
nothing less
oh wow
okay
Oh well needs more finetweaks in the filter
no need, ill be installing bluefin to my zram anyways
Yeah, and be sure to mount GCloud s2 as your ram
compressed gcloud s2 :clueless:
Why is there a grid in the dialog
i copy pasted it from a shadcn example
thats most likely the issue LMFAO
although there isnt any grid directly on the code
in their own example, there is no grid in the dialog
Dialog
A window overlaid on either the primary window or another dialog window, rendering the content underneath inert.
edit profile one
erm what the sigma
we might just need to use grid instead of flexbox
I'm gonna put it in a seperate component now, because this code is getting messy
also the drawer animation is sooo different in webkit omg
sigma made it
make the apply really big
we dont want people on non FDE
non FDE?
non-full-disk-encryption
who needs that pft
theres no performance penalty and its very important LOL
true
when you sell ur hard drive you are gonna wish you had FDE
why would i sell that
actually when you trash it
I'm way too paranoid
same
and i will just throw it into a woodchipper
or smth
where i know its destroyed
based momento
mind commiting this to my branch?
also make the wording lean more towards "please encrypt ur disk" LOL
"Encrypting your disk is highly recommended, if you want to encrypt it, blablabla"
i dont know
LOL
there
LES GOU
px-10 instead of p-10 :clueless:also gonna properly style this with dark and lightmode since you didnt
i dont kno how to do that
with regular ol tailwind
also, apply on the right cuz all the "confirm" stuff is on the right
for styles that should apply to dark mode, use dark: like
dark:text-glorp
yeaGLORP
goofy gooba
https://github.com/ublue-os/yai/issues/25 I will think in what else to add after dinner
oh yeah flatpaks
true
oh man also
something i wanted to know
do you know if we could put the buttons all on the same-ish place?
dunno if thats a pain to do with css
like this is just peak design
i wonder if we can put em all on the same places as the back and install buttons
if you make them similar width and height, yeah
we should definitely have the same height on all pages
the users page is slightly lower than the disks one
or we define a max width on the outer shell and you make a div container with width-full,
i think we can just not center the elements and do like a flex col layout or something
thats how that is done
the container in which the elements live in is now unified
you can go ham in there
noice
i know we are comparing ourselves with a incredibly low bar
but this looks just amazing
compared to like anaconda or those other installers
oops
i may have accidentally created a branch with your and my change
whatever
as long as we make this good
but it includes you
its both in there
whatever
yeah
my bad...
sgima-format-fde
Both of you rn
erhrmm what the sigma
whos the most sigma
i checked out your code with the github cli
and thats how that happenend
i guess
just pr it
me no care
just this
at all
Whoever's more of a branch gets merged in conflict
GitHub
feat: add dark mode support, new screens for disk encryption and mo...
This PR adds dark mode support, adds a disk encryption UI prototype and user creation UI prototype.
there
review it, merge it, ship it
the three steps to make yai yai
now i have a commit that says "make it more sigma"
i can die in peace
make them write a spec or it will be madness for us all
I mean I got somewhat of an spec in the issues
I mean, just saying, if you wake up tomorrow and there's a git snapshot of systemd in there, not my fault
LMFAO oh no....
:huh:
systemd-installer
for the yai backend:PES_Devil:
HHAHAHA
maybe youre using a systemd git snapshot
just dont know it yet
Yeah its called systemd-firstboot but without the useradd part
yeah then you'll be like "whose fault is it?" "Merged by Jorge Castro 20 minutes ago."
I mean I will get as nitpicky as I can be in the issues
based
ignored, not shipping
Other than that, Niklas has the fault for leaving tulip go berserker :v
wontfix + closed + L + loser
Also make that encryption dialogue only appear behind a "encrypt disk" checkbox
FDE by default
Ppl will assume is necessary to fill otherwise
forever and ever
make a
kde label so we know which ones are KDE specificFDE?
yeah most people dont want FDE lol
full disk encryption
Ah no
with tpm2 chungus
or at least the ones that dont know about it
yeah just leave it skip
Ah whatever, just saying whenever bazzite folks come and complain
ill change that around a bit
and not let tulip go berserk
popOS does FDE by default :clueless:
otherwise we have a rust encryption module in yai tomorrow
the joys of being based
yeah then i will direct all complaints directly to you
send it to tulip@proton.me
so fancy
Needs an icon on the right too
true
Good
And a second field, to confirm the password
sigma
didnt think about that
there you go
Not sigma
Also maybe an 👁️ button to show the password briefly
While held or a toggle or something
there might be some stuff for that on the form components
Input components built with React and Tailwind CSS - Origin UI
A collection of beautiful and accessible input components built with React and Tailwind CSS.
Btw, which is the current order of the screens?
Gonna set the order in the issue
And wanna see if we can avoid changing it too much from what you guys have currently
nah its fine
"welcome" -> "user (if KDE)" -> "disks" -> "install"
we'll most likely add more stuff
Yeah asking because I won't let you set the user one before the keymaps one😂
i think GNOME has that by default, itll be KDE+ only
No
Wait
that inital setup thing on the ISO does it for us
with the languages
maybe plasma welcome also does it?
With the live users package it doesn't ask for locales right
No
on GNOME it should do it
Yeah not in bazzite
We straight up remove that package
me and the zzite
Yeah you just don't open that when steam is the first thing that launches
oh the deck images are going to be on the desktop ISO i think
I can imagine people trying to install games in the live session
NO steam, NO lutris, NOTHING on the live session.
RAAAH
Yeah, but again, with the livesys packages, won't ask for keymaps regardless
Because the user is already set and the welcome thing won't showup
eh it shouldnt be too difficult to do this
hopefully
localectl does all the work for us
If not let's write our own
in rust ™️
with 2000 dependencies
No python. This is titanoboa
The saddest part is that is close to not being a joke
Also, add another confirmation password in the disk encryption
forgor about that one too
And the 👁️ button too
Close to buying some snapdragon device to test out bluefin lts arm64
But I don't wanna
ehrm just buy a phone :clueless:
i do wonder if you could install that on a phone
I mean, yeah but they will deal with the same issue than any livecd. Out of ram
Good luck with the cursed Android phone boot loaders
i can see people reporting bugs about this
id rather not have any gaming related thing on the ISO
or at least very few
Closed because of skill issue
we'll get tired of it after the 300th issue
like we do with the anaconda bugs
True
um also
linux has two user names
we have that fancy username that can be uppercase and stuff
and we have the unix shitty one
Will ignore emb
dunno exactly how we are going to work with this
Easy as that
Or better close as no planned
do you know how anaconda does it?
You mean the hostname?
Tulip Blossom, then tulip as my login usernameOh
i think we can make an automatically updating field that just generates a unix-like username for you
I think has something to do with GECKO in the passwd?
wha
GECOS my bad
https://linux.die.net/man/5/passwd
yeaah that thing
not super sure how we can handle that
Oh just take the username and lowercase then replace spaces with
-mmm
And remove any weird symbols
That's what calamares does I think (?)
anaconda does
Full Name and usernameon that I'm not going to be picky ¯\_(ツ)_/¯
Anaconda generates a username if you don't define one
yeah thats what i was thinking about
Wait
Fuck, does useradd not have a GECOS flag?
true
doesnt look like it
eh we can do it some other way
usermod -l
does it for usThere is also this one https://linux.die.net/man/8/newusers
The live session is space-limited, I couldn't build Titanoboa before we had the build
wdym?
Do you notice a strange 30-40mb device?
I'm not certain, but I think you're limited to that on the live iso
Or something similar
It may be a consequence of the filesystem being only a few GB
that thing was composefs
the live ISO doesnt have any restrictions, you can just do whatever you want
nvm, it must be the filesystem iso
just need enough ram afaik
this is so nice
Oh also, we need to show the user screen whenever the the image is not a GNOME one
yup
we need a configuration file somewhere on the filesystem maybe
Nah I think we can simply scan the image root filesystem for the xdg files of the session
Checkbox for strong password?
honestly its their system
i guess we should just let em not have a strong password
I agree completely, that's a major problem that I have with gnome
What I mean is an optional checkbox to check the strength
this seems more complicated than having a config file 😭
maybe like a env variable or something i dont know
Nah shouldn't be. Gonna try to see how
i think this kind of stuff should be overrideable somehow
Wait nvm
Yeah an env var or something
But how do we set that env var
/etc/environment on the live ISO
TITANOBOA_NOT_GNOME
LMFAOOh wait
Yeah we can do that in a step
In the livesys one
yup
or the polkit one
Can't we have a toggle in the justfile?
I mean
true that can happen on build time
Why would it be a manual toggle?
"if you use livesys you need this"
You don't know when is needed
But the build does when looks at what desktop environments are there in the livesys step
Nevermind, I'm missing context
Should it be possible to create multiple users from the installer? I can see it being fine either way
GitHub
titanoboa/Justfile at main · ublue-os/titanoboa
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
Yeah but we want to create one with uid 1000 first and foremost
This isn't a common use-case. I don't know of any installers that do so, so I'd say we shouldn't worry about it
But it's possible to implement
TITANOBOA_NOT_SHOW_USERADD?
Idk gnome feels too genericTITANOBOA_HIDE=user
for gnome
done
Mm that opens room to potential parsing in that var, aka more work for you
comma separated string
if it breaks its a skill issue
Okay
Maybe Im repeating myself but
mmm refactor rather than reorder
Wait nvm, we have the hook-post-rootfs anyways
@tulip set it to
YAI_HIDE=useroh true
forgor the name of the installer :P
Hmm, something broke for me since I last tried it a couple days ago
i have no idea why but it just sometimes breaks
rebuild it from the iso step
it should work
not particularly sure what causes this
mught be the fallocate on the iso step
wait what?
is there an issue?
theres some race condition that causes that
99.9% of the time it works and it seems to not happen on the gh actions runners
need to file that issue
I think the disk path is hardcoded
ooh i think i know it
i think sometimes dnf5 fails to install the grub modules but the step doesnt fail anyways
Plasma welcome is also having difficulty launching yai, although I could run it manually from a terminal
it should use whatever disk you selected from the installee
I'll file the issue
https://github.com/ublue-os/titanoboa/issues/53
thanks!
i think if we just do that bash strict mode thing itll work fine
maybe a check
Try with
dnf install -y something || dnf reinstall -y somethingEhh... That mostly depends on if they have aes extensions.....
Also even with those there is still a bit of a higher CPU load.
Also depending on the luks flags it can get pretty bad
Does the step have pipefail?
nope not strict there
I'm blind or something can't find the grub installation step
Anyways cant even focus at this hours at night 🫠
Rebranding as in, in the installer?
Yes, yai
Should yai branding be placed on the image or in the iso build process?
That would be more of an issue for yai itself.
Titanoboa is just for the iso building
@Niklas ⚡ i might not know how to set state
zustand is confusing
man
this installer is turning out so nice
gimme some new screenies pls
i've mostly already show the GUI stuff
just doing some work under the hood so its actually pretty good
heh
hey I have a dumb question
mm?
if the user screen is used in gnome
like if we did that here instead of in gnome first run
could we do something like install brew packages after they finish this screen?
oh the gnome first run shouldnt even prompt for a password
its bugged out i have no idea why
conceptually
I mean
we could really do anything on the installer
anything you mind can think of
like if we could also install brew packages if we created the user earlier here instead of in the gnome first run
tulip
can this be real:
[ ] Enable Bluefin-CLI
we'd probably jsut wire it up with
ublue-user-setup
it can be real[ ] Enable enhanced terminal experience
I have to workshop this
you gotta write down all ur ideas
we can do so much stuff now
vertical integration ftw :)
yo because I'm going to have that box ON BY DEFAULT and take our rightful place as the leader in uutil technology. powered by the future. And rust. Mostly rust.
lets literally just do that im completely fine with it
AND THE FONTS TULIP
THE FONTS
makes total sense
we'd just have the installer create a single file on
/var that tells the ublue-user-setup that it can install the stuff for youthey'll go in .local/share/fonts, it's perfect
/var/cache/installer_state.yaml or something :)
having this kind of integration is just so wonderfulok next question
how about a:
[ ] Enable developer mode
and then you inject the "is dx or not?" into the final thing, even if it has to fetch new layers?
i think we could just make the system rebase
on first startup
nod
:P
ok I'm filing these for non mvp
yup!!!
write all of them down
.
yai repo?
yup
LES GOO I MANAGED TO INTEGRATE STATE INTO THE USERS PARTTT
this was a lot more confusing than i though it would be
JavaScript and their frameworks
Is it possible to chroot into a bootc installation? If so, we could run homebrew from within the chroot
It is possible in theory with something the systemd-nspawn
yup
we can do it
either bwrap or nspawn
If so, the installer might need to be network aware
Yeah, need to setup an internet connection
Probably after we reach something working
That could result in alot of scope creep
Maybe a prompt to "Use your system settings to configure your network"
Before proceeding?
System settings?
in gnome at least there's some api thing to ping
e.g. KDE Network widget on the taskbar
I mean
and then you make a button that launches what it tells you and it opens the network thing
It's a full desktop environment
As long you set it up, may work
Similar to how Windows used to allow local accounts if the device had no network connection
"Yo click on that network thing on the bottom right, then click ok up here. Good luck, have fun."
heh
Is it possible to suggest based on the full name?
doing that
by default
IIRC, rhel suggests FLastname, while some other distros suggest firstL
then if you really want a different username then itll let you
While also sneakily telling the user that we're unix-like 🕵️
i think this looks nice
What is the gear for?
Is that a menu or a generate button?
it looks so good, we went from awful to going to be so goooooood.
webtech from the start, brilliant!
this just makes me so excited 😭 this is AWESOME
YO now i understood why bootc install with FDE doesnt have a password!
its just FDE with tpm2.0 by default!!!
alright
user section and disk encryption sections are working perfect now
we just need the installation to work nicely now
@somebody once told me yeah
what?
we are making a installer for the live ISOs
I am aware
just figuring out really what to include as a lil fun thing during the installation prompt
lol
tbh just let the user do whatever
speaking from avid distro hopper that probably broke a USB drive from installing shit too much
if they wanna play games, kewl
if they wanna breathe, kewl
I would just let the user walk around and see what it's like
also bro responded to a question in an entirely different thread
i've never heard of this "kewl" thing
no idea what are you talking about
another way of saying "cool"
kwel!
my older bro says it like that lmao
"kewl beans"
...why does that sound like a type of bean
pogs
kwel bens
*kewl 🤓
kewl
noice
ok I admit it
it's true
I actually wore out a USB from installing Linux too much
godspeed random SanDisk I found in my drawer
why does this make me think of the macOS installer
LES GOU
XTERMJS
Fireeee
nklas
why are the lines not getting wiped
what cursed thing arae you doing...
Very cursed things
yeah so like i cant wipe the lines at all LOL
ill PR this and go to sleep
Yeah so set state doesn't happen "immediately" in react
i added
react-xtermjs but i think we can get away with regular xtermJSit doesnt seem to be running at all tho
Hmmm
Yeah I might need to debug this later
also while you were away i implemented the users and disk password prompts
like i wired them up
so we should be good now
🔥 🔥 🔥 🔥 🔥 🔥 🔥
Really nice
check em out on ur machine qhenever you can
Really really really nice
we are close to getting this MVP done
Don't worry I will, now I just need to go to work 😭
@ledif found the bug
cc @imbev
oh
now i got the cool thing?
diffed a funcioning "try or install" ISO and a non-functioning one
there literally arent like any changes that i feel could make this not work
just maybe the gnome-user-setup vendor?
if anyone is better at analyzing diffs than me it would be awesome if someone could maybe point out something wrong
@tulip does it merge immediately if I approve or does it after passes all checks?
we dont have the merge queue yet
just dont... merge it
approving it should be fine
just making sure thats the actual issue
👍
Looking at the gnome-initial-setup changelog it looks like all of that is set in the vendor.conf file
huh
that might be what is going on then.
ill try to modify the existing in-distro vendor.conf instead of making one in /etc
Are you sure that X is uppercase
oh i fixed it on the latest commit
Another bash trick:
ARCH_SHORT=${ARCH_SHORT,,} to make it lowercase install of sedi did the same thing but with uppercase
hopefully itll work now
:P
Oh k nvm, sed fits there well then, didn't finish reading
ye
im so glad this thing actually like works on aarch64 too tho
hopefully now UTM will find this stuff automatically
@Niklas ⚡ clap clap, wheres my aarch64 tester?
Didn't expect that one
Though we would be stuck with x64
Not with this attitude
goated wit h the sauce
i want to eventually try getting UKIs working
dont see why it wouldnt work
would be goated having a UKI-bootable ISO LOL
maybe with that cool nmbl grub-emu menu
Mm wondering when fedora is going to push for UKIs
taking them long enough
we'll just move to wolfi before they do it
its just that just like this cloud native stuff, UKIs are "10 years away from happening"
cuz its sooooo hard
I mean. It's hard, because I don't think they have anyone interested in dealing with mako templates
its absolutely not
UKIs themselves arent that much trouble to generate
they'd just integrate them into bootupd and done
they already do it with god damn grub so why not any arbitrary PE artifact
secureboot is going to be extremely annoying due to legal reasons but man this is so close
also @Zeglius check this out
https://man.archlinux.org/man/ukify.1
You can do it
Ah yeah, tried it by myself. Was able to make one with the cmdline to boot
i remember making a UKI on chimera linux and it worked just fine
You probably were not using secureboot then.
Biggest problem with UKIs is signing them.
Other problem is that grub by default won't chainload EFI executables
Mm I'm more interested to see if they are going to switch to systemd-boot
That would solve the EFI executable chainloading, right?
Yes
Shim would load systemd-boot.
Systemd-boot would chainload the uki. It can still do the normal booting process as well.
There is another effort to make bootloader like environment with the kernel and then kexec into your install
Wow that's awesome
that's gonna slay!
its already looking really nice.
last step is getting the install wired up with usernames and stuff
its already "working"
slay slay
mmmm
the flatpaks/container seem to not be on the latest bluefin-live ISOs
no idea why tbh
@bketelsen hoping i can nerdsnipe you with this :)
.
It would be great to have the text and the button text customizable
that one is from gnome-initial-setup
the name is automatic
@tulip we should build this into Yai https://frimousse.liveblocks.io/
Frimousse — An emoji picker for React
Frimousse — An emoji picker for React
Frimousse is an open-source, lightweight, unstyled, and composable emoji picker for React—originally created for Liveblocks Comments. Styles can be applied with CSS, Tailwind CSS, CSS-in-JS, and more.
an emoji picker for your username
😎
ah yes
Gambling right there. We recently discovered the isos were failing because one of Just doc files was in Chinese 💀
welcome, :glorp:
wait no
Wonder what would happen in the username
I baked my first custom aurora ISO. But I see flathub isn't installed by default. Do I have to install it manually or is the switch to the fedora the new default?
fedora 42?
no 41
it should be there by default
no idea why it didnt show up for you...
I restore my snapshot to clean state and check it again
wait are you on a installed system?
and is firefox also removed by default?
oh you need to use the flatpaks list straight from aurora
yeah, I test it on a VM right now so it's easier to test 🙂
this is the default:
so I have to click on add flathub first to fix this
wait thats the live ISO right?
but on my production system flathub was the default
else you might eant to talk about this on #💫aurora
(no I build my own, wrong topic ?)
sorry
no worries LOL
i thought you were testing the live ISOs
ill still try to help you out
@j0rge
bootc install with progress bars coming soon
@Niklas ⚡
Nice!!!
@tulip always bringing the best Bling to the project 🔥
@Zeglius
Nice
I couldn't let you all have all the fun 🙂
Frontend complete for my personal set of images. The one user will be so happy with this 😆
Now to wire up the important bits that actually change the system
Now to wire up the important bits that actually change the system
I've decided:
1. I don't know enough Rust
2. I hate Typescript
3. I dislike frontend
4. Copilot is awesome
But I'm pretty happy with how it's looking
That looks so good!
It's entirely RedHat's Patternfly design library + a tiny bit of tailwind. I can't take the credit for it...
WHO needs anaconda-web when you have this
Skill Issue? :clueless:
just kidding tho i get you
It definitely is. Things just don't do what you tell them to do
to be completely honest this seems like the best way to do it
i like yai but im reimplementing a lot of stuff
its not like it matters tho, we can always switch up installers at any point
haha this looks awesome!
we don't have to reimplement much if we keep it small and lean and mean
next next done
yup!!!
its just that some stuff is kinda complicated
specifically the terminal for bootc install is being the worst part
but its gonna be nice
don't force it, it will come
and f42 beckons
this is just so fancy tho
im such a fan of this
that does look amazing indeed
its a fully functional terminal
im just restricting it to not be writeable by the user
we can do like
a post install button or something
advanced users only
then it launches a shell on the installed system or something LOL
yeah!
@Robert (p5) are you using tauri?
for ur installer?
Yup
if so i can share you my interactive terminal code
itll make ur job easier
ill PR it to yai soon
Oh, I want planning on having one.
I'm just going to use progress bars 🙂
im thinking about making like dynamic progress bars
like
depending on the output that bootc does we change the percentage
while that json stuff isnt landed yet
Didn't tauri have a plugin already to stream the commands output?
yup but its not using ptys
so we dont have bootc progress bars
O
check this out
oh shit i cant cancel it
@Robert (p5) @Zeglius
Wouldn't that be running in host, right?
yeah i just added
distrobox-host-exec :Palthough the bootc progress bar kinda gets stuck :(
Mmm, ptys are tricky I guess
so close tho
it seems to be installing on the background anyways
we'll hide this anyways so its fine
like behind a lil button or something
material you design my beloved
implementing flatpak installation on yai :)
ok so
we'll need to manually enable
tpm2-luks block devices on /usr/lib/bootc/install/*.confBlue95 live ISOs are now a thing
https://github.com/winblues/blue95
Also made a PR to fix the timezone issue with Aurora/Bazzite
https://github.com/ublue-os/titanoboa/pull/60
I think we should just default to UTC timezone, it's what everyone else is doing with live ISOs.
Can't we use the geoclue thing?
How does that thing exactly work? Like, it detects the time zone automatically or something?
Yeah but it reaches out to the internet to do it, so probably not possible to use it in early boot before networking is set up. I changed the PR to set it to UTC as suggested
you are already using it
default timezone stuff on GNOME and KDE uses it
its just a lil location service
O
Ah ofc doesn't work because no internet connection
I thought the location stuff was with beacondb
beacondb is the backend for geoclue
yeah if it's connected then it'll set the timezone
if will set it when the user connects to the internet anyway
Why is tanstack router in the corner of all of these screenshots?
Thats just in devmode
In Production Mode those things are not present
debugging for the rouyer
its pretty nice
also
and the other things is for tanstack query
why those instead of react-router and react-query?
no idea what makes tanstack noce
nice
well
react-query = tanstack query
and tanstack router because... idk
oh LMFAO nice
i kinda wanted to try it out
cuz you did t eant to use solid....
yeah i know....
i what
you wahths
please dont get a stroke
rempte stroke activated
Tried svelte. It was rather pleasant, the routing sticks to vanilla JavaScript.
Also no jsx
I much rather use everything that uses React or supports React components. Don't want to reinvent the wheel, so sticking with existing design libraries is far easier.
Astro is my favourite at the moment because it's super simple to do pretty much everything, and supports all the existing libraries for React components.
I mean react reinvents the wheel each few years so
Don't get me wrong... I hate working with React. But Astro means I can use React components without directly using React 🙂
guys just use assembly
why reinvents the wheel
as simple as that
what are you making with react
GitHub
GitHub - ublue-os/yai: WIP: Installer for Bootc Images
WIP: Installer for Bootc Images. Contribute to ublue-os/yai development by creating an account on GitHub.
.
if rust, then maybe slint?
otherwise qml is also kinda efficient
unless you guys really like react or something
personally can't get my head around it
@tulip🌷 @Kyle Gospo @j0rge I got a message from Jiri (one of the main anaconda developers). He wants to target Ublue's needs for the Anaconda Web UI. What are our thoughts around that?
i think flatpak installation, secureboot on the live ISO and manual partitioning (bazzite), right?
we are using bootc install rn for the installation on the new live ISO stuff
Right, I guess what I'm wondering is before asking Jiri and the team to spend cycles on something for us, I want to make sure we aren't just dropping anaconda since we have been working on the live ISOs.
I'm happy to provide requirements and work closely with them on it.
All of those are still things that silverblue will want to have.
True. Valid point.
Are only unique item from what fedora is probably wanting to do is specifying flatpak remote (but this is ISO build time) and enrolling a MOK.
The second one is a kickstart post
For sure. Also the Web UI doesn't work outside of a Live ISO environment currently.
This is no longer an issue for us
We can use the work we've put into live ISOs and run Anaconda instead
are we sure we are going to fix the actual issues we've been running into for the past years if we stick to anaconda tho?
most of the issues were purely caused because anaconda was so bad
What issues?
The main one I can think of is the failed installs, which might already be resolved by renaming that Chinese file name
1 billion cryptic errors on the anaconda installation, awfully big ISOs, usability issues that we cant fix (...)
actually like most stuff that is on #🛟bazzite-help
and #🎮bazzite
Cryptic errors should be an easy fix. @Noel - This is quite a big issue. Anaconda hides all errors from us, making it impossible to debug issues.
Big ISOs, I think we've resolved this, right? We can install anaconda into our live ISOs. But we do need to ensure we can use
bootc install inside anaconda rather than ostreeCryptic errors should be an easy fixcan we fix it? technically we can fix this on anaconda if we PRd it but good luck doing that
The conversation started because Anaconda wants to help support our use-case. So now is the time to request these things
true
So @Noel , along with Flatpaks and MOK enrolment, we want Anaconda to support
bootc install commands using local containers-storage and to have improved error logging.i believe so yeah
I would also reiterate Bazzite's partitioning needs for dual-booting with Windows, but I think anaconda is decent for that already
What's the primary cause for the partitioning issues? Anaconda can create partitions quite well. Is it more of an ostree problem in that support for custom partitioning is flakey? Or do paritions need to be setup in a specific way by the user, and it's easy to get it wrong?
I just mean that partitioning / dual-boot support was brought up when developing yai and I believe the Bazzite folks wanted to make sure that any new installer still has flexible partitioning capabilities. But if we're going to use a bootc-aware Anaconda, then that's less of an issue
I'm not going to reiterate the last 3 years worth of issues back to the anaconda team
true tbh
i mean, if reporting those right now makes anaconda at least somewhat better itll be a win-win right
like, if you need ublue to tell you "make your installer work with bootc" then I don't really have much to say considering this was all targetted years ago. All of those issues are years old.
I'm not going to stop anyone trying to work with anaconda but for bluefin I'm done, we just need to blast a disk.
we'll be able to use zstd as the compression for erofs on the bluefin-hwe:lts/aurora-hwe:lts images soon
LETS GO
can someone tldr the status of yai?
- installation works, its a bit jank but its working nicely
- not super sure if flatpak support works yet but it should™️
- there are a few mocked pages for user creation and FDE
- users page isnt functional yet (jsut needs a tiny bit of wiring)
- FDE (full disk encryption) cant be done until we can get manual parititoning going (upstream bootc issue with the tpm2-luks install option)
bummer on fde
for bluefin in specific we just need:
- installation page to look nice and not be jank
- hideable users page + hide the FDE option cuz we cant have that yet
yup. its really sad cuz it was so close
got the upstream issue handy?
lemme get it
https://github.com/bootc-dev/bootc/issues/565
https://github.com/bootc-dev/bootc/issues/477
anything that lets us create a password for the LUKS volumes is a fix for this
im thinking we should just straight up have really simple systemd-repart configs directly on the ISO right now if we really want FDE
@Robert I reached back out to Jiri. @antheas brought up a good idea which would be to have a meeting with Jiri about needs for Anaconda and better support for bootc based images. Would either of you want to be on that meeting? I don't have a time yet.
My proposal for this is you do the partioning, unlock, then deploy with bootc
Having bootc do the partioning will be a can of worms
Bootc-image-builder's anaconda does bootc to-filesystem and Anaconda does the partitioning. We just need this same functionality on WebUI
Is the WebUI purely a web UI? in which case, the backend functionality to use bootc should already be there?
How do they tell Anaconda to use bootc? Is it just a file or config we need to set ourselves?
Wdym?
Bootc has no config files for install to-fikesystem as far I'm aware
If BIB uses
bootc install in Anaconda, how do they configure Anaconda?pretty sure it uses rpm-ostree deploy
oh yeah it does i forgot. wait I'll have to double-check
looks like it's got a first draft here: https://github.com/rhinstaller/anaconda/pull/6298
GitHub
Add bootc kickstart command support by elkoniu · Pull Request #629...
This is a placeholder to sync bootc kickstart command support before formal PR being created.
Right, do we have some sort of reference to how anaconda selects an image source?
I don't understand why we're acting like anaconda is our future
we already decided
So you're saying there's no future for Anaconda in UBlue?
not for me
but like, they made this decision for us
2 years ago
look at the issues
we've been begging and whining, why do they care now?
we need to blast a disk, support secure boot, FDE, and flatpaks
I don't care about lorax
or any of that stuff
and no one else does except for people who think that doing installations this way is a good idea.
and the only people think that are enterprise companies, we make consumer images, blast the disk asap so I can get to work
plus we're on the cusp of bootc supporting other stuff, why would we stay tied to a distro specific installer anyway?
what happens if we wanna prototype flatcarfin or wolfifin?
Archfin
plus this 2 week old installer is already showing more promise
...but the bazzite dual-booters
"tulip and zeglius did more in two weeks than Red hat has done in 3 years" means we don't need anaconda.
bazzite can do what it wants, doesn't mean the rest of us have to
or want to
personally I think tying ourselves to an unhealthy project is a bad idea.
We can probably support dual boot with a couple weeks of testing
and I'm willing to say that publically.
I kind of already did in my talk
If we can support dual boot here, I will move to this
Anaconda sucks
hopefully you mean anaconda lol
hey if you cook maybe we will use it
of course I mean anaconda lol
I find myself having to support not using anaconda and all I got is "look at the 20+ years of history"
just focus on those UX gains because i have like 50 dual boot videos and the hobbyists are making "how to install broken steamos beta on your gizmo" vids now
nod
update from timothee. He thinks we're going down the wrong path. Feel free to tag him in an issue and he will tell us what to do.
I guess we're just cursed with this thing
anyone got an ubuntu usb stick lol?
Created a general one covering both the live ISOs and installer
https://github.com/ublue-os/main/issues/778
GitHub
Installation Improvements · Issue #778 · ublue-os/main
The installation experience across the UBlue projects is sub-optimal, with frequent errors, inflexibility and causes confusion. We've started investigating into live ISOs (ublue-os/titanoboa) a...
I can't see how bootc install and systemd-reparted would be dead-ends. Curious to know what their thoughts are
James dm me pls
I mean, we are deviating from upstream by doing this, can see the concern
what are we even doin tho
mmmmmm
The iso
we can just use kiwi if thats the issue
Though the major offender would be yai probably
not using anaconda no more
and i dont really see whats wrong with habing a custom installer
Unexpected behavior and differences compared to upstream
bootc install is the most upstream you can be tho 😭
systemd-repart tooNo is not
Anaconda still uses ostree to install as far I'm aware
i know but like
eventually they should and will move off of it
They have started looking at using bootc directly already
Very early stages, but they are looking at it
Good to know, though I'm concerned regarding how we are supposed to configure it
We can throw an interactive-defaults.ks into something like /usr/share/anaconda and it'll pick it up
That's how the current ISO works with ostree
So it'll just be a different kickstart contents
Is there a template?
Template? Not that I know of.
Documentation - https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/9/html/automatically_installing_rhel/kickstart-commands-and-options-reference_rhel-installer#halt_kickstart-commands-for-installation-program-configuration-and-flow-control There's a section for ostreecontainer, which is what we currently use, but will be replaced with something bootc-specific
Documentation - https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/9/html/automatically_installing_rhel/kickstart-commands-and-options-reference_rhel-installer#halt_kickstart-commands-for-installation-program-configuration-and-flow-control There's a section for ostreecontainer, which is what we currently use, but will be replaced with something bootc-specific
for what I read, this is more something for static configurations.
Well yes, it supports a single image. It's doesn't provide multiple options for images or anything like that
that assumes fedora is our only upstream. 😄
Before anyone says anything, I know Anaconda isn't the best.
IMO if RedHat is now working on improving support, we should at least hear them out. When bootc gains popularity in other distros, there's likely to be other third party installers available
IMO if RedHat is now working on improving support, we should at least hear them out. When bootc gains popularity in other distros, there's likely to be other third party installers available
I'm tired
I'll just publish bluefin images and skip the ISOs, maybe a rebase thing might be useful
Tbf bootc is quite new
I can understand why they did stick to ostree install method until now
do we even know what they are saying we are doing wrong tho?
the ISO creation maybe is wrong, we can use kiwi if thats "wrong", yai isnt wrong by any means, i dont see like what makes it bad or anything
I don't think they know either lol
I mean yai isn't complete plus there is the maintenance load of a possibly niche installer
we can just move off of it when a decent installer shows up
I mean
Which one?
whatever people make
we need a installer so....
I don't think there is an active development of installers for bootc
we still don't have flatpaks either
in anaconda
Well, Readymade
yeah
good there you go
if thsts good then we can use it eventually
Is still in development as far I'm aware
There is Flatpaks in anaconda. I think we just need to put them on the right place in the image -
/flatpak/repoPlus, we don't want to stack expectations on top of fyra labs just by selfishness
can we do this with the existing bib installer?
tbh i dont think so
Not with BIB, but with titanoboa, yes (I think)
not really, it's why Build Container Installer exists.
there is actually a hack to anaconda to get it to use flathub as the default repo. Otherwise, it's hard coded.
Well yeah but how are we supposed to configure that thing is the question
"there is actually a hack-"
The one I imagine?
We are able to put files in the rootfs in the ISO.
So we can do that with the right config files
yhats the past 3 years
"theres actually a hack-
I believe it installs everything from /flatpak/repo onto the system
if theres something wrong thats whats wrong tbh
Right, but we don't know which config files, or how anaconda is setup in that action, do we?
I did hear back from Jiri if we are still interested in hearing out the anaconda folks. I understand Kyle and Jorge are wanting to stay out of it, but having communication with upstream can only help us, not hurt us.
I extracted the ISO. It's only 4 or 5 kickstart files needed in a specific directory
make them use github
why are we backchanneling?
I want to establish contact first. I will point them to a github issue ultimately.
why?
please don't promise anything
why not?
because I'm trying to get rid of this thing
not double down
I'm not, I haven't even had a discussion.
Got a list of the paths? Gonna see how complex these are
/usr/share/anaconda/interactive-defaults.ks -> ostreecontainer --url=/run/install/repo/lumina-cosmic-stable --transport=oci --no-signature-verification
/usr/share/anaconda/post-scripts/flatpak_configure.ks -> Adds the Flathub remote and disables flatpak-add-fedora-repos.service
/usr/share/anaconda/post-scripts/install_configure_upgrades.ks -> Runs the bootc switch --mutate-in-place
/usr/share/anaconda/post-scripts/secureboot_enroll_key.ks -> Does some checks then runs mokutil --import
And here's the code which I believe reads /flatpak/repo and installs everything from there
https://github.com/rhinstaller/anaconda/blob/84cd12d77fca4472b2d43797b7cb6b4ba26fc980/pyanaconda/modules/payloads/payload/rpm_ostree/flatpak_manager.py#L51
disables flatpak-add-fedora-repos.serviceWe probably need this in ublue-os/main and ublue-os/bluefin-lts?
Yup
Wait maybe not in main, might be too opinionated
on a different topic I was starting trying out putting ublue bootc images in a folder on a Ventoy drive and then boot into a live image and deploy the images from the drive
oooo do those work?
on ventoy
also works for me
noice
also @Robert i think bib doesnt store the image contents the same way the old ISOs do
afaik ventoy just does
losetup -Pf (iso) then boots off of the CD_LABEL on the ISO
maybe for whatever reason the files cant be read or soemthingFor bootc install. I'm assuming we can install from an oci-archive. I'm guessing that archive might be smaller than container-storage
We could use Calamares with a custom bootc module
It's probably the most popular installer
noooooooooooooooooooooooope
No need for custom module, there is one that executes shell scripts
Wdym no
Might end up being the most distro agnostic installer out there
I'll see what it takes to setup Calamares with Titanoboa
Yai is very impressive, but Calamares is fully functional on many distros today
Also fits in nicely with kde
Yeah, and already handles partitioning
lets just use it for now i guess LOL
Let's see if we can actually use it for bootc
I mean I already mentioned it a couple of times
https://discord.com/channels/1072614816579063828/1318955957509685298/1333549910540157012
the idea is to not use calamares at some point
I'm open to any option, as long as we remove the maintenance load from ourselves
But the expectation to an installer to be as easy to configure/agnostic as calamares, and also being agnostic, simply spawn out there
Is close to a desilusion
Nobody is making installers to be agnostic
ill just get done with my exam week already at college and ill finish up yai, you'll see what im talking about then
I'm not in a hurry, the hardest part was already dealt with (iso)
But I'm going to look into more options, fixation in our house made tooling is a pitfall
We might need to wait for the KDE epel sig, but it's already packaged for Fedora
There doesn't seem to be a package for kf5coreaddons
Only kf6
I talked to Neal about a possible package in EPEL, however there are some technical obstacles.
I am very thankful we have so many folks in getting bootc ISOs fixed. It was just Jason and I trying to wrangle anaconda for the longest time.
What if we launch Anaconda from Titanoboa?
Probably would work, but you still have the baggage of using anaconda.
Not sure if the web UI works with bootc based images.
Or if it has better error logging when something goes wrong.
Why wouldn't?
Don't mean to detract from the convo, but I would appreciate a review for https://github.com/ublue-os/titanoboa/pull/61. I want to change the live user's name to "Chandler Bing" for Blue95 ISOs 🙃
Last I looked, not everything was wired up for the web UI. That may have changed.
I mean as long is easy to configure, and can be launched by executing a binary or script, should be ok
So much Sigma in that sentence
@Noel if Anaconda folks ask, tell them that we need to be able to use /var/lib/containers/storage as a possible source of container images.
I don't want to imagine somewhat hardcoding the container image in an OCI directory like what is being done in our isos rn
What's the problem with hardcoding
Well, not hardcoding itself, just the part where you place the image in any arbitrary way in the root filesystem
tbh i think the way we are doing it now (image in /var/lib/containers) works pretty good
we can use these tho if theres a reason to do it
O okay didn't remember that one
yeeee
thats from bootc switch
install supports the same stuff
cuz it uses skopeo under the hood
I jsut just hope we don't need to deal with lorax again
never ever again
we'll maybe use kiwi or mkosi at some point for titanoboa
but like
fuck using lorax i am tired of that bs
Mkosi doesn't create bootable isos
Probably will take a look at kiwi
kiwi is a bit over-engenieered for our stuff
but if we can just transform a
.img to a bootable ISO we are GOLDEN with this
mkosi does generate initramfses and i think it would be a nicer way to work with the rootfs
so... maybe?
no idea!I mean, for the initramfs is just a dracut call
Could .img be used to support RPI and similar platforms?
we have control of literally every single step of the ISO process
we can do whatever we want
we are just packaging it as an ISO right now but we could make .img instead for aarch64 embedded devices
oh my god
bootc install does not do swap by default
i dont have swap on my system right now 😭
? Isn't the zram setup by default?
no zram on the default RHEL kernel
A
theres zram on the hyperscale kernel tho
What? This seems odd.... Zram is like the best idea for swap
im pretty sure it didnt work by default
i think the hyperscale kernel has a config change specifically for that
it works perfect once you enable it tho
@M2
Interesting I wonder what is the rationale
Well, we are using bootc base images, which are meant to be the minimum required set of packages and configuration for a working system. Zram mustn't be important for all use-cases, so downstream would be expected to configure it themselves
but like this is the default on centos stream
like upstream centos stream
we are configuring it ourselves but wha.....
like theres no zram on the default kernel?
ill have to check out the sources
@Robert @M2 it does have it! its just that neal fixed it a while back :)
interesting
i wonder why they dont use it by default then
Is this file required for zram?
that
CONFIG_ZRAM=m is required, yeah
oh, the other oneSorry, I mean the generator file in the screenshot
yeah you need that for automatic zram generation
you can set it up manually and a few distros do it already
but the systemd zram generator config is really nice
Systemd zram generator is my preferred method
It can do the math in the config file
For sizign
https://gitlab.com/redhat/centos-stream/rpms/kernel/-/blob/c10s/kernel.spec?ref_type=heads i dont find the zram commit that neal did on the c10s kernel tho
That would be unfortunate
seems to be enabled regardless
its super nice
I think popOS does 50% of ram as zram by default
@tulip🌷 @madomado h
a
dnf install --nogpgcheck --repofrompath 'terra,https://repos.fyralabs.com/terrael$releasever' terra-release
command for installing terra for el10ill have to set up a VM anyways
and also uh
im on centos-bootc
wtf is c10
oh
centos 10
centos stream 10
centos stream 10
yeah uhh
centos stream 10
centos stream 10
centos stream 10
yeah uhh
wait why do you need rdm on el10
4k
i thought ublue is fedora
titanoboa uses the same container that will be installed as the ISO rootfs
ohh yea
btw if you ask me, I have absolutely no idea how to statically link libhelium
you don't
bluefin LTS uses c10s
it's GTK
you don't statically link GTK
well it's not gtk
it's kinda more like an alternative to libadwaita
it is GTK
libadwaita is gtk
btw taidan has been running forever and I've no idea why
its gtk all the way down
mado this is a ublue chat
so yeah um about the readymade stuff on the live ISO: titanoboa has just complete control over what can be on the rootfs so i think we can just plop libhelium there and done
yeah
so do that
it doesn't need to be statically linked so its fine
Terra EL10 is still going though
we're having a hard time porting packages to EL10 😭
you guys maintain way too much stuff
100x developers in a company what do they do
then 0x developer cuz you died cuz of stress
😭
we have 5 100x devs who have nothing else to do but maintain code to maintain their own sanity
where 5
aaaaaaanyway
I think I can't count /hj
idk i made it up
real numbers is like 3
ah
@tulip🌷 so all you need is a bootc backend in rdm?
should be doable then
pretty much yeah
we also need to copy flatpaks from the ISO
we are literally running
rsync on yai for thatoh, we kind of have a solution for that actually
meet polycrystal
GitHub
GitHub - Ultramarine-Linux/polycrystal: Barebones "automatic" Flatp...
Barebones "automatic" Flatpak installer for distribution-default Flatpak packages. - Ultramarine-Linux/polycrystal
needs a remount after bootc install afaik
way ahead of you on this one
yeah, we already have a solution for this
the container images requires the polycrystal oneshot daemon to be installed though
we dont need anything on-boot tho
it's a oneshot thing
you only run it once
it's meant to only be run once to copy the flatpaks over and that's it
like the scope is just:
- partition the device
- bootc install to-filesystem
- plop the flatpaks to the right place
- done thats it for the entire installer
at least for our usecase
@ublue
yeah polycrystal does the flatpak part and readymade does the rest
that runs on the installed system, right?
polycrystal i mean
iirc yes
so we cant use it 😭
aw man
lemme read the code
so... you need copy /var/lib/flatpak from the host to the target install?
yup!
alright
that is doable
very
its just that we cant modify anything except /var and /etc on the live ISO
after bootc install
yeah this is doable in readymade
What's cs10?
centos stream 10
Centos Stream 10? 😉
centos stream 10
Same container image + livesys package + whatever is ran with the postrootfs hook. So will differ from the container image to be installed
And I think we made it so can use a completely different container rootfs for the liveCD
What is libhelium?
Our (Fyra labs/ultramarine) application design framework https://github.com/tau-OS/libhelium
GitHub
GitHub - tau-OS/libhelium: The Application Framework for tauOS apps
The Application Framework for tauOS apps. Contribute to tau-OS/libhelium development by creating an account on GitHub.
fyra's alternative to adwaita
used in readymade
That's very nice, it reminds me of Material UI
counter strike 10?????
holy shid
True.
vac 12: actual working anticheat :clueless:
google does 2x on chromeos
https://issuetracker.google.com/issues/227605780?pli=1
afaik there's no drawback to this
since zram size measures the uncompressed size
I do it on my own image as well haven't found any issues but its not like I ever need to use zram
I did stress test it though
Not implemented it yet, but enroling MOK keys will be easy - just copy & paste the existing kickstart
Also I don't think the ISO will be close to the size we made with YAI. At least at the moment. Raised a feature request to ostree and anaconda which would help with that.
got links? sorry on a plane
Currently local. But will push it somewhere shortly.
Pretty much titanoboa but installing anaconda and webui, adding some kickstart files and using oci-dir to store the image rather than containers-storage.
oh ok so we'd at least have the liveCDs?
tbf I am fine getting yelled at for not using kiwi than moving to it since what we have works.
I guess we can go from "Eats Anacondas" to "Takes Anaconda out on a date and sets the table".
"Use kiwi"
I mean... If it is that easy, why doesn't upstream do that already?
shrug whatever, I'll deal with it so ya'll don't have to
Btw probably asked this before, but do we have a link to the issue for upstream atomic livecds?
Wanna dig what did they come with so far
can't find it, timothee (siosm) made a kinoite prototype
that might help with your keywords
GitLab
Create Live ISO installer and pre-installed disk images from Bootab...
The current installer for Fedora Atomic Desktops directly boots into the Anaconda installer and does not offer a Live ISO environment for users to explore the system before...
no wonder
Basically sums to this https://github.com/OSInside/kiwi/issues/38
GitHub
RPM-OSTree - git for operating system binaries · Issue #38 · OSIn...
With rpm-ostree, you can create versioned filesystem trees to provide atomic, reliable upgrades. "git-like" management of appliances would be pretty sweet.
more specifically this
For what I understand,
bootc install-to-filesystem in order to prepare the rootfs of the squash
Did we ever find out where the space reduction came from with the new vs old ISOs?
Maybe squashfs/erofs detecting matching files from the rootfs and the containers-storage(?)
honestly no clue, probably tulip knows more about that
Flatpaks are close!
It is just putting the files into /flatpak rather than the /var/lib/flatpak we were using
would a bind-mount for /var/lib/flatpak -> /flatpak work?
or a symlink
Not sure about a bind mount - /flatpak is inside the rootfs, so probably not?
But symlink, I assume so
Symlinking worked.
I just need to figure out how to get rid of the above error.
I just need to figure out how to get rid of the above error.
Well, the install worked with Flatpaks. But it booted into an emergency mode screen as initrd-switch-root failed.
is the root mounted?
what does the initrd see?
:/
Might know if things worked next week
But I think it's very close!
Things like this make me wonder if locally building would be faster..... That's pain
The issue is my internet. Building locally, I get timeouts downloading the flatpaks... At least this way I know the build is successful
next time, try with axel. Its a dumbed down aria2c with multiple simultaneous download channels
Does that work with GitHub job artifacts? IIUC they don't give you the download links - they do some redirect/JavaScript magic or something
I guess I could push test ISOs to Cloudflare too
yeah do that, i dont think will work with artifacts js
well, next time
dont go now and cancel the download
14 hours left on the download...
New internet is getting installed at the end of the month. Can't wait
New internet is getting installed at the end of the month. Can't wait
😲
at least FDM has a browser extension that intercepts all downloads initiated by webpages and takes the raw link, maybe other download managers have that too
you can make your own coreOS iso
how actually
I meant as in putting the image right inside the iso
this still requires actually downloading an upstream coreos image, and then embedding a script to automatically rebase to your custom image on firstboot
my question is how to build a completely custom ISO that already contains the custom image offline
ah i see
no idea on that front
I'm looking at how COSA does it
actually I wonder if we can do some freaky stuff and generate a custom ISO and just get ignition to work
so it does bootc install, and then run ignition for firstboot setup
just skipping coreos-installer and COSA altogether
and only use ignition for our stuff
bootc install to-filesystem --source-imgref=whatever --target-imgref=my-custom-other-imageyeah like that
ucore but without the core
what do you call it now
u
The primary goal of Fedora CoreOS is to be useful directly, and to have users deploy containers for applications, etc. Some people and organizations have asked about creating custom derivatives of Fedora CoreOS using this project. This is supported on a best-effort basis. Supporting these use cases is not a primary goal of coreos-assembler, though we try not to purposely break them. Also note that coreos-assembler is still evolving, and it’s likely that there will occasionally be breaking changes. You should subscribe to releases to make sure you’re aware of these. That said, we do try to maintain a stable interface between coreos-assembler and the “source config” repo. Maintaining a custom operating system build is a nontrivial undertaking that you shouldn’t take lightly. But we will make efforts to document changes and keep coreos-assembler relatively stable. coreos-assembler is used to build RHEL CoreOS, but coreos-assembler is not a Red Hat product at the current time.i keep entering the wrong chat wtf what if you have a bootc image that literally just runs ignition so you get all the benefits of coreos without the coreos part I should look into how coreos-installer works
what... bare fedora?
fedora bootc to be specific
ah
so basically CoreOS without coreos-installer
built with pure bootc instead of using the COSA toolchain and rpm-ostree
we made that already
actually we archived it just today
which one?
OH ccos?
yeah
that's just ucore + the bib action
yee
bruh
?
wdym you just archived it
because no one was using it
oh
I can transfer it to you if you want
yeah transfer it to ultramarine
I'll maintain it
one sec
what org?
Ultramarine-Linux
you need to give me temp access in that org to transfer
ok
hold on
done
incoming, then remove me. All done!
ULTRAMARINE SERVER
it was planned but I never got to it
since uhh everyone else was locked in on desktop
hold on I had an old project lmfao
GitHub
GitHub - Ultramarine-Linux/kanopy: A K8s-optimized operating system...
A K8s-optimized operating system, based on CoreOS. Contribute to Ultramarine-Linux/kanopy development by creating an account on GitHub.
need to redo this again
kanopy was meant to be coreos fork
Has anyone had success creating a user after running
bootc-install? I am basically calling pkexec chroot <ostree-root> useradd. It adds the user correctly but the home directory it creates is in the /var/home/ of the ostree deployed root which doesn't seem right. When I try to log in as the user, it accepts the password but has no home directory.
Here is the line in my installer: https://github.com/winblues/blue95/blob/c795ff93ce722aef49cabc6a096ca6b53908df54/files/system/usr/libexec/blue95-installer.py#L119
I'm thinking I need to use systemd-firstboot or something like thatoh @Robert did you manage to get anaconda-webui working on the live stuff?
wanted to test it out too!
i wonder if it works with c10s...
Yes. It works well
nice!
mind PRing ur stuff to titanoboa?
https://github.com/rsturla/eternal-images/pull/374
Here's what I have. It installs, but Flatpaks still need work
oh damn
i wonder why then? like
it should be just a copy+paste operation
i think the installation just needs to be
/flatpak instead of /flatpak/repoI didn't copy and paste, but I did try to copy line-by-line so I understood how it works.
It might work... But locally I get a lot of timeouts, so one of the runtime ones could have failed
oh yup
but like didnt you say the flatpaks made the VM not boot?
The desktop icons are showing up in the live ISO. But won't open
Everything boots
maybe a bindmount from
/flatpak to /var/lib/flatpak would work?
why does it need to be under /flatpak tho? is that hardcoded on anaconda-webui?/flatpak is in the rootfs. I think symlinks are better?
Yes, it's hardcoded in anaconda
damn
huh maybe just having
/flatpak be a symlink would workWe should only need the /flatpak/repo directory. That's all the existing ISOs have
i think you still need the other stuff
maybe thats why the flatpaks didnt start up?
so like why did you copy titanoboa to ur repo? did it blow up somehow?
Didn't blow up. I just don't fully understand Justfiles. And didn't want to blindly copy and paste things without understanding it
fair LOL
There's a good chance Flatpaks didn't work because the script failed to download the runtime packages (internet issues). Need to download the build from GHA to test that
omg true
that might be it, yeah
if you still have the ISO you can try running the flatpaks on ur terminal
it should fail at least in a more understandable way
Also my local machine is an Nvidia one, so even in the container it tries (and fails) to download the Nvidia runtime deps
And I'm testing the install in a non-nvidia VM
maybe you'll need to manually specify the nvidia stuff?
man this is also kind of a problem, like, how are we gonna get anaconda-webui on c10s?
maybe a copr?
Isn't it being built yet?
nothing on the official repos
or epel
Hmm. Yeah, maybe a COPR. I doubt they are planning on using the webui on EPEL10, so there's probably not much reason for them to build it
i wonder if we can wing it anyways
Got to walk the dog. Will be back in about 90 mins
everything is on the official repos, just anaconda-webui that isnt there
aight have fun!
Yeah. Simply installing that package made the webui work. No other changes
The main thing is the interactive-defaults.ks file needs to have the ostreecontainer command, otherwise it tries a regular workstation install
isnt there an explicit warning there? 😭
Yes. But it's how Silverblue does it
Like the official Fedora-built Silverblue ISOs
like is that the
bootc install equivalent of what we are doing?uses ostree container deploy behind the scenes
Now I'm actually off. Be back soon
aight!
omg if we could do liveCD + anaconda + flatpaks soon
then we'd be done
doing that!
or at least swap out the ISOs
im writing down an issue with everything that @Robert did
excellent
then its just implementing all this stuff and then we'll be done
hopefully
:)
part 2 of "we'll be done"
https://copr.fedorainfracloud.org/coprs/ublue-os/staging/build/8868108/
anaconda-webui builds successfully on epel10
so it should be fine?
HURRY
LTS GOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
lol
lts go
also omg isnt ur thing in like an hour
yep
nice
I'll be ready
GitHub
Primarily support
anaconda-webui · Issue #64 · ublue-os/titanoboaWe discussed about this a lot on discord recently, and the conclusion was that we should go with anaconda-webui as our installer instead of yai because we'll have support from more people and i...
could this also install a fedora image?
we're just sourcing from centos right?
it can install anything i think
oh so this will work across the board for us
not just lts
the ISO and targets can be completely different
yup.
the titanoboa checks test bazzite, bluefin, aurora, and aurora LTS
Yeah. I'm using it on my Fedora image, and it mostly works.
The install works. Flatpaks are a WIP
The install works. Flatpaks are a WIP
you try the new ui?
figured flatpaks are busted
Flatpaks being busted is just because we don't know exactly how it should be used. It will work
Yeah, webUI works well too
!!!!!!
So what you're saying is
Neal Gompa saved the day
Umm. I'm a bit lost.
This was working a few days ago 🙂 (On Fedora)
This was working a few days ago 🙂 (On Fedora)
It's just the live-anaconda part we needed right?
For CentOS, yes. Fedora was already good to go
NEAL NEVERMIND GO BACK TO BTRFS.
robert you almost make me believe that we'll have live CDs with the new webui before F42.
If the live iso is F42 it can deploy F41
robert was smart enough not to reply
I'm trying to trap him
Well... Everything works independently. WebUI works. Flatpaks work. Live ISOs work. We need to put them all together
If I could actually test the ISOs, I might be more optimistic.
ok so let's solve that
do we need to just a thing
or make a new bucket in R2 for everyone?
There's nothing except waiting. It's an internet issue that will be fixed in 3 weeks.
I can't build the ISO with flatpaks locally because flatpak times out (not configurable). I can't pull from CloudFlare because it takes over 8 hours.
I can't build the ISO with flatpaks locally because flatpak times out (not configurable). I can't pull from CloudFlare because it takes over 8 hours.
its working
How much of it?
i just need to know how to run anaconda LOL
Does the gnome setup thing not show?
nope, thats on purpose tho
livesys scripts does that
Did you install the webui ?
we can just remove that if you want it to show up
without it jsut to see if the regular anaconda works
Oh I think it doesn't work on the live env
You can maybe drop into the TUI
how does fedora workstation do it tho
oh shit
@Robert @James
@j0rge
its in portuguese cuz of my timezone LOL
we can embed those cool themes we made for bluefin here too
Awesome! LMK what happens with the flatpaks after install
no webui?
Don't we want to webui now?
thwip. :p
If anaconda works, the webui will work
just making sure regular anaconda works right now
yee
"regular anaconda"
it does need some configuration i think tho
heh its doing the wrong thing
@Robert
Did you add the ostreecontainer kickstart?
pretty sure
lemme check the rootfs
do you remember what was the correct place?
/usr/share/anaconda...?
/usr/share/anaconda/interactive-defaults.ks
:clueless:
Did you install anaconda after writing that file?
i dont think so
It might have been overwritten
omg i bugged out the gh actions thats why it didnt work
Try the artifact here
https://github.com/ublue-os/bluefin-lts/actions/runs/14316973733?pr=427
its just because i didnt put the right argument on the github actions
jsutfile expects
1, actions has true
@Robert its absolutely working now
YESSSSSSSSSSS
Did you rebuild with my new commit?
I added a step that fixes Flatpak installs
I added a step that fixes Flatpak installs
we dont need
/flatpak
jsut specify /var/lib/flatpak as the flatpak installation that anaconda will use
at least seems to work?
ill check it out LOLI think it will fail soon. Missing two things
lets see!
rebooting the VM
it showed up as successful
oh wow it deployed!
without the flatpaks for now
but still
No flatpaks?
Here's where it's hard-coded
https://github.com/rhinstaller/anaconda/blob/6e74fcef5512c805a39d005aa8c0b6e777c56463/pyanaconda/modules/payloads/payload/rpm_ostree/flatpak_manager.py#L51
And you probably missed this
https://github.com/ublue-os/titanoboa/pull/65/commits/dff21a2eb9fc1a7d80c16fda84fea62345c39ea3
holy shit
Also need a post script to bootc switch to the registry image
might as well flip the web bit! Please note that I have never seen the UI live so I wanna see it lol.
yupyup
ill do that right now with the titanoboa stuff
Probably makes more sense in bluefin-lts right now? titanoboa isn't anaconda-specific yet
i just added an anaconda option
Web UI is a single package last time I was messing with this.
May have changed.
yupyup
i added it to the staging copr
Yep. As soon as you install the anaconda-webui, it uses it
should be just that + an autostart config
I use the gnome setup thing, which starts it 🙂
also works!
i dont know why but it is super inconsistent on bluefin LTS
it sometimes shows up on the live mode (if the rootfs is based on erofs (at least seems like it)), new user mode (rootfs as squashfs w/ rw flag), or existing user mode (no idea what causes it)
when i build the ISOs locally it works fine, it shows that cool "try or install" thing
when its built on gh actions it doesnt work
when i was testing it locally i didnt have flatpaks or the container on the ISO so maybe thats whats making it think theres already an user on the ISO?
Ok so
just sticking it up and dealing with livesys-scripts seemed like the best idea
Gotta move the anaconda parts out
mm?
And use the hook instead
yup
its just that we are going to use the anaconda stuff for every ublue image
I don't want anything related to anaconda in the repo in fact
Other than well the container to OCI dir, that's not an issue
its just configs, we arent installing anaconda or anything on titanoboa
Well do it elsewhere, like the CI action
fair
hmm
I don't want to be tied to that thing
We can worry about that later. Let's get something working
All the work is in a PR right now
although we do need like a ummmm
we just had a meeting with neal
we may be onto something
?
@Zeglius i also would like to keep it "clean" of anacondas but theres some stuff anaconda hardcodes, like the container directory/format and stuff so i think we need at least some flag for an anaconda-supporting environment
Like what per example?
target container needs to be under
/usr/lib/ostreecontainer, flatpaks need to be under /flatpakThe directory path doesn't matter. It just needs to be an oci-dir and match the interactive-defaults.ks path
and some configs that might be able to just work on the hook itself
That's fine
Shouldn't interrupt the normal functioning
heh the edits
yeye
What I don't want to see is anything to do with anaconda in the main action nor the recipes
Exceptions are well the CI action
Zeglius is burned forever. I hear you brother.
ill just keep it into my branch right now then we can make it anaconda-agnostic
he's like, keep that shit outta here forever and for five generations
Yeah. Once it works, we can worry about how to implement it
Good, I had enough of these 30 mins I attempted to use lorax
real LOL
Don't want another
so @Robert how are you doing the flatpaks?
I haven't got anything fully functional yet, but symlinking /var/lib/flatpak/repo to /flatpak/repo and running
flatpak build-update-repo is closewhen i run
flatpak list on the liveISO it seems pretty madYeah.. I ignore that. Don't think it's anything we've changed today
i think it is tho
it used to work just fine
How? Just having /flatpak/repo makes it happen?
It shouldn't
We're not doing anything else
i think its the installations.d entry
How is it different to the liveiso one?
nono i mean the anaconda one at the end
like
/etc/flatpak/installations.d/anaconda.conf that anaconda uses
Yeah. I mean it should be the same as liveiso, just under a different name
the
liveiso one is just a provisory one that we use to get stuff installed in the container
it shouldnt be there on the ISO itself
lemme see if i can get it workingThe Anaconda remote is hard-coded in anaconda, so I thought it was needed
oh now it just worked 👍
yeah i think i mistyped it when i was testing it out
:)
🤡
OMG
anaconda seems mad still
At the ostree or flatpaks stage?
it need just a little more of tears and blood.
fails cuz of anaconda not knowinf the remote on the Anaconda flatpak installation
come on everyone, get out your ouiji boards and trinkets
symlinking doesnt seem to trick flatpak
yup. just a bit more!
Can you share the output of
ls -la /flatpak/repo?
Just a screenshot or something
Want to see if it has the summary filesjuust a sec
Also maybe try removing the installations.d from the rootfs
I can't actually see it on the old ISO, so the hardcoded value might be something else
aight so
@Robert
oh LOL THE THING
...
Missing the important bit 🙂
there you go
hahahhawdahah
How did you build this? Do you have the step that creates the flatpak repo metadata?
i just symlinked
/var/lib/flatpak to /flatpak
didnt do anything other than thatYou will need this command
https://github.com/ublue-os/titanoboa/pull/65/commits/dff21a2eb9fc1a7d80c16fda84fea62345c39ea3
did that on the live ISO
lets see if it works now
Yep. Those summary files are needed I think
IT DIDNT COMPLAIN
YEAH LES GOOO
omg the dream is real
hold me tulip
@bsherman bring it in chief
Don't celebrate too soon!
alright, installing this thing now
its pretty damn fast too for whatever reason
Maybe because we're bundling the OCI inside the squashfs? Could that make a difference?
shrug
OMG TULIP please new webui I need to see the future
lets see if the flatpaks worked LOL
That seemed too fast...
lol
flatpaks didnt get copied
:P
i wonder if we need to do something with the kickstart?
Can you try without installations.d?
sure
how does silverblue even do it tho
actually how do we do it with container-installer
Abstracted away into lorax or another similar tool I think?
ye
its some lorax template
flatpaks run just fine on the live ISO btw
Looking through the silverblue and old isobuilder action and I cannot see an installations.d
Everything else looks the same
trying it without the installations.d
symlinked
/flatpak to /var/lib/flatpak, generated the metadata with taht command, and added the ostreecontainer line on interactive-defaults.ks
finished the installation
Seemed too quick again 😢
@Robert it was waaay too fast
i think yeah
is that the new webui?
Nope
nope
Ah ok
still on the flatpaks
we are making the regular anaconda work then webui
ye
this is so close tho
anaconda doesnt complain, nothing seems "wrong"
maybe theres something neal told me
can you lie to it with mount instead of a link?
ill test it
you'll need something like this to make your stuff work properly though: https://gitlab.com/CentOS/Hyperscale/src/anaconda/-/blob/c10s-hs-v40.22.3.20/data/profile.d/centos-hyperscale.conf?ref_type=heads
GitLab
data/profile.d/centos-hyperscale.conf · c10s-hs-v40.22.3.20 · Cen...
System installer for Fedora, RHEL and other distributions - https://anaconda-installer.readthedocs.io
there's nothing here for ublue stuff: https://github.com/rhinstaller/anaconda/tree/main/data/profile.d
GitHub
anaconda/data/profile.d at main · rhinstaller/anaconda
System installer for Fedora, RHEL and other distributions - rhinstaller/anaconda
huh cool!
GitHub
anaconda/data/profile.d/fedora-silverblue.conf at main · rhinstall...
System installer for Fedora, RHEL and other distributions - rhinstaller/anaconda
But that bit is all working.
yeah
i mean, kinda
its defaulting to everything being enabled
ill try to bind-mount now
That didn't happen on Fedora...
That must be what Neil meant
ye
its the profile
we can just change it
whatever
i got the classic "device is active" error :)
restarting the install stuff again with the bindmount
oh man we can do btrfs by default for bluefin-hwe:lts
yes.
we shall capitulate to Neal's demands
I'll have my vengeance on that one on another day ... 😈
heh hell yeah
ok so, installed fine, booted fine, seems to just have not copied the flatpaks
even when
mount --binding /var/lib/flatpak to /flatpakping neal I know he's around
not super sure what makes it copy.... hmmmmmmm
It's got to be something tiny we're missing...
oh
@Robert build-container-installer seems to be doing something completely different
like what
No idea what that does...
Oh
Yeah flatpak uses ostree
it looks like its using ostreecontainer to install the flatpak repo
on another lorax file
(we dont need lorax)
We don't want it
its just some random file under /etc
Is not maintained
GitHub
build-container-installer/lorax_templates/flatpak_set_repo.tmpl at ...
Creates an ISO for installing a container image as an OS - JasonN3/build-container-installer
That's awful...
yeah it is some using some major jank
Oh. Check earlier in the file. It's updating anaconda.conf
omg this is why we arent getting the fedora-flatpaks repo
on f41 and stuff
GitHub
build-container-installer/lorax_templates/scripts/post/flatpak_conf...
Creates an ISO for installing a container image as an OS - JasonN3/build-container-installer
ill try to manually add that
same thing:
- symlink /flatpak
- ostreecontainer on interactive-defaults
- that other command for summaries or something
flatpak build-update-repo /var/lib/flatpak/repo
cc @James
also like
what about the person that made
build-container-installer?
are they here?
they seem like the perfect person to ask about this stuff@Noel I forgot jason's handle
SkyNet
why didnt we ping him like right at the start of this thread or something? LOL
I think that dude bailed long ago?
Because we didn't want to use Anaconda
true
we still don't want to use anaconda. Let's go with "we have no choice" so I can at least sleep at night
heh yea
man, first RPMs, now anaconda.
what's happening
its the curse
I've been living amongst the distro people for too long. I yearn for my cloud native mojo...
did you ping neal on matrix?
surely he knows
pinged him!
Btw how does one launch anaconda? Is it a desktop shortcut?
yup
desktop shortcut
the latest bluefin-lts-live ISO has it
we will wire it up to an autostart thing
Ah alright, for a second thought we were to lose the livecd session
oh nope
at least that part isnt going away
https://paste.centos.org/view/0f347665
everything flatpak related on the logs
I can ping him, but he will likely want us to try and contribute to the lorax based option that we already use 🙂
kill me, let's go back to the ouijia boards.
Lorax is dead, is no longer maintained
Saw already there is some migration planning towards kiwi
zeglius don't kill the buzz
i didn't think i could cheer this loud
Yeah sure, though there is still no livecd for fedora atomic
I feel there is not enough push for that
im cheering because lorax just sucks generally
also i thought y'all got livecd working
Yeah sure we got it working
Though would love it if we didn't have to do it ourselves
I guess atomic is still relatively niche compared to workstation
I think this will work
we just need the symlink thing to work
and then boom, titanoconda saves the day, and we'll at least have working live ISO with flatpaks and the new webui
then at the end of the movie Neal claims my soul, but what is the cost of one soul if it saves us all
LOL
well technically two because Noel was the first one taken
wait what happened to noel's soul
yeah only 100x smaller
but we will get there
Yeah, once we get arch fanboys into bootc
Made this discussion and will ping a maintainer tomorrow. Jiri seems interested in helping us usually
https://github.com/rhinstaller/anaconda/discussions/6335
If we work it out without the Anaconda team, let's update there to document it
GitHub
ostreecontainer | Install Flatpaks with Anaconda · rhinstaller ana...
Hey Anaconda team The Universal Blue team are trying to create a LiveISO for our Fedora Silverblue images. We feel we're 95% of the way there - we've got an offline live ISO working with em...
Back channeled Jiri to have him take a look.
Noel, nature is healing
Honestly, I've been taking quite a step back lately. Besides Summit, Fedora Flock, and maybe Minnebar, I haven't done a ton of work on Ublue as of late lol
it's ok noel
the healing is coming
soon this miniboss will be slain, then we can finally rez you
I didn't fear snakes before joining ublue.
why does it always have to be snakes
I dunno man. Then you decided to create titanoboa
MINNEBAR
excited to see you soon
did you get your tickets yet?
i forgor if we have a way to build live isos out of bootc images yet
because that's all I need
they did it with katsu initially but built their own thing (titanoboa)
yeah
i saw it at flock it was awesome
*scale
but how
@Kyle Gospo do you have the katsu config still
make sure you attrib if you use these capp
all i need was a thing to generate the iso image that you can run
bootc install on
that's pretty much it
end goal is to make an independent ucore iso image with no auto-rebase from fcos script needed
and all it takes is an ignition or cloud-init scriptYou want it to boot into a DE? Or just a terminal?
terminal
it doesnt really matter as long as you can boot into an environment where you can use the current live image with bootc
so basically what we're doing with live ISOs minus the anaconda part
i just forgot how it's done
FCOS live ISO has bootc on jt
but you still have to download your custom image from the internet
I meant as in directly booting the bootc image from the iso
bootc live isos
i don't know what you're thinking but we want to create a live ISO that:
- boots to a custom bootc image
- lets you run
bootc install with the currently booted image
which is what we have right now with titanoboaI was thinking for what you want since we can't dedup the booted live image and the bootc image you want the booted image to be as small as possible
iirc we already solved it with titanoboa
but yeah
my goal is to basically do this but put ucore instead of a desktop image
and I'll build an installer as an alternative to coreos-installer but uses bootc
polycrystal:
Could someone please remind me of the R2 URL of the latest Bluefin LTS Live ISO?
Or perhaps Tulip - could you rebuild a new ISO with all the fixes you've been testing for Jiri?
@Robert I am not sure if this link is getting updated but you can check. since its like 7 am here in brazil.
not sure when tulip wakes up.
Thanks! There's no rush. Jiri asked for a copy of the ISO so they can see what we're missing, but timezones do exist so nobody expects replies immediately 🙂
I don't think the ISO is updated - I can't see any GHA runs after we've discussed some fixes, so let's wait for a more up-to-date ISO
GitHub
Build ISOs · ublue-os/bluefin-lts@6da611d
Bluefin LTS, built on CentOS with bootc. Contribute to ublue-os/bluefin-lts development by creating an account on GitHub.
That was the first attempt, but Tulip was making changes locally.
ah, understood
ok I'm around all day to help push buttons, off this week, lmk. Just getting online now slowly
i might be stupid
Pretty much that's how it works yeah
Wait actually with that pushd might end up with an
rootfs/rootfs/var/lib/containers/storageI want to try the anaconda webui in one of my Xfce images. Is this the branch that people are using to get it to work: https://github.com/ublue-os/titanoboa/pull/65 ? I don't see it actually installing anaconda so there's probably a local hook that I'm missing
i jusr changed the format there
installing anaconda needs to happen elsewhere cuz you need different repos and stuff on c10s
the branch itself also isnt completely functional 😭
Ah okay, I'll wait until it stablizes then
@tulip Please can you push the latest ISO to R2 so Jiri can help us out with the Flatpaks?
oooo
the one that has anaconda should be there already
yeah I saw that
it was a bug
But it's missing a lot of the fixes though, right? Like no summary files?
fixed it
but you need to manually do everything else cuz that branch is all broken
yuuuup
also, bug in titanoboa
Yeah, I hit that one a lot. I just retry lol but we should look into it
weird
never got that one
katsu just generates a new partition directly with xorriso
i got that straight from the katsu source code
instead of doing some cursed grub-mkimage
oh huh
anyway, I'm gonna add a bootc backend module in katsu soon https://github.com/FyraLabs/katsu/issues/34
GitHub
bootc backend · Issue #34 · FyraLabs/katsu
self-explanatory: implement a bootc backend for building bootc-based images with katsu, should take in an image and optionally some disk formatting time to steal some scripts from https://github.co...
How difficult would it be to make an ISO that works and push it to R2?
Saying "run these 3 commands, delete this file" etc would be difficult to reproduce
By "works" I mean functions, except the Flatpaks aren't installed onto the final system
I.E. gets to the point we were at yesterday
titanoboa might become obsolete again 🧌
hopefully not at all!
we just need:
- summaries on the flatpak step (just need to add that part to the flatpak step to titanoboa on my branch)
- thats it i think actually
it was supposed to have more than one module but got too lazy
summaries on the flatpak step (just need to add that part to the flatpak step to titanoboa on my branch)This is already done
did you push to my branch?
just refresh the ISO on the bluefin LTS action now then! noice
Just the normal ISO workflow with "Upload to Cloudflare" selected?
the other thing that the anaconda configs werent working is working fine now cuz i fixed the bug on the action
yup
i think you dont need anything else than that
yai*
i wanna eventually deprecate titanoboa too cuz me no want to maintain this shit
kiwi seems to be enough for us
no not yai, that's readymade
Don't let Jorge see you saying that!
yai is gonna be deprecated with that
heh i mean, i did end up learning some cool stuff from neal
i actually meant titanoboa because I'm gonna add support for whatever it's doing in katsu
because it wasn't there
all I have to do is add a new backend
all it should do is just copy some bootc images into the root and generate the ISO the normal way
should be that simple
so that complaint about katsu being useless will be addressed finally
https://discord.com/channels/1072614816579063828/1072617059265032342/1357715160902668418 this one
my plan is to create a derivation using a containerfile for the live image specifically, export that rootfs, do the work in that rootfs, then copy the original bootc image into the rootfs
so the only part podman gets run is pulling, exporting, and generating initramfs
on readymade?
no, on katsu
oh good LOL
are you getting confused between katsu and readymade
it's not like anaconda where anaconda both builds and installs the system lol
nah i just didnt know the context
katsu builds the image, readymade installs (a readymade image, get it) to disk
very obvious naming :clueless:
LOL @Robert
it's named after this https://youtu.be/jg09lNupc1s
Ado
YouTube
【Ado】レディメイド
大人にだけはなりたくなんかない
ダウンロード/ストリーミング配信はこちら：https://umj.lnk.to/ado_rmID
↓レディメイド Inst音源↓
https://piapro.jp/t/pIC2
Vo：Ado
https://twitter.com/ado1024imokenp
Song & Lyrics & Arrangement：すりぃ様
https://twitter.com/iii0303_8
Bass：malo様
https://tw...
lol
Yeah got that as well, for some reason it works if I run the iso recipe alone afterwards
odd
Inadvertently, we solved this lmao
https://github.com/rhinstaller/anaconda/discussions/5888
GitHub
Consider deriving from fedora-bootc · rhinstaller anaconda · Disc...
The Anaconda ISO has historically been built with lorax, and more recently osbuild. One thing I'd love to see is support a spike on what it'd look like to build Anaconda as a bootc containe...
wtf did i do wrong
working pr if anyone wants to help port titanoboa code to rust https://github.com/FyraLabs/katsu/pull/35
nvm i did it
@tulip katsu will now support bootc as a first class citizen once I merge this PR, so you no longer need to actually maintain titanoboa anymore
🚎
@madomado
the way I did it is pretty funny though
We will keep it around regardless
Bash is easier for prototyping than rust
we just gotta get the cycle closed
anaconda working -> then we can change anything else
maybe katsu, maybe kiwi
so katsu now has a
bootc builder module, and accepts image, derivation, and context as parameters
derivation is passing in a containerfile to create an extra layer for the live imagei personally dont care
But I have more features 😉
Better a devil well known, than a devil yet to know
Besides tulip had quite a fight with the isos :v better to maintain the legacy of that battle
i mean
the centos altimages SIG does this stuff we've been doing with kiwi instead
i wanna get anaconda going then it would be just amazing to use kiwi instead of maintaining this
hyperscale SIG also is using kiwi
for live ISOs
i'm still just happy y'all have live isos in the first place
real
Yeah, if we somewhat can use kiwi while keeping it easy will advocate for that
like it doesnt matter how we make it its all the same shit
Keyword being easy
we'll just preprocess the rootfs like we are doing right now, then use kiwi for the ISO
a reminder that katsu can both build oci images and now build live isos from them — and is built to run in a container in the first place
Wait so needs to run podman inside the container
discord no emdash support 😭
please tulip, my brain
i don't like kiwi and have had issues with it in containers but that's my two cents
LOL whats with ur brain!
oh wonderful 😭
the problem is solved we don't need to sidegrade stuff
Tulip in her quest to overengineering
yupyup
the goal is to get working ISOs
ye thats what i was talkin about
Other than the flatpaks, what else? Was webui tested?
didnt touch it
Well that and anaconda team adding bootc support
but its not going to work if non-webui anaconda doesnt work
i mean it works
?
we are just using the older way of doing it
with anaconda
The older way being?
ostreecontainer --dir=/usr/lib/ostreecontainer
on /usr/share/anaconda/interactive-defaults.ks🗿
its the shit way of doing it
Yeah so doesn't work
ye
So if an error happens we would get the typical 1 or 2 and nothing else?
i think its different on the webui
we just gotta get the installer working cuz just having a live ISO is a big improvement
then we move to readymade or something else if anaconda is being too bad still
we are just descoping yai
For that we still need them to add bootc support
yup they were here talking about it recently
Recently as, after they asked for our feedback and Jorge had the meltdown?
meltdown?
oh i was talking about readymade
but this kind of shit is why jorge was completely right
having this much trouble on getting flatpaks working is so annoying
I mean anaconda yeah
pure meltdown
expecting an OS installer to install an OS is a bridge too far lol
😭 real
I mean, tbf bootc is kinda new
these issues have been filed for years
We are just going too fast for them
rpm-ostree and ostree arent new, all this stuff is mostly the same for them you know
I think once I told them no more free marketing from me until they fix things may have helped.
imagine an os installer that installs oses smh that's impossible
ready made images
on the plus side we don't have to edit the mission statement.
yaaay 😭
Well now part of these issues can be forgotten now that the backend moves to bootc
anaconda is still not great despite having bootc support now
No longer relies in anaconda that much
until we run into another random issue
because anaconda relies on a hardcoded path
and 1 billion year old things
Yeah but 99% of the deployment is done by bootc
Which is what I care about
anaconda has burnt a hole in my heart
I took a look at how anaconda installs an ostree image and is a nightmare
It handles manually the bootloader installation
im not surprised or anything anymore
Now they can nuke like the entire module and will still work
Because bootc handles all of it
they'll do it once RHEL16 needs it
dont worry
its gonna be soon
:huh:
they'll just be depending on a 10 year old library in the middle there just because
Like for real or?
might as well
webui doesnt fix this stuff either
its just using the shit backend again
I mean if they asked for feedback is that they are going to start working on it soonish
The WebUI was tested, and fully works in F41.
I didn't face any ostree errors during the deployment so can't comment on those error messages, but the errors for Flatpak are quite good.
Everything else we need is ready to go - we know exactly what we need to do for secureboot and bootc switching to the registry
I didn't face any ostree errors during the deployment so can't comment on those error messages, but the errors for Flatpak are quite good.
Everything else we need is ready to go - we know exactly what we need to do for secureboot and bootc switching to the registry
We need to see how the errors logs looks like when ostree fails
"device is active"
To be fair, that's probably what ostree itself returns
🗿
Yeah thats why is useless if they dont switch to bootc
They are working on it.
For livecd at least, is good enough, and if anaconda behaves the same exact way, its an improvement
Once we get flatpaks to behave, we need to move the anaconda stuff to the CI action (as reference), merge PR and wait for them to get the bootc bits
That's only in 2043, plenty of time to prepare everything
merged the thing
katsu live bootc isos are now real
documentation is my passion
new ananconda breakage now...
Ah yeah, attempting to install the bootloader manually
And that's why I'm so stubborn with anaconda switching to bootc :v
yepp...
like I'm convinced right now that the installer for fedora-silverblue et all... are on an older version or something
Readymadeeeeee, where are youuuuu
soontm
i asked mado to add a config option for bootc backend idk where he went
i forgor about it
@madomado aaa
either that or make the patch yourself :3
for rdm?
yessir
what do you want it to do
can't we just use wholedisk?
a
needs to be custom with blablabla cuz
bootc install to-filesystemI forgot
mado forgor again
yes use wholedisk
it is wholedisk
but instead of CopyBlocks it runs bootc install to-filesystem
thats it
madoim
full format with repart should work fine enough tbh
all it's gonna do is partition it similar to bootc install to-disk but instead of partitioning manually it should call repart instead
actually
hm
yeah that worksw
mado forgor again
wtf does bootc mean by this
@tulip Any idea why the current live ISO doesn't contain the latest titanoboa changes (such as generating flatpak repo data)?
I can see it being invoked in GitHub Actions
https://github.com/ublue-os/bluefin-lts/actions/runs/14335050064/job/40181981936#step:6:1258
GitHub
Build ISOs · ublue-os/bluefin-lts@b0270a6
Bluefin LTS, built on CentOS with bootc. Contribute to ublue-os/bluefin-lts development by creating an account on GitHub.
oh those are on the other branch
my anaconda-testing-whatever branch
maybe we just merge it?
It's using ublue-os/titanoboa@anaconda-webui-format
oh huh
the branch seems fine....
hm
The build shows those commands were run. But neither that, nor the /flatpak symlink is present on the ISO
OH yeah i know
i dont have any idea why that doesnt show up tbh
we are probably putting the symlink on the wrong step or something
we just need to
ln -sf rootfs/var/lib/flatpak rootfs/flatpak at some pointPlease can I have a quick review
https://github.com/ublue-os/bluefin-lts/pull/439
Enables the hyperscale repos during the ISO build script so Anaconda can be installed
Don't need to run the full image builds since it doesn't touch them
merge now or wait for the iso builds to finish?
The part I've changed has already ran successfully, so we can merge now without the queue
done
We are so close!
Pretty sure I know what we need to do, but not sure how...
Pretty much it needs to see /var/lib/flatpak/repo as the remote
flatpak remote-ls anaconda shows nothing, whereas we need it to show all the Flathub apps we've added.Pretty much it needs to see /var/lib/flatpak/repo as the remote
maybe we can add the flathub remote named as "anaconda"
LOL
I hate it, but maybe...
I'll come back to this tomorrow.
If
flatpak remote-ls anaconda shows all apps inside the repo metadata file we've generated, it should work@Kyle Gospo and @Noel expressed interest in wanting to talk over katsu and readymade with bootc for ublue. is everyone avalible at <t:1744387200:f> ?
CC: @Owen @madomado @tulip @Zeglius @j0rge @ultramarine dev @thirtyseven @lea
please thumbs up 🥺
my apolo cheese for mass ping
also for the late night ping for the americans
Is that PST or?
it's a timestamp so it'll change to your timezone
Oh okay
I'll be there
has to be around then because of mado in hong kong and cappy in thailand
11pm good enouigh
midnight for mado...
idk
mado already said was okay
He said it was fine
im entirely fine with both of those tools and we can just make them into like a github action or something
sounds pretty damn good to me
what if we package it on terra + terra EL and put readymade on the ISOs
bam done congrats we fixed a 5y.o. problem
they are designed to run on gha
that's the main usecase
already in both repos you can install katsu, readymade, anda, and soon taidan from terra
hell yeah
Based
i love (ab)using gha
we build 78 images across 2 repos in gha with a little help from a server we own and the arm cluster
Microsoft has enough compute :clueless:
*69 i was off
Tbf this list doesn’t seem too far fetched given you have 60 laptops
i only have 45-50 owen has more
I have about 100
aight so gamers
we just need a readymade build for f41-42 and el10
with the wip bootc stuff
then i think... this is done, right?
Does Readymade support Flatpaks? I don't see anything in the codebase that suggests it does
Unless there's generic postinstall scripts we can add, which would also enable the MOK enrolment
meanwhile Im testing to see if katsu works
dagger being the GOAT
no idea
i think it would be like a custom script or something
Okay seems there is an small issue with katsu when building the target containerfile
meow?
Gonna try adding --network host
uhh
you put it into your image
Alright that fixed it
Ah... So you're saying bundling Flatpaks into the OCI?
We made the decision not to pursue that route a while ago
We made the decision not to pursue that route a while ago
not happening for us tho
I see…
but like
then the other option is put it into taidan
but obviously that won't work for you guys
we can just copy the
/var/lib/flatpak to (rootfs)/ostree/deploy/default/var/libhmm
it works
if theres any way to do that
maybe I should add a postinstall module that just
runs a script
yup
?
hold give me a min
rsync -avhZ --info=progress2 -> copies the flatpaks n shitYeah. A drop-in directory with scripts or something would be great.
Even better if there's convenience environment variables for things like the rootfs path
wait a moment
so katsu simply uses a containerfile to setup the root filesystem of the squash right
if you can try also setting an env variable for the target rootfs cuz itll help so much on the scripts 😭
cant we simply add the flatpaks in that containerfile?
worst case we can figure it out btw
what about this
But then the flatpaks are tied to the OCIs. Massive OCIs and updates
I'm feeding this into stdin as json
No is not
is that ok?
sgtm
there are two different images, the ona in the squash, and the other to be installed right?
we dont even need any info straight from readymade on the scripts if thats a problem
the ideal thing would be to have a rootfs mount somewhere
ah don't worry
the root mount is
/
it's chrootedI'm not sure I follow.
I'm talking about getting the flatpaks onto the installed system
LiveISOs with Flatpaks are easy enough - we've proved it works
huh
we do need access to host files tho
uh
maybe a
/run/host mount?hm…
the live ISOs have
/var/lib/flatpak in themright right
we'd just copy them to the right place on the rootfs and done bam
@ultramarine dev maybe we should mount the host rootfs in the chroot as /run/host?
thats what distrobox does
?
what do you mean by that?
I'm making a new postinstall module that runs a custom script
they want access to the host
You are confusing the squashed rootfs image with the one to install
but the postinstall modules are chrooted
in tiffin
hrm...
yeah
gtm
/run/host works
do we just add it to the tiffin container?
yup
bind mount to /
i think thats all we need from readymade
Could you explain how you're expecting it to work?
right, impling that
if all goes well tiffin should just trigger the unmount on Drop
katsu does some stuff and he wants to put the flatpaks on the ISO rootfs using katsu i think
so how we are installing the flatpaks in titanoboa?
huh, funny
cp /var/lib/flatpak rootfs/ostree/deploy/default/var/lib/flatpakwe just chroot in the root filesystem with podman right?
nope
Yes. I know we can install flatpaks into the live ISO.
But the point is getting them onto the final installed system. So when you boot from disk, it's there
But the point is getting them onto the final installed system. So when you boot from disk, it's there
we just put the flatpaks in the rootfs using polycrystal
and that probably would be slow as hell cuz we are using fuse-overlayfs sadly
that cp thing does it
k which branch has the newest stuff (titanoboa)
can i ask which editor is this
is this nvim
is this riced nvim
thats yazi, a tui file browser
oh
smh i thought it's nvim with the tree plugin
Ok so, the livecd has the flatpaks at runtime, correct?
yuup
just use helix :clueless:
@madomado crazy idea but imagine wasm plugins for readymade
helix + yai + zellij the best dev thing ever
like zellij
zellij my beloved
So, yeah as I said, you are chrooting into the root filesystem with podman
rdm path to script btw
whar
/etc/readymade/postinstall.sh?So yes you can bake the flatpaks into the image
nope, titanoboa just uses some other container so that we dont get the host runtimes on the flaptak stuff
/etc/readymade/postinstall.d/*.shZeglius: cant we simply add the flatpaks in that containerfile? p5: But then the flatpaks are tied to the OCIs. Massive OCIs and updates Zeglius: No is not Zeglius: there are two different images, the ona in the squash, and the other to be installed right? p5; I'm not sure I follow. p5: I'm talking about getting the flatpaks onto the installed system p5: LiveISOs with Flatpaks are easy enough - we've proved it works Zeglius: <Posts screenshot of Containerfile> Zeglius: You are confusing the squashed rootfs image with the one to installIt seems you're talking about putting Flatpaks into an OCI image. We don't want to do this.
okay
we are mixing concepts
yeah
We have two OCI images with katsu
actually this would work too, whatever is easier to implement for you is the best solution
the one that gets exported to a squash, and the one that we install
With titanoboa, you are installing the flatpaks in the squash one
the funny
because I'm extremely lazy
sounds good to me
so yeah, we dont need to break our heads to add flatpaks separately, just install them in the containerfile
we arent installing the flatpaks on the container
Ok. So you're talking mainly about Katsu (LiveISO creator), whereas we're talking about Readymade (installer)
we are installing to
/flatpak that is /var/lib/flatpak on the ISO rootwhen you run podman run --rootfs, what do you think that is?
its a container
we are thinking about way too much shit at once
three people saying three different things
lets organize this
zeglius, what you wanna do with katsu
katsu as a replacement for titanoboa right
sounds pretty damn good to me
If we wanted to yeah
less maintenance for us
Currently Im trying to build an iso with it
ye
but like what are you talking about the flatpaks tho
We can also lay this all out in the call soon here too. So many things to go over in text lol
we cant use bootc install to install the flatpaks if you wanna do that
Im not saying that
Again, both are separate images
yeye
even funnier: just execute them
instead of calling
shI ack
mado what if you just use
std::process::Command to execute the whole file, so now literally any executable under the sun can execute in there?
Python scripts, Ruby scripts, shell scripts
well that means you need shebangs?
what if no shebangs
yup
undefined behavior, but atp it's their fault
nah
let them call python from the scripts or something
whatever they want to do
sobbing
speaking of where do we wanna call?
why not do it like katsu,,,
we literally do that in katsu mado
vc in here or jitsi or smth
idk im pissing myself here
The tty vc channel in this category?
bet
set -x; for f in /etc/readymade/postinstall.d/*.sh; do sh $f; done
this is current code
set -x; for f in /etc/readymade/postinstall.d/*; do ./$f; done
this feels a bit
dangerous?
it executes everythingyeah but it's in /etc
wait
wait
instead of putting them in
/etc/ put them in /usr/share/
so it will be immutable when run inside bootcsomething something
/usr/share/readymade/postinstall.d/
ok
honestly just do all of them by this point@Kyle Gospo @Noel @j0rge @tulip @Zeglius @Robert (p5)
(Hope I got everyone)
good luck! I've got other things going on but happy to see progress, let's goooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo
did
lets gooo
come in kyle
pushed to here
flatpak
shit script if you dont have it
tulip get in vc
@tulip join if you can
this is a threat
kyle is very scary
ill get in vc soon just a sec
GitHub
GitHub - ublue-os/katsu: The vicious image builder
The vicious image builder. Contribute to ublue-os/katsu development by creating an account on GitHub.
@tulip @jade @madomado @ultramarine dev @Owen
anaconda is fucking weird
you need some funky stuff to make the flaptaks works and we dont know it yet
it works™️
we might as well just support just readymade
it works on both anaconda and readymade
because it puts the flatpaks right inside the sqfs
what would be ideal, or better, what does readymade do today that would work w/ this setup?
that postinstall module that @madomado made
it would be just
rsync -avhZ --info=progress2 /var/lib/flatpak (rootfs)/ostree/deploy/default/var/libhttps://github.com/FyraLabs/readymade/pull/50
available here
GitHub
GitHub - bazzite-org/titanoboa-readymade: bootc installer - eats an...
bootc installer - eats anacondas. Contribute to bazzite-org/titanoboa-readymade development by creating an account on GitHub.
lol when a bunch of random people on The Internet do in no time what the Big Distros can't seem to deliver.
Welcome – Fyra Developer
The Fyra Developer Documentation
GitHub
GitHub - ublue-os/titanoboa: bootc installer - eats anacondas
bootc installer - eats anacondas. Contribute to ublue-os/titanoboa development by creating an account on GitHub.
this is what we need to do
https://github.com/ublue-os/titanoboa/issues/66
GitHub
Switching to Readymade · Issue #66 · ublue-os/titanoboa
Todo Fyra rm hardcoded branding move Readymade to Terra readymade-config-ultramarine UBlue add Readymade to image config enable GNOME OOBE
lets keep discussion in here so we don't loose track
@jade when do we want to plan on meeting up next? Not to set any deadlines or anything, but I'm curious how quickly we can each get this going.
im engineer wrangling so one sec
@lea @thirtyseven @madomado @Cappy Ishihara @Kyle Gospo @Noel @j0rge @tulip @Zeglius @Robert
GitHub
feat: add WIP live ISO workflow · ublue-os/bluefin@048d683
The next generation Linux workstation, designed for reliability, performance, and sustainability. - feat: add WIP live ISO workflow · ublue-os/bluefin@048d683
workflow is working
we just need a config now
Please reference this issue in PRs and issues for doing this work across projects, a simple link will tell GH what to do
Readymade in Terra now
Backports and all
To answer this, honestly whenever lol. We should have our end done pretty soon but if anything comes up on either end I'm down to meet again. At the latest might be nice to meet before anything gets released
GitHub
[el10] switch helium and hydrogen to proper naming (#2058) by gerbl...
Backport
This will backport the following commits from frawhide to el10: #2058
Should fix the issues with building the readymade package in el10: https://github.com/terrapkg/packages/actions/runs/1...
readymade wasn't building on el10, this should fix the broken deps
reviewed
good catch man
oh shit nice
Readymade running in the Blue95 live ISO. It crashes when selecting the language but still cool
NICE
huh why does it crash tho? do you have any terminal logs?
It's the same issue gerblesh reported in https://github.com/ublue-os/titanoboa/pull/67
GitHub
feat: initial install of readymade by gerblesh · Pull Request #67 ...
Initial work for installing readymade onto the image (from terra). Currently readymade isn't available for centos 10 due to dependency issues (terrapkg/packages#4299) so this doesn'...
@madomado you should ask about rdm bootc stuff here instead
currently
/etc/ostree/prepare-root.conf is
and the image used is containers-storage:ghcr.io/ublue-os/bluefin:lts
and
Is the image inside containers storage? I.e. is it already pulled?
uh
how can I pull it
because most likely no
podman pull?
or is there a way to make it pull automatically using something other than
containers-storage:?
I'm going with podman pull for now thenYeah. We run pull inside GitHub Actions to make the image available in the rootfs container storage.
You might be able to use something like registry:ghcr.io/... so it pulls automatically, but untested and that wouldn't be our use-case
You might be able to use something like registry:ghcr.io/... so it pulls automatically, but untested and that wouldn't be our use-case
trying
but hmm I mean
actually so I need to check if it is pulled
if not then I need to pull it I guess?
that's a surprising amount of blobs lmao
You could do that. Or expect it to be defined inside the config file and don't worry about it
I guess
well
well well well…
Ah.. umm.. not sure how to get around that 😆
yup… w
When the image is inside a live ISO, it doesn't do that because there's some fancy squashfs magic that hard links and deduplicates the contents of the OCI with the rootfs contents
can I get uhh
do you guys have like an iso for me to test readymade on
with the container pulled
Someone made a PR. Will check how complete it is
If you just need a ISO without readymade, and install readymade in the live environment, we have one of them
ah yeah that works
I'm on mobile so can't easily find it, but could you search this thread for 'bluefin-lts-live.iso' and there will be a download link
found, thanks
Oh.. That's not using containers-storage. It's using oci-dir
1 sec. Loading up my machine and will make something
Where should we fetch the readymade libhelium dependency from? It's not available in your um repo or default fedora repos
Ohh, it's in Terra, and so is Readymade
So we should just switch to sourcing them both from terra
yeah
Yeah that's one of the reasons we just dump the container image rootfs in a directory, instead of doing everything in a container file (at titanoboa)
Is simply too annoying to deal with nested podman
You would need to use fuse-overlayfs in the inner podman instance or something at build time
Wait nvm forgot this is not katsu, forget what I said
https://download.projectbluefin.io/bluefin-live-41.iso
Our (soon to be) GTS ISO with the image in containers-storage, without any installers. Although I've not booted it up to see if it looks like we expect.
Making a new build that contains readymade too
readymade doesn't work yet I think
That's completely fine. If we have a pipeline that includes readymade, it will be ready for when support is there 🙂
good to know
Bluefin PR to fix the ISO workflow
https://github.com/ublue-os/bluefin/pull/2336
Can be merged without queue
GitHub
feat(liveiso): add flatpaks, fix readymade, correct image tag by p5...
add flatpaks, fix readymade, use correct image tag
Can be merged without image builds
ok so basically, we're getting the infra ready, readymade folks just need to push RPMs and our next runs should be good.
We just need Bootc support in readymade, to figure out the config file contents and to make scripts to copy over Flatpaks/enroll MOK keys
I thought tulip had the config file?
can I run the action? build-iso-test.yml?
Not sure. I can't see any other branches or anything
I wanna see it running
Sure, you can run it
There's a new ISO in Cloudflare already, with the installer present. Obviously it doesn't work yet, but we can play around with it
oh ok
I'm more of the feedback loop guy this morning hahaha.
This is the download url
where should I put the v4l2loopback package install in bluefin?
since it's from a copr it doesn't go in packages.json does it?
and it's not replacing anything so I think it doesn't go in the shell script
https://github.com/ublue-os/bluefin/pull/2337
ok this should work
I'm pretty sure terra isn't enabled during the packages.json install
GitHub
bluefin/build_files/base/05-override-install.sh at 3ff7e659d827648c...
The next generation Linux workstation, designed for reliability, performance, and sustainability. - ublue-os/bluefin
we are using sentry things in here
oh sorry moving to bluefin dev
that's only on F42.
We also have to figure out fwupd for F42 as well.
yeah I dunno on that one
we may have an ISO today so working on that
If you're hoping for a fancy new liveiso, it won't be today...
v4l2loopback install from terra is added. If it fails. it just moves on
working awesome here!
I am just here to click buttons
it boots so fast!
I am trying out the bluefin iso but
this is what happens when I click
Install Bluefin LTSno secure boot
use
bluefin:stable not ltsit also happens to old bios
just disable it
true also that
i tihnk theres the f41 iso somewhere
Here!
With readymade installed
oh nice you managed to install readymade too?
where can I download?
oh
the loop is closed. 😄
thanks
amazing les go!!!
what's the imgref again?
actually I can just check later once I get into the livesys
ghcr.io/ublue-os/bluefin:41
i think
I'll try to boot with bios instead then
because don't want to figure out how to disable secure boot
on virt manager?
yeah
need to make a new vm and disable it on VM creation
just put a regular UEFI without secure_code
bios -> UEFI
without secure chungus
there you go
should work
yeah that is kinda what I did with the lts one
what the sigma
im pretty damn sure that works
ok this works
heh aight
something wrong with the lts one I guess
shrug
wait
bluefin is a kind of dragon?
🦖
afaik its a ummm
I thought it was tuna?
actually ill just refer to the dino expert
@j0rge
the betrayal
It's a dinosaur
Wasn’t it a tuna fish
it's on the forum, I added a tuna video too
omg true
Universal Blue
Isn't bluefin a tuna fish?
Bruh Like the project, but not understanding the dinosaurs. Every time I shutdown or reboot, the plymouth dinosaur logo makes me feel like I’m visiting a 4-year-old’s bedroom. “Raaarrrr!” I get someone on the team is like, super enthusiastic about them, but the name “bluefin” isn’t even a dinosaur, it’s a ginormous mercury-lade...
MAGURO
oh shit nice
@j0rge bootc blast the disk
And we can enforce whole disk installs, completely removing the custom partition option IIUC
uh
I ran out of memory and uhh
it killed the tuna,,,,
bootc blast the disk!
How much memory was it?
Even I didn't have issue with ~5 Gb of ram
it's not because of the vm
Oh host
my host ran out yeah
so the kernel OOMed
At least yours oom, mine freezes :v
I was running rust-analyzer for readymade + the vm
try pressing
alt+prtsc+f to invoke the oom btw
helps a lot
you need to enable kernel.sysrq = 1 in /etc/sysctl.d/ thoOoooh nice
@madomado completely different topic. Can you guys add the key.asc to terra40-source? It's missing. But it is in terra40.
@Cappy Ishihara
bootc failed hmm?
is this a bug with bootc itself?
copying static EFI failed?
what the sigma
It's bootupd that's failing. Checking
huh i wonder if c10s would break then
invalid args???
cursed fs
cappy, person above asked for key.asc
oh shit also you need
bootc install finalize
after bootc install to-filesystemjust that? no args?
rootfs as an arg
just like the ummm
ah ok
other one
doesnt it auto finalize
i dont think so
nowadays
its on the finalize docs
no idea how this works
Do you need that? I thought it was something internal
i read the docs on the new bootc-install-to-filesystem
looks like it??? i dont know
and there's an flag to disable auto finalize
so I assume it's implicitly set to auto finalize
it's kinda not in the official docs?
oh we'll also want to have
--target-imgref= as like a config or some shit at some pointits not there???
damn wtf
oh @madomado
--boot-mount-spec=(rootfs'es boot partition)
just found ityeah
Should that on the bootc command?
so e.g. /dev/vda1 in this case?
you need to mount /boot in the root, mado
did that
on the install afaik
odd..
--boot-mount-spec=/dev/vda1?no idea
or the path to
chroot/bootprobably an uuid or some shit
if you are already on the ISO you might want to manually run bootc install
actually
wtf?
well it's a mounted partition for sure
yeah it should implicitly do that
seems like it needs to be
/boot on the hostthe error is invalid args, so something must have went horribly wrong with the ioctls
the live ISO doesnt mount
/boot to anything by default btwno? I did the same thing with /boot in the chroot and it works fine
tested with manual bootc install on bluefin lts
maybe try manually bind mounting the chroot's boot to
/boot on the host and see if it works
bootc install to-filesystem?yup
complete "i dont know" momento
@M2 should be there now hopefully
I tested with /dev/sda (sata emulation)
GitHub
Dual boot support for bootupd with static GRUB config installations...
bootupd & static GRUB configs don't support dual boot yet. Previous title: Install of Silverblue installer image with bootupd enabled fails due to grub2-mkconfig failure Describe the bug Af...
so
...
a hunch: virtio is fucked somehow?
probably not
@madomado idea: try rebooting the VM with SATA emulation instead of VirtIO disk paravirt
so it appears as /dev/sd*
real
just a hunch
will that fix the bug?
probably not?
I do not know
try it and see anyway
it's a hunch
let me try --boot-mount-spec first
anyway, a minimal bootc install to-disk layout is this:
- ESP
- root
no xbootldr partition here
with a huge btrfs/xfs/whatever partition too
thats the root
i am actually on a
bootc install to-disk disk right now
there you go
the bootc install to disk layout is 2 partitions on EFI
3 on BIOS
the first partition is actually the MBR placeholder
funny
would strongly prefer having /boot and /boot/efi be seperate. ostree uses symlinks in /boot
yeah
there is no xbootldr
They are going to be gettign rid of /boot shortly
https://fedoraproject.org/wiki/Changes/BootLoaderUpdatesPhase1
i mean, the guy that made ostree had this as the default so i think its ok right?
no xbootldr here
only MBR header, ESP and root
i don't understand how it's supposed to work without being able to read /boot
from what I'm assuming, it straight up just has GRUB read the root partition instead of xbootldr
wonder how he does luks then
since that is a default option
luks is locked behind a config because the default luks thing is broken
it just sets it up with tpm2 and it cant be unlocked with a password
LOL
i tried luks and seemed to work for me
yeah
tpm2
based
merged EFI and
/boot partitions + huge luks rootfs
in my system /boot is a mount to that ESP partition
and /boot/efi is just a folder
I'm confusion
same tbh
this layout is super weird
so /boot is the esp?
i cant seem to find anything separate from it
that's cursed tbh
oh.
so boot-mount-spec should be the esp instead maybe?
i have no idea
that would make sense with this layout
please absolutely do not quote me on the bootc implementation
what about… our layout
/boot is.... on my rootfs?thank you
wha
well the normal layout
which is /boot + /boot/efi
can you do an lsblk tulip?
lemme get the json thing
good nighty chat
message too long
sec
just do an lsblk
screencap it
or lsblk
no arguments
yea
curious
uhmmm is this your host?
yeah
im right now booted off of it
they are doing the same shit the live ISO is doing
do you have an lsblk of an install-to-disk
this is it
what?
i did
install to-disk for my hostwell, /boot in this is on the root partition
yeah i also got confused about that
how come they do that
what's in /var/mnt?
since that looks like an ESP sized item
i manually mounted my ESP there
/dev/nvme0n1p2
it isnt mounted by default
how are they even mounting the boot partition how does this even workill boot into my initramfs and check the layout there
brb
oh shit
/boot is /sysroot/bootostree black magic
\
|
whoops
LOL
/was just reading the massive message string when I woke up
\
-
loading bar
just add a few
\rs and we get fancy TUI loadingHi there, how do you install this?
I run install to computer, crashes when I choose a language
yea its not working yet
we are slowly getting it working
i was testing some stuff out right now actually
so theres no way to install it for now right? is the goal of the iso then just testing?
yup!
its there specifically for testing and then getting it working, whenever it works we'll publish it on the forums and stuff
got it, thanks
does bluefin still have download link to the bluefin-latest branch?
i dont think we have a latest branch ISO anymore
just the stable one
they should be refreshed i think
ok cools
not sure tho! might want to check it out first dunno
looks very nice tho, love the dino dark wallpaper
its actually super close to being functional too :)
we'll probably release mid cycle on f42 with it if EVERYTHING goes well
yea that one got removed i think
oh man this looks great
@j0rge check this out
ill record my screen
it failed but it looks just so gourgeous
oh wonderful mkv doenst work
@Crono the dvd player thing
With a little branding, this is gonna look fire.
MAN imagine putting a little logo on the target hard drive
this is looking good, I was planning on tidying up my crappy readymade PR/testing this out if you're able to VC
im able to vc right now
if ur up
yeah I'm in tty
im here right now
RTC connecting wow
sec
GitHub
bluefin/iso_files/configure_iso.sh at main · ublue-os/bluefin
The next generation Linux workstation, designed for reliability, performance, and sustainability. - ublue-os/bluefin
@Gerblesh
@madomado
do you have a config for bootc readymade?
we are trying to test it out
cc @jade @Cappy Ishihara maybe you guys know it?
here @Gerblesh
YESSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS
❤️ the music and dvd logo
ok all we have to do is make that work
boom
ITS SO CLOOOSE
THEN WE SHALL BE DONE.
like man its ALMOST DONE
HELL YEAH
Just need to dd a disk! All we need, dd the disk!
@tulip removed the
stage!() macro that was throwing a panic! at unwrap() and got a systemd-repart errorPOG
i cant read tje log cuz im on my phone did it work?
no it did not, systemd-repart is was having issues with
/boot/efi
Failed to open source file '/boot/efi': No such file or directory
try moubting the target ESP to /boot/efi
@madomado just in case
I know this is a brand new feature so, it might be a couple minor changes still
oh also remove the grub2 thing from the config if you have it
yeah I removed that
so apparently there is no ESP partition
seems to not have partitioned the drive
repart should've done it for you
it says "future partition"
I see
it tries to copy /boot/efi but there is nothing there
in this repart config:
doesn't seem to be needed to copy over the esp partition stuff because bootc install should take care of it right?
oh yup there wont be shit there
where are these files located?
in the install
you need to copy the static grub binaries off of bootupd's directories
/usr/lib/bootupd/something
I'm confused
bootupd supplies you some grub stuff
theres the configs and something else that you need to put on the ESP partition
bootupctl shows you the stuff that it installs for you
why can't we just setup the efi partition and let bootupd handle the rest?
yeah thats what we should do
or does it require files to be installed /boot/efi first
just get rid of the snippets that clone boot configuration
on the readymade source code
I just want to remove the last line in the repart config
well cant you?
idk where these configs live on the system
just remove it
bootc install does it for you
OH
ooooh
those get made by readymade i think
partitioning is working without that copy
🙏 that bootc install works
OMG LETS GOOOOOOOOOOOOOOOOOO
video meeeeeeeeeeee!
I'll probably get the same error as mado tbh but we'll see lol
oh, nm
let's hope then
OMG EHAT
please dear LORD make this work
lets see what happens.
even then i think itll work honestly
might take a bit
it's been "populating" the FS for a hot minute lol
oh shit nice!
install and check btop
you'll see skopeo copy there
oh god
no it's copying the entire rootfs
not using bootc install
badbadbad
LMFAO
HAGDHSHSHHSH
🤦♂️
DANGER
DANGER
Hold onto your butts!
I was so confused I was like wtf is it "populating" the rootfs with lol
so it got the wrong copy mode
by accident
0: neither $COPY_SOURCE nor [install] bootc_imgref = "…" are supplied
:dispair:
not even
we just need different repart configs
and clearly it didn't load the config toml properly somehowdid it try to load the env variable at all?
like their repart configs both partition and populate the system, where ours should simply partition and let bootc do the rest
yupyup
doesn't matter it should still be able to load the value from the toml
if you get it working plis send ur the patch itll be awesome
my "patch" is hacky in that it manually sets a bunch of config values and gets rid of a macro (which is probably used for something important-ish) but yeah I can try to clean it up
sgtm
so those env vars are also not being loaded
and toml is being ignored
but manually overriding in the code works
anyhow it's complaining that the rootfs is not empty and has the
/dev folder
I see
it's being run as a subprocess of itself
readymade I mean
and the non-interactive subprocess refuses to load the config for whatever reason
also does bluefin-lts use btrfs now? Because the repart configs are all btrfs atmoh we can use it :)
hyperscale kernel has support so now it just works
oh ok
also another thing is secure boot with the installer ISO
should just work
yeah we shouldn't be having these files, why does repart do this?
i think its readymade doing it
it shouldn't be
ok so it is
huh
but where
oh I see
the error mado was getting was because /boot wasn't mounted to the rootfs
as in the chroot
nvm scratch that different error
out of curiousity, why can't we just use https://bootc-dev.github.io/bootc/man-md/bootc-install-config.html and use
install to-disk avoiding repart and manually mounting/setting up EFI partitions. They seem to also support tpm2-luksye that one should be completely fine for wholedisk configurations
except luks
luks doenst work
how so?
cuz it sets up luks without a password, jsut tpm2
then you cant unlock the volume cuz theres no password
can't you chroot and enroll a password?
or is that impossible during install?
you straight up cant add anything
like a password or anything
is there an issue for this?
GitHub
LUKS volumes need configurable password and/or recovery keys · Iss...
See problems described in #476 and #421 for context. The bootc install to-disk --block-setup tpm2-luks feature needs to provision either a systemd-cryptenroll recovery key or a default/backup passw...
sorry can I get tldr? was eeping
GitHub
readymade/templates/ultramarine.toml at main · FyraLabs/readymade
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
this is the config and then just add copy_mode = "bootc" and bootc_imgref = "containers-storage:ghcr.io/ublue-os/bluefin:41"
@Gerblesh
it didnt work doe
we ran into a few issues for whatever reason
we tried that
both of them under
[install], right?yeah for me that resulted in the uhh
yeah
you got super close tho
like, do i put
wholedisk as the install type?hold
sorry I forgot to push my changes
lmao
like whats the full config that you used, thats pretty much what we need :P
LMFAO oh thats why
the toml config wasn't loading in the subprocess for whatever reason in the live image
idk it could've just been my image
oh I see
that's probably why things were broken
still confused on the bootupd error
there
we can straight up talk to colin about this
yeah I'm still facing that issue
are you on the fedora bootc room on matrix?
both of you
uhhh possibly
no
I forgor
afaik
btw I wrote the code for the uhh
might wanna be there if you guys wanna troubleshoot this
boot mount spec thing
did it work?
but I didn't test so I commented that out for now
that seems like thats what is blowing it up
I eeped
we can test it now :clueless:
try reverting the latest commit and recompile
I am now
i dont think you pushed anything
to that branch
wtf
oh yeah errored
:clueless:
ok it's live now
now its ready
lets made
OH SHIT WHAT
huh i wonder if bootupd is gonna work now
cc @Gerblesh @j0rge
why is it pulling the image?
its using the local image
under
/var/lib/containersbut it's using the docker transport?
im pretty sure its the right imgref
oh so it automatically pulls from localhost if it's containers-storage?
this is straight up the output from bootc ya
@madomado at least this seems to be a different error
no idea why its going on ngl
this seems to be the last step.
the image installed successfully and everything looks great
is it bootable?
not yet, we need the bootupd configs
the image itself is mountable
everything is already amazing we just need bootupd to cooperate
oh!
oh this is absolutely bootable
the bootloader got deployed succesfully,
/boot seems fine, too
the configs are there.
lemme rebootbooted grub, seems to be missing the
grub-static configs as expected
that's the same error
oh same
os error thing\
hm
trying to boot off of the thing to see if it worksis this with --boot-mount-spec or no
no changes to your latest commit
k
its bootable btw
to the initramfs
:clueless:
oh I see
we just need to have the static configs on the base image
do we not have them there?
so… the heck with that error?
i seem to have grub2-static on my sytstem (bluefin LTS)
bootupd is failing to write the grub configs to the rootfs
/usr/lib/bootupd/grub2-static/grub-static-efi.cfgso the resulting system is unbootable
it uses this
grub2-static config that i have no idea what the fuck that is
maybe theres something missing on the ISO or the image itselfdoes this exist on the image?
is that on the livesys?
oh the files are absolutely there actually
on the live image
and on the container
oh so why is it erroring
huh
lets just talk to colin
ill message them on matrix
GitHub
bootupd/src/grubconfigs.rs at 604866ea4f5743c2db5cb4706fa076d6e4c3d...
Bootloader updater. Contribute to coreos/bootupd development by creating an account on GitHub.
actually @Robert you have much better communication skills than i do, would you mind sending colin a report about this? 😭
about the switch root part, does that have anything to do with grub configs?
why do I feel like it's something else
i think so
i probably just did the thing wrong
hm, ok
on my system i have a few configs under boot
lemme just send you guys a tree or something
oh theres a ton of shit
this part is fine on the thing
grub is installed, everything is ok
the
grub2-static config is missing from /boot thowhy would it be code 22
i have this on my centos, bootable, working system
(sec lemme get it)
dunno why fucking "invalid argument"
definitely doesnt look right
since bootc install is running on the host it should use the host's files for this
i think
right?
i think we can give colin an ISO with this readymade version
config seems to exist, its also the right path and stuff
what about target?
target variable?
"vendordir"
it absolutely exists since this block even executes
"installed_efi_vendor" seems to come from somewhere else
it may exist as in that it's a valid string but it could be an invalid path
as in the basedirs don't exist when copying or something
BOOTC_BOOTLOADER_DEBUG="-vvvv"?
no…
bootupctl backend install --write-uuid chungus (invalid argument)is that an error from std::io?
possibly?
wdym?
the invalid argument thing?
it feels like so
i mean,
oh shit i think we can wrap bootupd to get the verbose logs
ill try to wrap the
/usr/bin/bootupctl to get all the args that bootc install supplies it
reinstalling with the verbose shiti think this is readymade related
ISO doesnt have
/boot/efi at all there
and bootupd should do the job for us i thinkuhhhhhh
weird cuz i didnt run into
OH
wrong config!
ah
sorry i forgot i had to put the configs there again
aight looking golden
lets see what it tells us
oh man
i think bootc install is way too up-to-date
my bootupctl on my system has a
backend command
whereas my bootupctl on the VM doesnt have it, instead its just straight up the backend itselfHAH YEAH!
THATS IT!
can someone please run just
bootupctl on ur guys system?
ill report this upstreaminteresting, this is also broken on F42
whats ur bootupctl
plis
reporting this to bootc
bootupctl --version and bootupctl --helpbootupctl 0.2.26
both of those outputs
what.
this failed?
is that on the new thing?
on the f42 ISO?
last time I checked
I'll try pulling the new installer
retrying
just wrote this wrapper
please dear lord work
YO
its the
--auto option
and the backend subcommand
i thinkyeah not particularly sure
just complaining about BIOS boot
sudo bootupctl install --write-uuid --update-firmware --auto --device /dev/vda /mnt
sudo mount /dev/vda3 /mnt && sudo mount /dev/vda2 /mnt/boot
auto breaks it for whatever fucking reason
dunno why it cant find the thing
YO
@Gerblesh
sudo mount /dev/vda1 /mnt/boot/efi
with the
--auto option it absolutely works
ill reboot nowwdym?
so
bootupctl seems to not be mounting the right thing
thats the ESP partition
/dev/vda2 is /boot by default
but theres no need for that on bootc apparently
shrugso we don't need /boot partition, only EFI?
ITS BOOTABLE
RAAAAAAAAAAAAAAAAAAAAAAAAAH
LES GOOOOOOOOO
is this EFI or BIOS?
UEFI
oh ok
HELL YEAH
I think my VM is BIOS by default
it fucking WORKS
are we not supporting BIOS boot then?
tahts why you need to customize the VM before install
UEFI first, BIOS later
we'll most likely need to support it for local testing
alr hold on I'll add the UEFI firmware
ofc no flatpaks on the target because we didnt add the cool shit yet but thats it
it's a checkbox in virt manager right?
need to re-make the VM
customize vm before install
then change it to UEFI (no secure boot)
i think the boot UUID is wrong
@Kyle Rankin another awesome moment for the "titanoboa adventures" blogpost
just scroll up a bit
ok so, we need the
boot-mount-spec
recompiling with the cool thing
@Gerblesh mado commented it out but like
probably works
@madomado ur thing almost works!
im getting /boot/efi has no uuid
it should work fine after that tho
trying to manually bootc install ithmm…
@madomado i think i figured it out
we just straight up dont need the boot partition
just mount the EFI directly onto
rootfs/boot/efi
and bootupd works
didnt boot tho
huh
uh… what if we mount the efi to /boot
at least im pretty sure that you can just have
/boot inside of the rootfs
so /boot is the efi? maybe?
so just
/boot?probably?
efi is a separate partition
its not mounted on my system at all
/dev/vda1 pretty much
oh shit it absolutely booted
holy shit
ok so
we want
@madomado this should be it!
managed to install it again!
with that format
ill show logs from the installed system
why not /boot ?
boot inside of rootfs
what does
bootc install to-disk do?EFI as a separate partition
same as this
exactly the same
wow this is actually kinda crazy
should we have a separate repart config then or does this work on normal fedora?
no more discussions about go d damn fucking
/boot
this is normal fedora
its bluefin:41
reiterating: this is what we absolutely need
--boot-mount-spec=UUID=(uuid for ESP partition)
with that layout
and it should workesp → /boot
rootfs → / and /boot/efi
?
esp -> /boot/efi
rootfs -> / and /boot
no need to mount /boot at all
just create a efi folder under rootfs/boot/efiok
@Kyle Gospo this fixes the "ehrm just use 5gigs" problem
for what? UKI?
not super sure about UKI
but having multiple ostree deployments, yeah
oh yeah
now we can pin 1 billion deployments :clueless:
"ah but this will break in 10 yea-" lets just close the loop
I think we'll need a separate installation method with different repart templates for this @Cappy Ishihara
should we report this to upstream bootc?
nah it was a local issue
things are still broken on F42
if you copy the bootupctl binary to somewhere else it turns into bootupctl backend
as in there's an invalid argument
itll be broken
unmount the ESP and replicate that layout i told you there
oh
in the ISO?
then bootc install and bootupctl install should work
yee
its hard to replicate but it works and the installation is entirely bootable
I have it in the repart config but not the live environment
might want to try putting that on ur local readymade sources
wdym?
iafaik its using repart templates that are on the repo
oh I changed those
that layout was generated by repart
and I'm still getting error code 22
you also want the commented out thing
.
yeah I uncommented that
you might wanr to manually run that huge bootupctl cimmand
and see what happens
its 5 am now for me ill lseep a bit
gn
night gamesrs
well tgats it btw just try doing it manually
have readymade fail the imstallation
change the layout manually by unmiunting and creating rootfs/boot/efi
I see, why don't I need the boot spec in the manual bootupctl cmd?
please sleep :d
then bootupctl install that
oh shoot
yeah you should sleep
LOL night
answer my question tmro
you need it if you wanna install the thinf w bootc install cuz then it kniws how to mount rhe ESP
go to sleep ffs
sleep
okay
I'm still getting the invalid argument error
Not for luks
If root is encrypted then grub cannot read it's config or find the kernel/initramfs to boot
I am so looking forward to reading it!
it shouldnt work right now then tho
cuz /boot is inside btrfs
no idea what is going on there
Grub can read files from btrfs
It cannot write to Btrfs
interesting
Grub can even read from older zfs pools
peak linux
Lol
I'm sorry but we wanted the date....
Pretty Name is what's used for grub boot entry names.... And it's used in a lot of other places
Could get rid of the FROM stuff and just have latest-XX.XXXXXXXX in parans
all things considered it's nbd
bootc should be getting the rootfs installed just fine, its just bootupd thats tweaking
maybe we dont need to mount the ESP for it?
no idea ngl
Can we summarise what we have in a discussion upstream (bootc/bootupd)? It might be something easy that's just not documented well
we are trying to run
bootc install to-filesystem on a previously formatted filesystem (we are running it on the host)
bootupd seems to not be able to install itself due to some mounts not working somehow
skopeo copies the target image just fine, but bootupd although it installs the grub binaries it fails to successfully install cuz it fails to install the grub2-static configs to somewhere
the fs layout that seems to work for me is / and /boot/efi being mounted in the target folder
when you have a separate /boot it seems to just failAre we still trying to resolve the same error as this:
yup
exactly that one
at that point everything got installed
but it failed to copy the grub2-static configs to
/boot
this is what is failingDo we have the exact "bootc install" command we're using, so I can share it in the issue?
bootc install to-filesystem --source-imgref=containers-storage:ghcr.io/ublue-os/bluefin:41 /tmp/(randomfolder)
this was a lot of fun to figure out 😭And that random folder is the path to the target mount?
/tmp/rootfs for example
if theres a mounted /tmp/rootfs/boot, it fails
but it works if you have a mounted /tmp/rootfs/boot/efi and not the /tmp/rootfs/boot
using --boot-mount-spec doesnt seem to work btw
if we put the UUID for the EFI partition it fails, not sure about the non-efi /boot partition
would actually be much easier to replicate with a systemd-repart config
wow yeah @M2 its straight up mounting btrfs!
grub.cfg is referencing the btrfs rootfsso thta won't work if /boot is inside a luks encrypted partition
ye
gotta know how to use
/boot thereyou would need a seperate partion for /boot
yeah but its not working
--boot-mode-spec?
dunno why tbh
even when you run bootupctl directly it cant get it working
thats the issue actually
"invalid argument" when you have a
/boot partition mounted thereHave we tried looking at how osbuild is using install to-filesystem to see what's happening?
Are you mounting /boot to the right spot?
havent seen it
seems like it?
dunno tbh
it might be trying to find the
/boot on the host instead of the target rootfs there
ill try to iterate over what i can think of w/ boot-mount-specSo mounts in both /boot and /boot/efi doesn't work?
(osbuild install-to-filesystem docs)
checking it again right now
put the
boot-mount-spec as the UUID for /boot
hoping itll work fine
@Robert
shrug
sudo bootc install to-filesystem --boot-mount-spec=UUID=df98c74d-e7f8-44ad-aeb0-eb57928db98b --source-imgref="containers-storage:ghcr.io/ublue-os/bluefin:41" /tmp/.tmpJUZa7c/
(that UUID is /boot on the target)
at least this is a better errorthat's not mounted
Are you running in a UEFI or BIOS VM?
but it's a subvol
UEFI vm
uefi
no idea whats going on ngl
trying other arguments n stuff now
by default osbuild is making a /boot and /boot/efi
ye
how are they installing the bootloader configs 🤔
Would be good if we can keep this discussion updated with new findings, or to correct anything I've detailed wrong
https://github.com/bootc-dev/bootc/discussions/1271
I don't fully understand what's going on, but I tried my best 🙂
doing an amazing job! thank you!
so what I'm seeing inside of this
boot has a mount a possible mount point inside of sysroot
In /sysroot there is /sysroot/ostree and /sysroot/boot
/etc/fstab only has a entry for /boot to mount it ro
very close to getting this working
even if we get bootupctl going like we cant make it so bootc install mounts the ESP thing for us there
https://paste.centos.org/view/3257e91d
strace for this
another thing
there's a race condition where the kernel doesn't immediately update
/dev/ so readymade cannot fetch the uuid of the espLOL wonderful
and it happens randomly
maybe like
have it retry or something
yeah dw I will just
put it in a loop
LOL nice
/efi isnt mounted on the rootfs so it fails
but like, we cant make bootc install do that for usalso idk why they didn't design bootc to run multithreaded
you can always PR it :clueless:
it's stuck here for pretty long and uses like >90% cpu
its using skopeo
I kinda didn't know it's written in rust
skopeo experimental (something)
pretty cool right
nyeshu
its very disk heavy btw
yup I'm still getting invalid arguments with or without uh
boot mount spec thingy
no
/mnt/boot/efi:
https://paste.centos.org/view/3257e91d
w/ /mnt/boot/efi:
https://paste.centos.org/view/1292939a
its trying to find a hardcoded label?boot is labeled boot and root is labeled root
future enhancement
would be nice if we fixed the btrfs sublabels so things like snapper just work ootb
kyle.
stahp.
it's just a label, cleans up fedora's shame
something to think about after we're done
should work fine with snapper.... main issue for snapper is that you would need to make another toplevel subvolume for the root snapshot which shouldn't be necessary with ostree
for your home dir, you would need to destroy and make again on a subvolume
fresh install using osbuild's unattended anaconda installer
it might not even be using booupd for the bootloader install
bootupd*
on the install
like bootc install is doing
anaconda just copies shit from boot and puts it in the target system
might be under "ostreecontainer" or something
https://pykickstart.readthedocs.io/en/latest/kickstart-docs.html#ostreecontainer
doesn't look like anaconda is using bootc
but when making the raw and qcow2 it is using it
there is a verbose flag for osbuild
install-to-filesystem
GitHub
osbuild/stages/org.osbuild.bootc.install-to-filesystem at main · o...
Build-Pipelines for Operating System Artifacts. Contribute to osbuild/osbuild development by creating an account on GitHub.
it mounts /boot, /boot/efi, and / all into place first
It's even using bootupctl for the bootloader install
yeye thats when osbuild is making the ISO
anaconda itself just does the weird anaconda stuff
thats not an iso build
that's making a .raw disk
i think it also usws that on the iso build
aight so that seems exactly ehat we wanna do LOL
yes
whats the command?
oh they dont have --auto
--generic-image : Perform configuration changes suitable for a "generic" disk image. At the moment: - All bootloader types will be installed - Changes to the system firmware will be skippedthey use this flag and then manually run bootupd afterwards it seems to me at least
yeah, they have that flag
also seems that we need a MBR else that breaks
cuz it tries to install grub for BIOS and fails if we dont have tbe MBR thing
ok so
seems that the esp needs to be labeled "EFI-SYSTEM"
or "EFI System Partition"
:clueless:
bru
GitHub
bootupd/src/efi.rs at 604866ea4f5743c2db5cb4706fa076d6e4c3d2ba · c...
Bootloader updater. Contribute to coreos/bootupd development by creating an account on GitHub.
wonderful
i love hardcoded variables
lmao
I mean at least it works
it should definitely be documented though
yup its literally that
lemme try to manually install it to see if it works
we just have to change the repart config to be EFI-SYSTEM
yup.
label shouldnt matter tbh
it shouldn't but it does
oh yea
cursed
it does fuck this
i hate it
dunno how i specify a label on the
systemd-repart config thingLabel=EFI-SYSTEMyeah
from the man page
man repart.d
and we don't need the boot mount spec right?dunno
ill see it
yupyup
the repart configs that readymade uses are under
/usr/share/readymade/repart-configs/wholedisk/*So you've got it all working outside readymade? 👀
yeah
well the bootupctl part it seems
ya
I'm trying to install with the label changed to EFI-SYSTEM
we just need the right partition name now
this is so fucking cursed lol
@tulip how did you fix this invalid argument error?
partition name needs to be EFI-SYSTEM
and try to add that
--boot-mount-spec
as UUID=(/boot uuid from lsblk)it is
you shouldn't need --boot-mount-spec
if you mount boot under sysroot/boot
are any of the live-isos working to generate an iso?
https://github.com/ublue-os/bluefin/actions/workflows/build-iso-test.yml
This is the workflow we're using at the moment. It generates a liveISO with a readymade version without bootc support
GitHub
Build ISOs (Live) · Workflow runs · ublue-os/bluefin
The next generation Linux workstation, designed for reliability, performance, and sustainability. - Build ISOs (Live) · Workflow runs · ublue-os/bluefin
the ostree sysroot or just the chroot?
i just modified the readymade configs to get rid of the
copyfiles and added the label= to the efi configlooks like in the chroot
seems to be installing
still getting invalid argument
without the mount spec
same
trying to check the verbose logs
BOOTC_BOOTLOADER_DEBUG="-vvvv"you have to set that env var in rust
because the readymade installer is ran as a subprocess of itself
when tis running on bootc install it fails
works when im running it manually
sudo bootupctl-broken -vvvv install --write-uuid --update-firmware --auto --device /dev/vda /mnt
same command
would it be easier run bootupd seperately then?
nope cuz bootc install will fail anyways
not if we use --generic-image
it won't run bootupctl and we'll need to run it after or manually install the bootloader
Oh I see
nvm
"all bootloader types will be installed"
so that flag just installs for both mbr and BIOS
do you have to manually mount the partitions as well?
yup
so maybe it's the mounting that's wrong
manually mounted on this one
where did you mount them to?
/mnt?
/mnt, /mnt/boot and /mnt/boot/efiyeah it did something similar with
/tmp/.tmpQew7go
but it also mounted /dev/vda3 to /run/bootc/storage?this should be like at the end i think
on the finalize step
what that's so weird
so why does it break
because the command works when running manually
it installed successfully when mounting and running bootupctl manually
Please can someone update the discussion with what we've changed and what stage we're at
I'm gonna try running bootc install to-filesystem manually and then give an update
ok so manually setting up everything on /mnt and doing bootc install works just fine
My guess is that we're doing something funky with mount permissions or something
yup!!!
no idea why its blowing up specifically on bootc install
i think its something wrong with the mounts
it's not specifically bootc install
it's the mounts
because doing it manually worked
are the mounts wrong when we are running bootc install then?
on readymade
idk
they didn't look to be when running lsblk
left a comment summarizing everything, I think our issue pretty clearly lies within readymade itself and the way it handles mounts though
Thanks!
I'll leave the issue open until we resolve it - Colin may turn up and instantly have an idea about what's wrong
maybe yeah
aight so
for the readymade repart templates we want all the wholedisk ones
but then we remove the "copyFiles" stuff and add a "Label=EFI-SYSTEM" on the 20-efi one
mado added a patch that sets our file location to
/usr/share/readymade/repart-configs/bootcwholedisk
they should all run sequentially with that human ordering from sort -> so, 10->20->blah
manually mounting the partitions seems to work for whatever reason, running it through readymade directly doesnt
not super sure what is going on there
@Gerblesh did you manage to run bootc install to-filesystem directly?Yeah it works all the same
It's sorta hard for me to fully grok the mounting code in bootc_copy (I'm not very familiar with rust) but it looks fine on the surface
im manually mounting once again to see what is going on
I'm curious if we could just like see what the generated layout looks like in comparison to manually mounting or something
And like exit the program there before bootc install
i feel like we might be running bootc on a sandbox or some shit
Are there any special mount options in the code?
also am i dumb?
shit its /var/mnt
You need a trailing /
its var mnt
Oh wait that’s a good idea lol
repart create them in that order, but we mount them in order of "how nested is this mp", so it's / then /boot/efi
also mado do you have any idea why it works manually but not on readymade?
we made a discussion on the bootc-dev stuff
this seems to be the last blocker
no clue…
uh well how did you mount /
sudo mount /dev/vda3 /mnt LOLcan you send me the issue again
getting it
1271?
yuuup
GitHub
install to-filesystem fails at bootupd backend install · bootc-dev...
When trying to install a bootc image using "bootc install to-filesystem", bootupd is failing to install the backend configuration with the following error: | Deploying container image...d...
We changed the EFI partition label to be "EFI-SSSTEM" to help bootupd detect it properly. And now bootc install to-filesystem when run manually works.
interesting
EFI-SYSTEM actually
typo on his part
thats hard coded by bootupd
cuz thats the default name for coreOS' boot partitionsob
painful.
i hate it too
its also hardcoded to
EFI System Partition
specifically like thatGitHub
readymade/templates/bootcwholedisk/20-efi.conf at w/bootc · FyraLa...
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
here maybe we can try adding Label=?
yup it worked when i did that
also we need to remove all the CopyChungus=
forgot the exact key
yeah ok
the
boot.conf one is fine too
need that too
apparently
just fails specifically on readymade…!?
https://github.com/FyraLabs/readymade/blob/w%2Fbootc/templates%2Fbootcwholedisk%2F50-root.conf#L4
could this be an issue…?
huh.
like
when you
sudo mount /dev/vda3 /mnt, sudo mount /dev/vda2 /mnt/boot and sudo mount /dev/vda1 /mnt/boot/efi
it works
i think its not mounting the subvolumes by default, right?
i think its mounting the default subvolume from that btrfs partition which is 0 by default (just the btrfs real root)actually yeah I don't think they do
aaaaaaaaaaaaaa
so repart is mounting it
maybe its fucking it up cuz theres two mounts on the same place?
but we don't mount them for bootc
ooh
our mount logic doesn't understand uhh subvols so
good
If I remember correctly, repart has a flag that outputs an fstab file. Maybe you can use that for a temporary one and use it to mount?
Also there is a flag for mount that lets you read from a fstab file in a custom location, so a dry-run systemd-repart call could be done after bootc install just to generate the temporary fstab
There is so much context in this thread is it hard to follow. So right now on the bootc branch of readymade you are running into issues with bootupctl not correctly installing bootupd?
yeah it'd be nice to get a summary to see where others can help
copy chungus
yup! exactly that
oh actually bootupd isnt installing the grub2-static configs to the target so it isnt bootable
seems to be something wrong with the mounts or something since running it manually works
We need both /boot and /boot/efi to be different partitions if we want to support anything beyond just a basic install
ya
i thought it didnt work before
ah…!?
alright
Yeah on the post it all worked with an extended boot partition when mounting and running bootc install manually
Are there any special mount flags y'all are using or is it basically the same?
no mount flags
Huh
can you guys check if the command is the same
why
try to figure out if you can merge those two
IIRC it's LUKS
im not saying remove luks
sorry
/boot
im saying merge /boot and /boot/efi
you need /boot for luks
probably
you cant place the kernel under the root partition for luks
because you can't put the kernel into encryption and you can't put it into the esp
but why cant you place it in the efi partition?
uh…
actually you have a point
why not
no idea
time to simplify into a single 2gb efi partition
I just went through AW and they seem to always have /boot
Not what bootupd does I don't think
lets just do the default thing
we can change the partition layout later
GitHub
Add an
install command by cgwalters · Pull Request #30 · bootc-...Usage example with /dev/vda being an extra mounted disk:
$ podman run --privileged --pid=host --net=none --security-opt label=type:unconfined_t quay.io/cgwalters/c9s-oscore:latest bootc install /de...
Huh
What if we try SELinux disabled
We need a shirt like "it was DNS" but for SELinux
dont disable selinux because you will blow up the annotations
set it to permissive in the installer if you want
its permissive on the ISO
Sorry setenforce 0
Yeah permissive
So wtf is this "invalid argument"?
i had it when i was incorrectly mounting some shit
no idea why, no idea how this broke
So is the mount logic screwed up then?
Also question about LUKS
Don't we need to configure GRUB/fstab to handle luks?
lets think about luks later :P
the initramfs should do its job i think
Oh ok
How was it incorrectly mounted?
i do not know
this thing is so hard to debug
i was trying to just mount it like the
/mnt bla stuff we did
and it failed on bootc install to-filesystem
then it worked??? i need to check it again
its a mess
i dont remember what i didAlr I'm gonna try exiting after the mount and see if there's anything funky
try cancelling it right when it starts bootc install
then try running it by yourself
i cant do it right now cuz im fixing some stuff in
ublue-os/packagesok made it stop and now manually running from the terminal
same
/tmp mounts and shit right?yeah
sudo bootc install to-filesystem /tmp/.tmpPfQly5 --source-imgref="containers-storage:ghcr.io/ublue-os/bluefin:beta"
maybe we are just missing sudo?
idk I think the installer is run as root in the subprocess
it could have weird perms tho idk
that would be funny if that were the only issueit worked??
nope :(
same error?
yup.
try running
bootupctl install that thing
i forgot the commanderror: boot data installation failed: Installing static GRUB configs: Copying static EFI: Invalid argument (os error 22)
ERROR Installing to filesystem: Installing bootloader: Task Running bootupctl to install bootloader failed: ExitStatus(unix_wait_status(256))sudo bootupctl backend install -vvvv --write-uuid --update-firmware --auto --device /dev/vda /tmp/chungus
does this work?
I wish I knew...
what the hell 😭
ok so it worked
huh
so we are missing something on
bootc install
does the generic-image thing work?
i dont think it willidk hold on I'm gonna try again
I manually remounted the EFI partition and renamed
I'll rename it using repart now and try
as in repartition, etc
maybe you can try putting the cool
UUID=chungus --boot-mount-spec optionMountPoint=/boot/efi:umask=0077,shortname=winnt
this is in the efi
some special mount options
are those the correct permissionsmaybe remove them?
no idea
but bootupd manually works
so i guess its not that
drwx------. 3 root root 4096 Dec 31 1969 efi is what I have on my system
the host I meanyeye
so I think those permissions might be correct
that's so weird wtf
so umounting and remounting the partition works
I don't like this
what.
bootc install works when you do that? LMFAOwell the bootupd thing
so if you try to install fresh off of readymade mounting
it shits the bed, and both bootupctl and bootc complain
but if you just unmount and remount the EFI-SYSTEM partition it'll work
idk how or why it's like this
lol
nothing seems different about it
maybe just remove all the options there on the repart thing?
yeah that's what I'm trying
removed the mount options
in repart
iirc it shouldn't make a difference
Ok(sys_mount::Mount::new(source, target)?)
this is the line for mounting
oh it might be weird because /boot/efi might have conflicting mounts or somethingi also thought that
shrug
oh shoot yeah this is weird
.map(|(source, mntpoint)| {
let target = targetroot.join(mntpoint.strip_prefix("/").expect("cannot strip /"));
tracing::debug!(?source, ?target, "mounting");
if !target.exists() {
std::fs::create_dir_all(&target)?;
}
Ok(sysmount::Mount::new(source, target)?)
})
.collect::<Result<Vec<>, color_eyre::Report>>()?;
mind sending a
lsblk --json output from the VM here?
as like a fpaste url or something
there might be something funky therehttps://paste.centos.org/view/77e6f3f8
aight gtg
hm.
oh shit.
maybe its mounting the stuff on
/run/bootc/storage so it cant find the damn /boot and /efi thingies
cuz they arent under /run/bootc/storage
does unmounting everything then adding that --boot-mount-spec thing work?
omg so thats why
its finding the UUID of the rootfs from /tmp/.tmpwhatever
then its mounting it on /run/bootc/storage
it copies everything to /run/bootc/storage/ostree/repo then cant find /boot cuz its not on /run/bootc/storage/boot
it doesnt care about our mounts
that doesnt explain --boot-mount-spec not working thoheres the bootupctl call on bootc btw
lib/src/install.rs#1398
also we can add the users later by just editing the etc on the initial deployment
this is mine:
/ostree/deploy/default/deploy/bda6641326ac5b93c4ba9cf494402210c3228c348ff68389e4728e7937cf39b4.0/etc
has /etc/passwd and stuff thereWhat's the diff?
I believe /run/bootc/storage is just where it's loading the OCI image to during the "fetching layers" stage.
It looks exactly the same as /var/lib/containers/storage and /usr/lib/bootc/storage (on a bootc-booted host)
ok so the lsblk outputs are exactly the same
wtf is going on when I remount then?
Does anyone have a script to quickly get a liveiso up to the point we're stuck at?
uh hold on
You have a bluefin live ISO right?
ok so
This is the config
you'll also need to copy over:
https://github.com/FyraLabs/readymade/tree/w/bootc/templates/bootcwholedisk into /usr/share/readymade/repart-cfgs/bootcwholedisk
then you'll need to clone and build https://github.com/FyraLabs/readymade/pull/50, (
cargo build --release) and putting the binary on the VM
then you go through the install process and wait for the error.
what I've also done is add an in process::exit(1) before executing bootc install in the copy_bootc function and then building to test manually running bootc install to-filesystem but yeah that's the basic rundown. Idk how i'd fit this all into a script tbh
@Robert ^So kinda ignoring Readymade, it's working? I.E. manually applying the repart configs then running install to-filesystem?
ish
manually mounting works but mounting using readymade is broken
but yeah ignoring readymade it's working ish
I'm just confused because after remounting the lsblk is the exact same
except bootupctl runs without errors
Sometimes I really hate devcontainers. :/
Been trying to get gtk4 libraries working for the build, but everything I try fails
Been playing with devcontainers a lot lately.... I really find it frustrating that VSCode is only thing that really works consistently
Like this
Like.. Why???
This might be more of a Rust/pkgconfig/ubuntu rant rather than devcontainers
Oh. It might be because devcontainers promotes an ancient Ubuntu version which doesn't meet the requirements... gtk4.pc is older than 4.10
oh can you swap in a fedora container?
You can, but then you lose all the community "features" like docker-in-docker
oh I meant just to get the thing to work
oh, maybe bumping up to the ubuntu prerelease container?
maybe those deps got revved
I forgot libhelium was a build dep, so needs a Fedora devcontainer 😭
Will continue with this tomorrow.
you win today, linux. See you tomorrow.
for gui apps, just use distrobox, devcontainer is not worth with other than ootb containers
FYI - I've built the Readymade binary from the PR and made it available here so people don't need to build it
https://download.eternal.sturla.tech/readymade
Thanks!
We should probably contribute a dependency list
honestly I'm considering… calling bash and then mount then instead at this point
that could work
worth a shot
super jank but if it works it works 🤷♂️
@madomado did you manage to make it work running
mount within bash?busy
it works!!!!
somehow
lol
idk I guess the mount function we were using from rust wasn't working
WHAT
patch.
NOW!
it's really jank
couldnt care less
jank is the spice of life
hold on lemme fix the thing were we rerun repart for whatever reason after running bootc
idk why
i am currently also doing jank shit
ok whatever I'll send a patch
got rid of all of the boot mount spec
it also doesn't unmount everything at the end
cc @madomado
does it straight up just work?
like
wow
you sure?
yeah basically
bootc install works
what fails then
the installer is still a bit broken because it expects the mounts to be unmounted after copy_bootc but like
can you unmount it?
I think it's like the setup_system or whatever
or does bootc install break cuz of that
oh yeah fs after it's done installing
i mean like, on that comment you added there
I was just lazy and I'm not super familiar with rust
oh no
cc @Robert
because the mount() function returns some form of mount object that it iterates through and unmounts
so we'd just need to run through a list of partitions and unmount manually probably
try fucking hard coding the unmount
LMFAO
the whole fancy map() can probably go away and be replaced with iter
then we just iterate again unmounting
ill test ur patch and try cleaning it up a bit
ok so chatty jeeps says I can just iterate through the fancy map thing with a for loop
i'll try that then
sgtm!
actually unmounts, not sure if it's really "idiomatic" rust but it works compiles
"if it compiles it works" :clueless:
this is the rust way lol
so readymade isnt erroring out at all for you anymore?
trying it now
ooo lets see!
if I had to guess it'll probably error out later but we'll see
i hope not
but at the very least it'll run bootc install all the way through
which produces a bootable system
YES.
if it works the last step is making a postinstall script for copying the flatpaks
and writing user info for kde
oh there's one more thing
on the install
in the GNOME OOBE
the "enable 3rd party repositories" freezes the entire thing
so we should probably find a way to remove that page
ok so it's complaining that the target is busy
umount: /tmp/.tmpnuFcYA/boot: target is busy.mount -R /tmp/.tmpnuFcYA
?
oh I can just umount the entire root?
thanks
ok so now it's just erroring in the
setup_system function
yeah it can't write to the /etc/fstab file because etc doesn't exist@madomado here's the patch to use
mount from shell to make bootc work. Keep in mind I'm not very familiar with rust so the code might not be the greatestunmounts and mounts properly, just errors somewhere in
setup_system because /etc isn't on the chroot (ostree has some weird ostree handling)thanks
I'm getting back very soon that helps
so readymade failed to create /etc/fstab
probably because there was no /etc
is that intended or something
Yeah I think it's just the way etc is handled with ostree
Doesn't bootc install to-filesystem already setup the fstab?
yup it does everything
we dont even need to touch the rootfs after install
(we will cuz of thr flatpaks but nothing on readymade itself)
Ah
Yeah
We will for user creation and probably LUKS depending
I also think the fstab doesn't account for subvolume mounts or anything special
you sure?
it seemed to install fine for me
Yeah because how would they?
it works tho
Well yeah install works but they don't have the proper subvol mounts defined in repart
GitHub
install: Add a generic
install finalize by cgwalters · Pull Requ...Basically I want to get Anaconda to run this, then we can perform arbitrary fixups on whatever it did between the install and reboot without changing Anaconda's code.
Motivated specifically...
We can use the command here rather than relying on ostree internals
Also it seems like modifying fstab is suboptimal because of composefs
nah no worries
fstab is on etc
you can edit it fine on the thing
its on the first ostree deplyment
We can use the install finalize command to edit the fstab rather than relying on ostree internals which could change down the line
I see, it's for changes done on the root based on ostree:
We hope to provide a bootc-supported method to find the deployment in the future.So we don't have an API just yet to handle that
I grabbed all of this stuff and figured it out.... I'm doing it with alpine
Next up is figuring out how to do it with local features instead of hacking together in the containerfile
Dev container Features contribution and discovery
TL;DR Check out the quick start repository to get started on distributing your own Dev Container Features.
Should we do a sync-up and write issues for any remaining bits?
Ok, would tomorrow morning work? Same time as last?
what time was the last one?
@Gerblesh @tulip🌷 could you all get an updated status of where we are at with things in this issue? https://github.com/ublue-os/titanoboa/issues/66
GitHub
Switching to Readymade · Issue #66 · ublue-os/titanoboa
Todo Fyra rm hardcoded branding move Readymade to Terra Split config into packages so you can install readymade-config-ublue or whatever UBlue add Readymade to image config enable GNOME OOBE Relate...
There has been a lot of conversation in here, just want to make sure it gets onto the issue.
I see there are some related issues.
holy shit
holy shit
@j0rge
it just worked on the latest readymade commit
like completely
@Robert @madomado @Gerblesh
like what.
it booted
wait what?
@Kyle Gospo CANCEL THE APOCALYPSE.
yeah it just worked LMFAO
Shit, I already started deleting the project Jorge.
lol
Is there an undo?
can you see if a GTS image can get spit out.
just need a readymade release now i think
shit I was already rebasing everyone onto Alpine
so we get it on the iso easier
You know Arch would be a better choice :clueless:
@Robert mind compiling it and putting the binary on your cool file hosting?
so we can get it there n stuff
man im impressed this just worked
like completely fine
<-- removes bluefin-rebased-to-snaps branch ...
What did you do?!
whatever you did don't fuck with it!
compiled readymade from https://github.com/FyraLabs/readymade/pull/50/
well let's go then!
then added the
bootcwholedisk directory cuz the package doesnt have it
and changed the config to the latest one cuz i have an old ISO
and... thats it
we just need a release and a bump on the packageCould you double check to make sure it has the right upstream location?
Bootc status.
lemme see
if not its a super quick fix
Right!
ofc it does :)
Heck yeah.
we just need it to enforce-container-sigpolicy on the bootc install and we gotta add a
target_imgref thingflatpaks?
secure boot?
not yet because we need a super simple post-install script that ill write right now actually
works
has the shims and stuff
no reason why it wouldnt
Easy.
I've logged out for the day but can put it up tomorrow
sgtm!
By "secure boot" I think Jorge is referencing enrolling the MOK keys
Tulip, here's your motivational soundtrack: https://www.youtube.com/watch?v=bZ8LevUbU9M
That can be a post script.
yeah
Post steps work
Hopefully.
can also be a post install sciprt
let's gooooooooooooooooooooooooooooooooooooooooooooooooooo
LOL nice
to think I was just going to shoot Noel in halo all night
You could try!
i can quickly put readymade on the staging copr just for testing this specific branch
with the configs
yes please
then we rebuild the ISOs and done
let's rev this bad boy
what can I do to help, how's the action look in the titanoboa repo?
everything is perfect there
Is LUKS still broken? Or was that sorted too?
havent tried it
it didnt even show up for me
BIOS support isnt there either btw
oh we don't support bios anyway
should work with like a single repart config or something
it'll be a nice to have because of local testing VMs
banger soundtrack btw
fair
Thank the Lord
Don't fix this
This is a threat
LMFAO ok
we can unfix it for ummmm
titanoboa directly
so if you are on BIOS it wont work there
LOL
I mean.
If it already works can we keep it?
eh
whatever
we can have like an option to support it or not
whatever tbh i dont care at all
making the staging package
I'm lazy with creating VMs using kvm lol
The latest readymade commit creates etc on the disk root rather than modifying the actual deployment
eh at least it works!
this is workable
Also the fstab generation should probably be skipped for bootc
Idk I guess we should probably just skip the "setup system" function for now and come back to it for KDE installs etc
I'm just a little apprehensive about releasing ISOs because a lot of behavior is a bit buggy/broken atm so we'd just get known issues reported
well I don't have ISOs so as long as next, next, next, blast disk works I'm good.
works for sure
that part is fine
I just need the mvp, all the other stuff take your time
like if the thing is just a github action and an RPM we could just have it automatically push an ISO to a test bucket every night to test
oh the gh action is fine already
its there we are golden there
ack
we just need an up to date RPM/readymade binary
and done!
@madomado @Cappy Ishihara @jade
dunno which one of you guys i can ping, but, is there any chance we could get the
bootc-supporting branch into the terra package? its still a tiny bit undercooked but it would be just amazing to have an entirely closed loop for this
i can try to package this into like a lil copr or something but readymade depends on libhelium and i think copr will crap itself if i try to do it
the latest commit works perfectly (with a few caveats), but it needs the configs and stuff
with that one we can already like publish "super alpha early ISOs" or somethingI’m down for this if mado approves
After dealing with ISOs today I think Kyle and I fused into Gogeta. Let's do this, let's just make a kickass bootc installer.
@Kyle Gospo you in?
Touring a robot factory
I'll be available in an hour
Come on dude is that what you spend your bazzite money on?
Ya
correct way to spend that money :letsgo:
I was promised halo.
All I got was a replacement to anaconda.
I tried to get your innocence back, but alas, I was too late.
is it planned for people to add, say, custom steps to the install process via a conf file
that would actually be kinda nice for certain people
That is so awesome
yeah so the proper path for etc should be something like
<DISK_ROOT>/ostree/deploy/default/deploy/<DEPLOY_HASH>.<INDEX>/etcyeeee
GitHub
/etc handling with bootc installer · Issue #53 · FyraLabs/ready...Currently in the installer it creates etc in the disk root, however with ostree we should be modifying the deployment directly and calling bootc install finalize (after changes). Currently with the...
should be heading to bed now but yeah
so bootc creates the fstab and crypttab stuff?
will fix
and will release very soon
afaik yeah it does it
for us
not super sure
Yeah it creates fstab and potentially cryptab though not too sure on that. I know you can provide kernel args from the bootc install cmd though. I think we will probably need to implement our own LUKS/encryption handling rather than using bootc because the config it has is rather limited
As in it only really supports tpm2 decryption
Idk we could try to upstream some more config options
Password wouldn't be too difficult, there seems to also be some demand for configuring PCR levels
i'm gonna spend my ultramarine money on more chromebooks
all that money from a certain fox pfp
want to also include your configs?
lets not do that
for like nothing
upstream configuration is always a pain
alright
bumped
approved rather
what's needed for disk encryption to work?
i am not sure what would be necessary for bootc in specific
im guessing it would be just the rootfs mounted with luks, and bootc should figure out the rest
the initramfs should do its job too
GitHub
bootc install to-filesystem - having issues with passphrase LUKS ·...
Hello, I'm trying to install bootc on top of an existing system that's configured with a LUKS passphrase. Reproducing this is easy, just a standard CentOS Stream 9 install - check the box f...
Does the bootc TPM LUKS work at all or is it completely borked?
you could test by adding the uhh
adding "custom" as an extra allowed installation method
no not at all
TPM LUKS straight up doesnt let you add a password to the thing
So like TPM can't even unlock it?
even if that works, how are you gonna unlock ur thing without a tpm2 device?
If I were to install it with swtpm or just on hardware with tpm
like if you change motherboards or something
we cant use it
I'm just mainly curious about how it works with decrypting and all
For mounting the root
i mean, how it works on any other distro?
initramfs unlocks the rootfs, mounts it as
/sysroot and pivot_roots to it
dracut handles all that
the mounting itself would be done by the ostree dracut module
initramfs would somehow detect what is the rootfs, probably via rd.root=UUID=luks-uuid-here
we'd just set it up with a password and tpm2GitHub
readymade/src/backend/repart_output.rs at main · FyraLabs/readymade
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
rd.luks.name={uuid}={label} rd.luks.options=tpm2-device=auto
yup there you go
same thing with bootc
that requires proper setup in crypttab so we'll see if bootc knows what to do
we can set it up just fine
/etc -> rootfs/ostree/deploy/default/deploy/365bee4bc879df58c956b291679bea2b1ec2910e5d705994b4dc0a0792a1aae2.0/etc
(unique hash per deployment)
oh does that get included in the initramfs?We could just probably walk through and pick out the only dir under the deployment but yeah
we arent changing the initramfs on the deployment at all
ya
Is that generated with the container?
yup
no need to touch any of that shit
Oh...
probably not
Can we pass it via kargs?
how does it read the crypttab tho?
seems like we need to just add the kargs
pretty sure
!?
how the heck does it read the crypttab you're right
bootc install to-filesystem --kargsKargs are really easy because with bootc you can just add them as a CLI option
ya i think it doesnt read it :clueless:
dracut?
dracut
i think it just finds it through the kernel cmdline
after
pivot_root
fstab gets mounted by a systemd unit at super early bootSkibidi ostree :glorp:
skibidi toile
you probably need to regen ramfs when fstab changes right?
nope
no…?
Are we able to test without crypttab?
a unit on
local-fs.target parses /etc/fstab and mounts it all for youhow does it know which part is the root
initramfs knows the root via the kernel command line
rd.luks.chungus
i mean how is it even reading it
well on a normal system without encryption you don't even specify the root part
you do
here is my command line
/proc/cmdline
root=UUID=cd533c89-676a-43b3-8278-e63b95c25301
non-FDEah
ostree= for the ostree deployment on the ostree moduleOh nice we can just take the repart opts, and pass them through to bootc
so it's done by grub-mkconfig
nasty
hate grub
now the question becomes does it handle crypttab as well
post-initramfs i think
together with fstab
theres no crypttab on the initramfs i think
i tested a bunch of shit with dracut when i was prototyping titanoboa
and it was so fucking hard to add a fstab to the initramfs
like it was super explicit
so i think crypttab isnt on the initramfs
oh shit nice
At least the cmdline
also we need to
--karg=quiet --karg=rgbh
cuz plymouth doesnt try to load without that
im getting tuxes on my bootAre you trying it?
i installed my current system with
bootc install to-disk from that live ISO
(bluefin LTS based live ISO)Oh I see
Oh yeah we should probably also enroll the secure boot keys, maybe a postinstall script or something?
yup nothing readymade related
no need to touch it
so, we want FDE right
FDE and possibly BIOS boot
we just need to mount it right, same thing as before n stuff, add the password prompt, and add
--karg=chungus to bootc installOh is the password prompt not there yet?
BIOS boot should be just adding an MBR header for it
doesnt show up on wholedisk
if you only have wholedisk
Oh OK so that's just a small 1mb header in the wholedisk for it?
ya thats it
bootupctl should install the thing just fine with that
i dont want to support BIOS tho
being completely honest with you
Fair
Just thinking because anaconda and friends support it but ig we don't really have a priority of legacy support anyhow
kill anaconda and all the nastiness that came with it
we can and should do it
i hate seeing how many threads are on #🛟bazzite-help about anaconda
c10s literally does not support older CPUs for example, and i couldnt be happier about it
i know no one with a haswell CPU is reporting bugs
access denied :nami
where?
oh you need to add urself to all the things
<id:customize>
a
I see
also thank you mado
so much
for working on this
mew
Testing with kargs
whelp that was easy
just worked
LMFAO nice
what did you do?
cc @madomado
SKIBIDI123 LES GOOOOO
had to add the encryption key manually since there isn't a prompt yet but it works
@Niklas ⚡
yeah no crypttab lol
hell fucking yeah
idk if tpm works but password unlock works
shoukd work fine, initramfs does it for you
systemd is super intelligent about it
:pikachuHello:
skibidi toile
skibidi skibidi toile
also we still need to do the whole MOK enroll thing but I think we're almost there
will be handled by post scripts
flatpaks too
dont need to think about it
betbet
And user creation 😭
also we need that cool karg
also on scripts
should be easy
if you say so
i believe it
we do need a bit of data directly from readymade for that
cc @madomado that cool json thjng as a env variable would be amazing for the user creatio
so basically just the prompt for encryption and users
pogus
big pog
anyhow I'm actually going to head to bed now, my intrusive thoughts won and I got up just to add the kargs 😭
yeah good idea
skibidi toile
@tulip🌷 you too
go go
:nami skibidi
is it not implemented?
I thought I did that
didnt see it yet
probably is
will look into that
ill read the code later
2h later will be back
funny uni
you dint need to implement the user creation wiring btw
just the data
YOOOOOOOOOOOOOOOOOOOO
READYMADE IS WORKING OOTB ON THE NEWEST ISOS
@j0rge @Kyle Gospo @Robert @Zeglius @Niklas ⚡ @Kyle Rankin
https://download.projectbluefin.io/bluefin-live-stable.iso https://download.projectbluefin.io/bluefin-live-beta.iso
https://download.projectbluefin.io/bluefin-live-stable.iso https://download.projectbluefin.io/bluefin-live-beta.iso
:letsgo: :letsgo: :letsgo: :letsgo: :letsgo: :letsgo: :letsgo: :letsgo: :letsgo:
Does it use systemd-repart for dual boot?
yup
still needs to be more fleshed out i think tho
havent tried it much
what the sigma where da passphrase
ill check if i can make it work
Empty password by default? I've done that for automation scripts
im guessing it should have a passphrase there by default
like a little prompt or something
cant seem to trigger the passphrase UI
Yeah the passphrase UI doesn't work
Idk if it's implemented yet
We could add a repart template but I think we should just worry about wholedisk for now
Going to sound like a dickhead, but not even gnome can rival that padding 😂
looks awesome tho
big fan of the padding
a bit less then i approve
this is absolutely fire. based. whatever you kids say. awesome!!!
Real. Based af. Very fire.
LMFAO YEAH!!!! BASED!!!!
SIGMA!!!
i knew this would be coming
also the install fails if you are on BIOS
plis use UEFI
😭
can we change it to say "based system"?
love the installer. Looks great, plenty of room to grow.
its so refreshing to see something nice like this
am I supposed to enable third party repositories?
nope
shoot
nope, where did you see this?
is that on GNOME initial setup?
yes
i mean it does not break anything i think but the button shouldnt be there iirc
i think the bluefin package didnt get bumped yet then
it shouldmt affect anything
i said yes, then the whole thing hung.
LMFAO oh no
I have notes/issues, should I start a thread?
yup!
first boot thread
maybe you can write them on the issues on readymade?
gotta do it on #💾ublue-dev
trying to get flatpak working
tiffin seems to be mounting everything, seems completely fine
but it doesnt run the scripts for whatever reason?
this seems to not be running at all
will fix rdm tmr sorry guys uh
took too much time doing uni math
no worries
I have exams tmr so yeah
good luck with ur uni stuffs
will there be a way to add custom install cmds
don't worry I have no idea what I'm doing
im trying to fix it myself
oof
noice
then put scripts on
/usr/share/readymade/postinstall.d
itll run on the installed system, with host mounted on /run/hostI would prefer to not setup loop devices during first boot :p
with chroot or smthn?
chroot w/ sandboxing
GitHub
GitHub - FyraLabs/tiffin: Rust Library for Linux chroot jails
Rust Library for Linux chroot jails. Contribute to FyraLabs/tiffin development by creating an account on GitHub.
ok cool
was making sure since the post install stuff I need to do involves setting up loop crap
OH @madomado i think i know what makes it blow up.
if you know
dive or arch-chroot, that's basically thatthe installed system does not have
/usr or anything
LOLalso uh what happened to the pr
ah…!?
how
eh that one wasnt necessary
is that some ostree
thing
initramfs sets up ostree for you
ostree mounts /usr and shit
ah…
uh
theres literally like no runnable binaries there LMFAO damn
well i guess ummm
maybe we just mount the thing and run the postinstalls on the host?
I wonder if glibc even exists
on a really weird path under ostree, yeah
you kinda can't run the postinstalls on the host
we just need the script postinstall module
just that one
and if we need to put stuff on the installed system we can do smthn like cp h.txt $sysroot/etc/ maybe? or full path
we wont even touch any of the other ones LOL
unless we do some weird stuff, one of
1. don't make it a postinstall module (put it in backend/install.rs directly or something)
2. well, chroot into /run/host
what i can think is like, maybe we should just mount the host's /usr to the chroot
and
/etc
both as readonlyah
then delete every mount after postinstall
we could bind-mount them from /run/host/usr
nice. yeah!
will do tmr
ill try to do it rn
that'd be great w
just like, does this make sense with tiffin?
huh i think ill just straight up mount /usr instead of bind-mounting /run/host/usr
is there not a
.bind_mount thing or something
but I think that'll work either wayeh lets see
Between tiffin and readymade it seems very indian
indian?
Tiffin
Tiffin is a South Asian English word for a type of meal. It refers to a light breakfast or a light tea-time meal at about 3 p.m., consisting of typical tea-time foods. In certain parts of India, it can also refer to the midday luncheon or, in some regions of the Indian subcontinent, a between-meal snack. When used in place of the word "lunch", ...
tiffin is thai or something?
iirc
Readymade is often refering to readymade clothing. And in south Asia especially you have to specify you are lookinng for readymade clothing since most buy tailormade/stitched clothing
nvm it's not thai
https://discord.com/channels/957104883490775061/957104884799393834/1161257562860179486
I misremembered
readymade a song name
many RDMS components are named after songs
GitHub
GitHub - FyraLabs/readymade: Install ready-made distribution images!
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
As the convention of making up codenames for system components after J-Pop references, we have decided to name the installer after Ado's single, Readymade, which happens to have a cool meaning to it as this installer essentially installs ready-made squashfs images. The lyrics themselves could be seen as an insult to people overcomplicating things for themselves (e.g. Arch installation).
Ado
YouTube
【Ado】レディメイド
大人にだけはなりたくなんかない
ダウンロード/ストリーミング配信はこちら：https://umj.lnk.to/ado_rmID
↓レディメイド Inst音源↓
https://piapro.jp/t/pIC2
Vo：Ado
https://twitter.com/ado1024imokenp
Song & Lyrics & Arrangement：すりぃ様
https://twitter.com/iii0303_8
Bass：malo様
https://tw...
this song
and as you may have guessed it's named by cappy
because well who's the ado fan here
I wonder if the ostree CLI has a tool for this
probably :P
we can probably figure it out via bash tho
It would be cool to try to mount it all in an OCI container or something
I think that might work
So like run the rootful container and mount the etc and usr folders
ok managed to get flatpaks working
:p
gonna PR soon, just making sure it works when you run
bootc install tooNICE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
LES GOU!!!!!
then we just need FDE and DONE
this is actually so exciting i love this 😭
I'm so ready to add more requirements 😉
step by step else we wont get anywhere
@tulip🌷 is there plumbing for a prompt yet?
I'm not super familiar with rust/adwaita but I might be able to try
prompt?
like for the FDE password
oh for the FDE prompt?
yeah
the plumbing is there
but i dont see wheres the UI
I mean for the UI yeah
like all the FDE code is there
i dont know wheres the prompt
yeah ik
did they merge the FDE kargs?
or do we not need those?
not yet, we need them
did you PR them?
might want that
nope
I send a patch
I'll make a PR
for FDE in general
ill help you once this flatpak PR is done
k
you able to VC for it?
yup
wanna vc?
sure
"dlts connecting"
weird
discord is weird with this stuff
which browser are you on
using vesktop
i'll use firefox instead
honestly vesktop is useless now
the regular client works on wayland w/ audio sharing now
ur muted
GitHub
feat: plumbing for FDE in bootc by gerblesh · Pull Request #56 · ...
Currently this PR includes plumbing for FDE in bootc, I'm planning on adding an actual prompt later down the line
YEAH!
FLATPAKS ARE WORKING!!!!!!!!
yesssssssssssssssssssssssssssssssssssssssssssssssssss
@Gerblesh
wow the bluefin run of the live iso was 21m
oh wow what happened?
that's already 10m faster than the old one
I mean the one you ran last
I assume I need to wait for these to get merged? Can you explain the loop to me?
yup we are working on them
those would be merged, then we'd made a release, and bump the package on terra
ok so when those merge, we wait for a release or are we slurping in a test artifact?
this is the flatpak install script btw
nope we are done
gotcha
whenever its bumped on terra we can refresh the ISOs
gotcha
:) hell yeah!!!
=>
_
@Gerblesh
is this one new?
nope been there for a while
beta and gtses are there
ok so I tried this one already
https://github.com/ublue-os/packages/pull/360
@tulip🌷 they have the UI for the prompt
it just doesn't pop up for whatever reason
YO?
yeah
ok so thats a BIG improvement
idk why
wheres the code for it?
src/pages/installationtype.rs
brew isn't installed on the live image with the betait shouldnt be
its about 500MBs of RAM usage
oh I see
you can install it by just running the setup unit tho
systemctl start brew-setup.serviceand we removed the "3rd party repos" button?
https://github.com/FyraLabs/readymade/pull/56
a little confused on how the prompt is supposed to work, trying to follow the docs here: https://relm4.org/book/stable/components.html#the-model-2
GitHub
feat: plumbing for FDE in bootc by gerblesh · Pull Request #56 · ...
Currently this PR includes plumbing for FDE in bootc, I'm planning on adding an actual prompt later down the line
that's a clever idea!
I love how readymade opened a browser window to their docs
that's awesome
cc @madomado
only person that knows about it here tbh
ill try to help you out soon, im just cleaning up a few thingies on the flatpak PR
yup! lets make the live ISO like work but it should be super stripped down
the RAM usage is about 2gigs on my VM
bit less
can we drop a bunch of flatpaks but only on the livecd?
its just amazing
like, it's going to copy all of them during the install
but we don't need like a bunch of stuff on the live install
well you need to add custom to the list for now
I'll fix that
oh we can hide them
oh I meant to save resources
naaah like
if we want to ship those flatpaks having them be avaibable there for launching wont make them take up more space/resources
oh if that doesn't then let's leave it, it would be weird if the installed one was different, the main point is to see the entire product before committing
its 99.9% free :)
ok so just no homebrew, EZ no problem
yup!
we can strip down the ISO as much as we'd like in the future
I love the way it's gonna be tho
the live cd boot is so faaaaaaaaaaaast
it can and will be a lot faster
and it's like all of bluefin
we are currently using squashfs because of the amazing compression it gives us but like
we can use erofs in the future once we figure out how to make the ISOs smaller
cool cool
well if you remove anaconda
ok bed for me, if you end up generating a new build to test ping me, good night!
live ISO, no anaconda there, just readymade
all bluefin, no condas
hm
ibus?
its the entirety of bluefin
but it's required nvm
it can be smaller, we'll figure it out
the beta ISOs are just 4.7GB
that's quite large
have you seen the bazzite ISOs
maybe guys let's stop using flatpaks
just ship rpms
smh
stop the flatpaking
nope
its due to inefficiencies with the ISO stuff
almost 10gigs i think
:nami
this shit is small for ublue ISOs
we've never been able to make them smaller
because nothing supports installing straight up the host image instead of another thing
so our ISOs are effectively 2x as big as they should be
ah
so… readymade solves that right?
nope
but it, with titanoboa, fixes literally every single other problem we've had with our ISOs
also @madomado do you know if im
rm -rfing something with this
this looks right but i feel like im missing something LOL
https://github.com/FyraLabs/readymade/pull/55/files
I think…?
this mostly looks good i think
testing it
oh man nice! it worked!
We need some way to have an ISO that is a bootc installation itself so that you can use the host live image to install. I feel like it should be possible I'm not sure what is the main problem to solve. I'm assuming bootupd
its done
OH
yeah i dont know
thats something we should talk about with colin
Like could we put all the user state in squashfs all livecd like and put the composefs in a separate thing that maintains the composefs/ostree image integrity
i feel like we could do that
probably via a custom dracut module
That's v2
I just finished my math final
I have absolutely no idea what happened
Oh yeah adding custom gives the checkbox but it doesn't give the text prompt when hitting next for the password
wtf
Yeah I was kinda confused as well
How we looking
Close ish to encryption support, the password prompt just isn't there
ack
this is me every time i go to my stats exam i get like a 90 on the exam prep then go take the exam and get like a 70
but the prep is weighted higher than the exam so it's okay:?
(my friend typed something very offensive, frick him)
where? didnt see it
shrug
automodded
"this message was blocked by your server/moderators blah blah"
this is done now
can be merged fine
tested it, seems completely good
GitHub
fix: make it so postinstall scripts module works on bootc by tulili...
Bootc needs some more mounting, and this PR makes it so the scripts run via shebangs.
WIP: we need to clean all the mountpoints off of the bootc rootfs after closing the container
Where is titanboa at right now?
I think just FDE
If I go into titanboa right now. If I do just build blah... What do we get?
not sure
Also. Do we use mount devtmpfs or do we rbind /dev and /proc?
clone titanoboa,
just build (blah)
should be it
its not a good experience yet
we were prioritizing just getting the thingie going
all handled by dracut
no ideaWhen you build the iso
There is the chroot environment
That's booted into the live-iso
its all in podman
privileged podman
Okay. So let's see if this can be done in rootless podman then
we dont chroot into the live thingie at all there
i think the only part that you need rootful podman is ummm
the last step
the ISO step
cuz it needs to mount something
Will try to see if that can be done in a userns
Getting into the weeds with Buildah: The buildah unshare command
Sometimes users wonder about the constraints of rootless mode for Buildah and Podman container engines. With rootless mode, we are pushing the boundaries of ...
where are you all currently working on this?
Every branch in titanoboa is stale
Titanoboa doesn't really need any changes.
The majority of this work is being done in the Bluefin repo, or tested locally
just main
just go with main
the other ones are kinda dumb
oh i think m2 wants to make it run rootless
if possible
also not disregarding ur work!! you did a great job with the anaconda branch @Robert
just really hoping we dont need it
GitHub
bluefin/iso_files/configure_iso.sh at 2308109988e3c4aac2aac8a4043fe...
The next generation Linux workstation, designed for reliability, performance, and sustainability. - tulilirockz/bluefin
oh thats my fork
I was figuring the readymade stuff was what @M2 was looking for
oh i think he wanted to work on the titanoboa justfile directly
both.
The last time I used titanoboa it failed to make an ISO
huh.
how did it fail?
does it work with non-LTS now? What was the secret?
like where
nothing it just worked LOL
we were just testing the LTS ones first
or it failed when I tried with Bazzite on day 1
it might work fine now then
we have checks for every PR there
that include bazzite
yeah I'm guessing it will, just wondering if there was an extra improvement that got it to work. I'm guessing it was the squashfs stuff
i dont think so...?
theres a lot of things we changed since day 1
so it probably works fine now with bazzi-
OH I KNOW
its
xorriso, we changed it to iso-level 3
so now we can have the huge live OS image inside of the ISO just fineswitching from podman push to podman pull. Podman push to that specific overlay inside of an unshare requires writing to /run/containers which is a no-go
waiting for flatpaks right now. main one i'm most curious about is xorrisofs
also why are we using fuse-overlayfs and having it be an image in containers-storage? bootc install can use oci-archives
Can squashfs do the same level of deduplication on oci-archives?
I'm not sure if it will, since oci-archive is a tar file
No idea?
xorriso needs to setup a loop device....
so cannot do from a podman unshare
i dont think xorriso needs that
its the ummm
theres a manual mount there
so we can make the efiboot.img image bigger
whatever makes it smaller is what we need! no idea what would be necessary tho
was the easiest way of testing it
we can fix it by just changing it to an oci-archive on titanoboa and readymade's configs
Unfortunately still need to be real root for xorrisofs
@tulip ok can I run a build in the titanoboa repo?
oh i think i already triggered the build
there hasnt been a bump for the readymade release
and theres a bug on the flatpak implementation still :(
its wiping /boot LOL
I thought we were just going to slurp in readymade and merge in what we need?
Is what we need in readymade's main branch?
I can push it to my bucket and we can pull it directly from there? --- Binary pushed to https://download.eternal.sturla.tech/readymade RPM pushed to https://download.eternal.sturla.tech/readymade-git~c0eeb68-1.fc41.x86_64.rpm
I can push it to my bucket and we can pull it directly from there? --- Binary pushed to https://download.eternal.sturla.tech/readymade RPM pushed to https://download.eternal.sturla.tech/readymade-git~c0eeb68-1.fc41.x86_64.rpm
I think we need to just merge the feature PRs and get some builds going.
Maybe we fork readymade into ublue-os and use that as a staging ground for commits we upstream?
Or we setup a COPR directly from the readymade main branch
yeah we should have just set it up that way in hindsight lol
i tried to do that but theres a ton of terra deps on readymade
ugh
why aren't they making a branch or releasing?
its not done yet even if they do it
the current implementation isnt bootable cuz its wiping the wrong thing
im guessing like
we shouls install readymade from terra then use the binary from @Robert
right, I care about the loop not the actual thing
we should be landing a snapshot every day
hmmmmm
yeah
OH
@j0rge lets compile readymade and put it on the bucket.
just do it in the action
snapshot out of the main branch or something
cant, the ISO hook is on the container itself
could we just build it in the packages repo?
oh, we can't just copy the binary in the containerfile?
i think we need libhelium there too
copy thr binary?
@Owen can one of you make a readymade release?
il try to put it on the staging copr
with like a git snapshot or something
right
I don't understand why we're stopped basically
like what's blocking us?
fron getting the latest readymade commit?
a readymade release + a bump on terra pretty mucj
ok can we ask them for that?
i mean, we'd need to do that like all the time if we want git snapshots from terra
right
exactly
are we not doing that?
if we manage to srt it up on the staging copr the loop is faster
can't we pull from the git branch daily?
nope
i dont think so
thats why im thinking we should just put it on the bucket or something
I pushed the RPM to my bucket, so we can use that for now
thinking we should build the rpm and push it to rhe bluefin bucket on-demand
They are building the RPMs in GHA, so I copied it over to R2
any other way is kinda bureaucratic LOL
i wonder if building readymade in the same job as the ISO would work
We can probably copy and paste this workflow, checkout the readymade repo instead and push to R2
https://github.com/FyraLabs/readymade/blob/main/.github/workflows/rpmbuild-debug.yml
GitHub
readymade/.github/workflows/rpmbuild-debug.yml at main · FyraLabs/...
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
what about this tho
then we dont need to put it on the bucket or anything
yeha let's just do that
then let's just run a build and new ISO regularly
we can do it
readymade-nightly can be real
Yeah sure
or yeah, a -nightly would be great, we're so close, we just need to finish it off
Ah or just readymade nightly
I can look at that soon here
https://github.com/terrapkg/packages/pull/4354
Readymade nightly real
Once merged lol
@tulip there we go!
LES GOU!!!!
ok so where are we at now
You said you had one thing left with the flatpaks, and then FDE?
yup!
thats it!
I also have some ideas to make the xorrisofs command run without needing real root.
It's specifically using a loopback mount and for the partitioning. We should be able to do this ourself and maybe we could use udiskctl to setup the loopback mount for us.
Unsure about though about getting the partitions formatted correctly yet
nice! whatever works tbh
xorriso is just so shit
i also think genisofs doesnt work for our things btw
I'm just wanting to be able to make it inside of a rootless podman container
Podman in Docker appears to work. Except the xorrisofs step.
also, does ummm
does just running it through docker work?
havent tried it at all
Yes
Well it should. Since privileged docker is basically root
But docker in docker doesn't seem to work
Probably more just an issue with mknod not propagating....
honestly no idea
that thing is a nightmare
um the flatpaks are workiong good actually
i thought it had deleted the
boot partition but i guess it was just a bug somewhere elseYepp it's exactly that.
The new loop mount doesn't get propegated
looking perfect!
oh yup i also ran into that at some point
getting closer!
now its just FDE
i think
And a script to enrol MOK
someone else will have to figure that out
i have no idea how to do that
but it should work
like just run the MOK stuff
GitHub
bluefin/iso_files/configure_iso.sh at main · ublue-os/bluefin
The next generation Linux workstation, designed for reliability, performance, and sustainability. - ublue-os/bluefin
for anyone brave enough to make the MOK script
don't we have this somewhere else already?
we do, someone just needs to find it
if no one does it ill try to find it
should be on build-container-installer
GitHub
bluefin/system_files/shared/usr/libexec/check-sb-key.sh at main · ...
The next generation Linux workstation, designed for reliability, performance, and sustainability. - ublue-os/bluefin
?
nope, we also need to remove that thing :P
the ublue-motd package has that
found it!
GitHub
build-container-installer/lorax_templates/scripts/post/secureboot_e...
Creates an ISO for installing a container image as an OS - JasonN3/build-container-installer
So I think I partitioning without root working.... It's sketchy
if its understandably sketchy we are good.
the entire ISO building things are really shit
Who likes manually calculating everything and catting together files into one file then labeling the partition table
This is merged now
This made a disk.... And it mounts with udiskctl.
Weirdly the filesystem is owned by root and not the user by default. Fat is owned by the user
After all of those cursed things..... I think podman machine makes far more sense for ISO building
I think I will split things into commands that are you have the tools already inside of a build container vs using rootful containers.
Right now there is a hard dependency for some of the podman commands that docker doesn't implement.
We're not going to do yai right?
I figured it's time to archive the repo?
Nope you can archive it
yup archive it
were we able to get a new ISO spun up?
PRd installing readymade from the nightly package to bluefin.
Tulip - please can you commit your Flatpak script when you have chance. Then the only unknown is getting FDE to work.
Tulip - please can you commit your Flatpak script when you have chance. Then the only unknown is getting FDE to work.
yeah even if fde doesn't work we can at least get some testing in
I've been attempting
maybe we can russle up some help
where are you stuck?
uhh
we're merging in all the other stuff P5 just updated and we're on nightly readymade builds now
I mean I'm super unfamiliar with the UI in general
Does it work when not using the Bootc backend?
it looks like they have a dialog and everything but there's no code to show it
or at least none that I can see
So it's not working for UM either?
it's all in: https://github.com/FyraLabs/readymade/blob/main/src/pages/installationtype.rs
idk I'm guessing not
could we prompt with a terminal?
wdym?
open a terminal to prompt for the password?
oh that would be cursed
So it seems Readymade doesn't use passphrases for LUKS - only keyfiles and/or TPM2.
Does that mean they don't need input boxes?
I don't know how either of those work 😆
oh they have password for LUKS
this is what happens when I show the dialog lol
idk why it's like this 😭
but hey at least it's showing the dialog
@Robert setting the password manually in the code works, we just need to figure out the whole prompt thing
would it work?
it can always be improved later
it would work but it's probably better to just fix up the current dialog
and easier to some extent
we could also make another page instead of the dialog
Yeah, if the backend is wired up already, the frontend shouldn't take long for someone who knows how this frontend framework works
that works for me
whatever closes the loop asap
yeah the backend works, I'm just totally lost in the frontend here
what's the tech stack? when I'm asking around what am I asking for?
I expected to see HTML or javascript 😆
that would be too easy
GitHub
GitHub - Relm4/Relm4: Build truly native applications with ease!
Build truly native applications with ease! Contribute to Relm4/Relm4 development by creating an account on GitHub.
it's an MVC framework
idk I haven't worked in relm or rust enough to really grok it
Oh!
LACT!
that guy was just here!
ok I pinged
https://download.projectbluefin.io/bluefin-live-gts.iso
ok, the installer isn't on the dock, so that's one thing I can fix
Also add it to auto start in /etc
should we do that? I was just gonna leave it on the doc
that might be a good idea
it has a language select and then a screen with "try now" or "install" so I think it is meant to auto start
flaptak script is already on the main bluefin branch
its inline w/ a heredoc on the configure_iso.sh script
it works perfectly, im just figuring out some solution for readymade blowing up the boot partition sometimes
@Gameborn in here, but it's late where you are
we're trying to hook up a dialog so people can enter their FDE password in the installer.
but we can talk about it tomorrow!
@j0rge the ISO failures are just a conflict with the nvidia-open ISOs
all the other ones are fine
where? which failures?
we just need to fix the naming of the artifacts there
something specific not working with relm?
github actions failures
or just looking at options
we just dont know relm enough
oh I mean the installer itself
/boot on the target system is getting uncleanly unmounted so its breaking
i have some ideas on how to fix it
is there an issue?
the installer is working pretty well tho
not yet
ill write it down
is that this?
nope that one is a known issie
BIOS doesnt work
absolutely should be written down tho
If you're using a VM make sure you have UEFI enabled
I specifically asked them to not spend any time fixing this, we don't want to support legacy shit
yeye i knoe abour this one
fair tbh
we probably should make a check or something in the installer at some point
oh ok
lemme set up my thing
ill have a fix for the /boot issue in a bit
right now the installed system is unbootable
And the encryption thing is just a matter of figuring out the prompt, right?
yup, backend is super easy to implement
So that's the two bugs stopping i
an MVP?
yup.
Cool cool
also um custom partitioning doesnt have bootc handling right now
i found that out yesterday
can we cut the button out of the UI?
but that one isnt a blocker for the MVP
not having it seems to hide the FDE checkbox
we should file an issue
writing all of these down
ok so trying again with uefi
just file them in the repo that's what I am doing
if you want it to boot rn i can send you the readymade build that doesnt do the cleanup afterwards
I just turned on uefi in virt-manager and going through the install again
yeye but its not going to be bootable once you install it
is the fix committed somewhere?
i dont have a fix yet, making it right now
kk
the fix is to just use an older commit that doesnt have the cleanup
(rn)
oh my god
i finally fixed it
oh my jesus christ
had to reinstall bluefin over and over and over again to get this working
its a dumb fix but it kinda works
ill test it three more times to make sure this works
GitHub
fix(bootc): decouple mounting/unmounting logic so bootc_copy's `/bo...
There was some race condition that made /boot be deleted all the time, I just entirely decoupled mounting from the bootc thing straight up because it makes it easier to debug and I also made it so ...
man we need someone that knows at least a bit about gtk4 😭
cuz like the FDE issue is just so close to being done i just dont know how to do this
gameborn is here they made lact with the same toolkit
they're just asleep rn time zones
but surely we can figure out a text dialog box np
i think hes using gtk3 + rust..................? no idea
yup this is like the last thing
There's a libadwaita version of lact so they probs know gtk4
oh true?
man why did they not use ummm
that
on the flatpak?
that seems cool
Desktop agnostic to use gtk
I read it on a discussion or faq on the lact repo I think
we do support bios because i made cappy do it
because we wanna keep supporting bios as long as we realistically can
I see they're using custom macros
does the LUKs password prompt work on ultramarine?
Iirc yes
I tested FDE a bit ago and it worked
interesting yeah it seems to be broken on fedora bootc, no idea why
the prompt I mean, bootc requires some kargs but those were easy
@tulip turns out I am the one who caused the regression on the dialogue not showing up
@tulip issue, the pr I wasn't like
idrk why I didn't notice that
but you were using
.spawn() to spawn the processes
that will cause racing conditionsyeah LMFAO managed to figure it out eventually
theres a new PR fixing that
LMFAO OH NO
i do wanna see how you'll fix it tho, very curious :)
I think I pushed the stuff already
oh LOL ok then
i did change some thingiws
ill resolve the conflicts and ping you again
ok?
nvm I only pushed the fix for the dialogue aaa
owk
GitHub
fix(bootc): decouple mounting/unmounting logic so bootc_copy's `/bo...
There was some race condition that made /boot be deleted all the time, I just entirely decoupled mounting from the bootc thing straight up because it makes it easier to debug and I also made it so ...
there arent any conflicts
should be completely fine to merge
@Gerblesh @j0rge @madomado HOLKY SHIT FDE WORKS
LET'S GO
LET'S GOOOOOOOOOOOOOOOOOOOOOO
filing an issue on the arwork
can you screenshot me the progress screen next time it comes up?
this also means its completely suited for bluefin now
flatpaks and FDE work
yeah that's why I'm like, artwork time lol
HELL YEAH.
I have may's wallpaper for you this evening since I will be gone
YOOO
ill PR it right before may then!!!
is it that super secret one?
yeah
very nice.
that one looks amazing
is there a new ISO? I saw the thing merged
not yet
wait cuz like
we need readymade to be bumped too on the nightly thing
GitHub
feat: plumbing for FDE in bootc by gerblesh · Pull Request #56 · ...
Currently this PR includes plumbing for FDE in bootc, I'm planning on adding an actual prompt later down the line
needs to be merged
yeah, or figure out when the nightly fires off
this is actually super close to being usable by bazzite too
the "custom" install type is a bit broken cuz it doesnt handle bootc yet but like, we are close
alright let's finish this thing, we're almost out of time
yeah!
ok heres the progress thing
@j0rge
got it
I just took my own lol
self own
oh nice LOL
where do I file the issue is this in bluefin or in titanoboa
readymade itself i think
actually both
cuz readymade needs support
and we need to put it on bluefin
they're not going to put bluefin artwork in readymade
can't we just replace the images?
dunno
no idea how those work
titanoboa then
oh i think i found it :)
its just a css thing
style.cssNICE!
super easy to replace
LES GOOOO
so we can do it without them?
hopefully
gresource at some place in the fs
https://github.com/ublue-os/titanoboa/issues/66#issuecomment-2817171623
how's this for the art idea?
NICE
this is kinda like what we did on the anaconda theming
yeah, so like, the summer wallpaper sections fit into those three boxes
honestly would be super cool
omg mado help us
how do we do branding
oh speak of the devil, click a button lol
tryting to figure out the ummmm gresource stuff
the answer is: we're still figuring out
LMFAO ok
unfortunately they are bundled in the binary
ooh.....
during compile time
like isnt there any way we can load it straight from the fs?
the installation progress ui is done by lea and I don't really know how it works
we can run with a branding fork tho?
Can the links and text be adjusted or is Ultramarine hardcoded in there?
that's compile-time resource
i think the links can yeah
this is awesome btw
oh ok so it's not a hard blocker to a working thing
yesssssssssssssssss
oooooh.
is this plymouth?
yup
bluefin, with flatpaks and FDE installed via readymade
ok so what do we need, gerblesh's thing merged and then ajusting the text, links?
yup
or is there more wiring left to do?
i dont think so
thats all we needed to do
thats what we need for bluefin at least
the labels and stuff are hard to not hardcode because translations
but I can unhardcode the urls
for the images, need to ask lea
kargs for nvidia maybe, but that we can work out on actual bluefin
no problem we don't need the images as much as we need the merge and release because that means we can do ISOs asap
is that possible mado?
it 100% worked
doing that
lovely!
tulip then later in the css we can make the buttons match the gnome slate color we default to, etc.
ngl the summer wallpaper's little sections fit those three boxes perfectly hahahah
i think that is a libhelium thing, taidan (their OOBE thing) already has it
kk
does it boot normally after you type in the passwd
i.e. does it boot into the desktop?
just to confirm
yup it works perfectly
completely perfectly this is amazing
GitHub
Nvidia support on Bluefin ISOs · Issue #70 · ublue-os/titanoboa
This is more of a tracking issue than anything: We need customizable kargs on Readymade or a kargs.d entry on the fedora Bluefin release like we do on Bluefin LTS Titanoboa itself should have a cus...
nvidia uspport
@j0rge ping just to make you aware of this
bumped to 0.12.1
yeah there's a kargs issue open in OG bluefin
it's straightforward let's just do that we need that to match LTS
i think ill fix them then
should be super quick
https://github.com/FyraLabs/readymade/issues/53
is this working now?
GitHub
/etc handling with bootc installer · Issue #53 · FyraLabs/ready...Currently in the installer it creates etc in the disk root, however with ostree we should be modifying the deployment directly and calling bootc install finalize (after changes). Currently with the...
helping zeglius with a quick pr
nope not yet, this is probably not a thing readymade needs to implement directly
k
ok ill be working on nvidia support right now
titanoboa first
does anyone know anyone with nvidia hardware?
we'd need someone with that for testing this
we can do a call for testing
hell yeha
like, we have the loop down right
so we kick off fresh ISOs
announce, mention that we need nvidia testing.
then as you push in bluefin we just respin new ISOs each time
also does that bump the
readymade-nightly package?no problemo
they bump automatically
nice.
sgtm
that will happen 12h later
ok so we move to regular readymade for now then
sounds great tbh, whenever we close that issue ill do that call for testing thing
yeah let's get new ISOs going and see where it stands
one last thing apart from readymade#58, need to show pages/installationtype even when there's only 1 type allowed in configs
yup.
we dont get FDE if theres 1 thing
yeah
this is so exciting
and then we'll just have auto ISO building nightly to a test bucket and then that'll make getting the KDE stuff and all the rest easier because we can just build on the spot.
because that'll be bazzite, aurora, and aurora lts
the irony that we have ublue adopting rdm before us
bluefin only needs to blast the disk, it's scope is very un-linux distro. 😄
we have ublue adopting rdm before gta6
LMFAO based tho
that means we'll be the canaries
we managed to fix it @Gameborn
hell yeah
ah okay
so ya'll dig this relm4 stack?
the kurage stack
my short opinion on relm is that it makes gtk-rs code nice and readable, but it's often annoying to write and kinda hard to figure out at first
that's why I made a wrapper around it again
@tulip ok so am I clicking the button or do we need to switch readymade-nightly back to readymade?
mostly having to do with the fact that you write ui using basically a DSL in a super complex macro where you don't get good code completion
>qt bad
>makes gtk
>gtk bad
>make wrapper around gtk called relm4
>relm4 bad
>make wrapper around relm4 called kurage
that's like 4 levels of wrappers at this point lol
kurage>gtk-rs>gobject bindings>gtk
i think we need to switch it to readymade for the ISOs right now, tomorrow the nightly release should work
based
the code completion part is honestly like you get no choice it's rust-analyzer being crappy
right, ok I'll be drowning in children in about an hour so if you switch it now I can get one round of testing in
then I'll be back in a few hours after that
@tulip I'll PR it
actually I'll revert roberts
you sure?
well if the ui is just normal functions it's typically fine (excluding the fact that functionality is often split across traits that you have to import separately in gtk-rs)
https://github.com/ublue-os/bluefin/pull/2369/files it's just this right?
yup, looks like it!
yup
it's just that until your macro syntax is fully valid it can't really parse it into output that's handleable by rust-analyzer
https://github.com/ublue-os/bluefin/actions/runs/14560195355
ok building and pushing to cloudflare!
fixing titanoboa
oh I wasn't able to find the .desktop file to adjust, filed an issue
I'll file the autostart one too
the readymade thing on the dock?
i think the bluefin package didnt get bumped
oh ok
missing a bump
ill bump it
kk
force-merge + copr build
copr build going
nice.
titanoboa is fixed
oh I thought we were fixing nvidia in bluefin repo
for the kargs
or is it "both"
oh we can drop building beta ISOs, that channel is kaput
its a ton of stuff
all on the tracking issue thing
kk
https://download.projectbluefin.io/bluefin-live-stable.iso
https://download.projectbluefin.io/bluefin-live-gts.iso
snagging gts now
4.8GB too!
NICE!!!!!
this is HUGE ngl
LOL
this is so much smaller than before
yeah but half the size
nononoo
plus we can move a bunch of the flatpaks to post install later
i mean like HUGE in "WOW!!!"
oh hahahaha
yes 😭
i mean, arent ubuntu ISOs almost that damn size?
LOL
when it's less than 5GB we can have cloudflare push to the CDNs more aggressively
NICE
our ISOs are smaller than ubuntu's LTS ISOs
oh, you can turn off beta AND the hwe ISOs
we don't need to make any hwe ISOs
i think those are already turned off there
lemme see
that's our LTS
clean!
we are using erofs compression there
if we switch to squashfs its gonna be super tiny
if thats the live ISO i mean
nice.
this was also before a ton of changes so maybe its just going to be smaller overall really
we can improve on this a lot.
best ISOs on the linux block!
stable's installing now
hey is there a way we can make grub default to the top entry instead of Troubleshooting?
gts installing now
cleared that
no troubleshooting anymore
cuz that this is just useless
error with FDE, the install without FDE is still going
do you have ummm
the hard drive mounted
somehow
sudo umount /dev/vda*
this is a bug in readymade, we should be able to unmount it if repart craps itself like thatsay it's not mounted
not sure what causes this then
probably a race condition somewhere
k I'll report it.
dude 6GB is gigantic
I can't find that acceptable tbh
its teeny tiny compared to the shit we are dealing with right now
…!?
why is your iso gigantic
this is why we need readymade so bad
ah that's the old one?
every single piece of tooling for bootc ISOs is shit right now
yup
ah
k I was like yeah
our initial ISO tooling was duct tape and it broke
what we have here is the most functional ISO we've ever had.
this is why we need this our project is currently dead in the water we're rolling with old ISO snapshots
FANTASTIC
LES GOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
third party repos screen is still there
will be fixed once images are bumped with the new bluefin package
WELCOME TO YUMMY TUNAFISH
ok filed the repart one with readymade and a few in titanoboa
the keyboard input screen is weird for whatever reason
seems to not be set to any keyboard input by default
ok done nvidia-open ISOs should be good to go now
alright, tons of progress, this is amazing!
trying FDE with gts again
i know it works, i showed you a screenshot and stuff
We don't really need them since we set modprobe stuff now. Kargs were needed when we didn't make our own initramfs
no idea whats going on with your thing
it's fine we can keep going
dont we need the
kargs.d entires?Nouveau is probably fine in a live ISO...
i mean, we have the nvidia proprietary drivers there
theres no reason why not to use them on the live ISO
will just be better i guess
bazzite will for sure need that
like I can see nvidia people wanting to test before committing
since that hasn't been an option for them before
true
@M2 we can entirely get rid of this
I think if we can get the autolaunching and the dock icon we should be good for announcing an MVP tonight
now that we are using bootc install for the installation
the "enable third party" thing should go away since we bumped all the bluefin RPMs?
- dock icon/"enable third party" -> bumped bluefin package + image + ISO
- autolaunching should be just a single snippet on the
configure_iso.shok so I can kick off a GTS build to freshen them up, then rerun the ISO builder.
yup!
exactly that
Definitely needs a test but those kargs scripts are from before we made initramfs
yup!
like all this stuff most likely can be removed i think
someone needs to test if nvidia works at all when installed now
nvidia and whatnot you can just keep working on it doesn't block me announcing stuff
Not the framework stuff
https://github.com/ublue-os/titanoboa/issues/70
this is effectively done
unless you get it before I get back lol
yup
yup! should be fine
oh and the grub thing
what's the tldr on that, I filed it
already fixed it
the troubleshooting thing right?
i just removed it
YESSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSs
WE ARE DONE!
(not really)
LES GOOOOOOOOOOOOOOOOOOOOOO
this is beautiful
super happy
ill copy the workflows and make bluefin LTS ISOs too maybe then
DOLLY YOU ARE BEAUTIFUL
amazing?????
Who knows how the SB keys should be added? Should be a post script
OH MY GOOD IT TOOK US FOUR FUCKING YEARS TO GET TO THIS THIS IS SO GOOD.
@madomado you are LEGEND.
you're all LEGENDS.
GitHub
build-container-installer/lorax_templates/scripts/post/secureboot_e...
Creates an ISO for installing a container image as an OS - JasonN3/build-container-installer
we'll just copy this
for secure boot
ping me for new iso size btw
Is it just a copy and paste job?
not exactly, needs just a tiny bit of work there
cuz im not sure wheres the sb key on the ISO
https://github.com/ublue-os/titanoboa/issues/76
found another one
super fast fix
someone file the SB one
yeah that's fine let's just get them filed and going
@tulip why not just checking for an empty string
https://github.com/ublue-os/titanoboa/pull/74
file it first then fix it so we don't forget.
quick and dirty fix
empty string also breaks just
?
cuz we cant have it be like
just build 1 "empty_string_here" argumentOh
if you figure out a fix it would be aweomse but no idea what to do there
Mm you should move these args to be recipe vars instead
i might just change "NONE" to
0
recipe vars?
maybe yeah
no idea how to implement it thoI think I have an example there lemme look
Grub?
Grub?
grub?
Ah yeah the hook_post_rootfs
Look how is being used in the build recipe
So the today I learned.... Podman can be used as userns chroot tool....
--rootfs ?Yepp
amazing right
Docker doesn't have an analog
interesting
I used to use arch-chroot for something very similar using its userns flag
this is SOOOOO Good
@tulip what's the tldr on the partitioning
Titanoboa as written right now is explicitly podman.
From it's push/pull/chroot stuff you need to use podman
works, bootc install doesnt work yet
like is the boot still too small? (for the nvidia people)
should be just a tiny bit of wiring up
no I mean the layout
we can make it bigger
lemme see how big it is
2G
we probably want to make it a bit bigger
wait, ext4 lol
yup,
/boot needs to be readable by gruboh oh
yup its entirely podman right now
we can and should improve this at some point
cuz this is a big nasty
Nasty dream of rewriting this to be a dagger module
mine says 1G
That looks right
ok we'll want to make the thing bigger maybe
seen too many people running into boot partitions being too small
yeah
do it LOL
filed
1G is good for up to 8 or so amd deployments or 4 Nvidia ones
First I want some semblance of rechunker
we can improve on this tho
we arent stuck anymore
to be completely honest i'd put fucking 3G or something as the minimum size
Why?
rechunker?
yeah I want to implement it with dagger
We don't use Btrfs bootable snapshots
Ostree is better for that use case
fixing the signed image one
on readymade
And when people get stuck... It's either they have a stupidly small boot partition or the pinned like 8 deployments
I keep reinstalling
I haven't seen some one get stuck on 1G when using their system normally
this is so good, we finally have it
ok let's leave it at 1G then?
no reason to deviate from fedora for this one?
man i was already PRing this one :clueless:
how it's the last issue filed, go in order. 😄
You can bump it to 1.5G to make Nvidia users less likely to footgun themselves by giving them space for another pinned deployment..
But it's frankly not super necessary.
at least this thing is dynamic
so itll be min 2 or something max 4
shrug i dont have strong feelings on this
@j0rge live nvidia-open ISOs are in the bucket
If we get rid of any auto kargs stuff we'll likely get rid of unintentional stucks
auto kargs?
ublue-system-setup
rpm-ostree kargs calls on that??
Some still exist in bazzite
considering your usecase you should make a new package that has a
%triggerin kernel thingy that strips dracut automaticallyWe have hardware setup scripts that apply kargs.
mm
thats leaving it to be dependant of runtime state
nvidia
.kos are way too bigdracut --force --parallel --regenerate-all --hostonly --strip --aggressive-stripNope. All of the kargs stuff predates us even making initramfs as part of the build
we build it straight in the containerfile so no need for a package for that
then let's just leave it, I thought it was an actual problem
readymade also does that automatically fyi
closing the readymade PR then
we cant use dracut on bootc host
we cant do shit you shouldnt do shit to bootc host there
ah
it's in the dracut postinstall module so actually you probably disabled it already
yup LOL
we are just using the script postinstall module
And this is only possibly an issue for Nvidia users because the drivers being included bloats the initramfs
so what do we do then?
2G by default?
im fine with that
I'm going to rebuild
bluefin:latest so we can keep testing ISOs, I don't want to push gts/stable until we're further along, save people the bandwidth.The only AMD user I've seen that hit this issue custom partitioned to 300 MB....
this should let us test to see if the RPMs got bumped, etc.
the world doesn't explode if we set it to 2GB
the PR is there
k
i can just reopen it if we want to
1G for an Nvidia user is up to 4 deployments. Default need support for 3 for rollback, current, and next.
4 makes it so you can pin 1.
2G is probably fine. But really just might waste 1G of space because some users are "special". On AMD 2G will be space for over 10 pinned deployments and I have a lot of questions there.
we can make it adaptable
the repart files are on the ISO
we can just change it if the ISO is nvidia
done
:)
what no
don't special case it that's complexity for no reason
just leave it, we have more important things to fix
yup
fixing those
ok so is this list of flatpaks the final place it's going to live?
https://github.com/ublue-os/bluefin/blob/main/iso_files/system-flatpaks.txt
we are reading
${GITHUB_WORKSPACE}/(some_folder) to get the list of flatpaks
can be anywhere on the filesystemI'm just wondering why we have like three places to define flatpaks
are we ditching the previous list, is this the new list?
I don't care where it is as long as we only have one.
that one is the new list i think
we just need to fetch it from somewhere on the gh actions
we have three lists cuz then we have the list on gha
you appear to be fetching it since the ISO picks from this list
nono i mean the actual list, like
lemme show you
https://github.com/ublue-os/bluefin/blob/9657f31837b6c83f13035c28ad2a99abe0901633/.github/workflows/build-iso-live.yml#L85
flatpaks-list: ${{ github.workspace }}/iso_files/system-flatpaks.txtright
we can just curl it from some github repo or something
thats not being fetched
so this is the new canonical location of the flatpaks right?
that path is relative to the repo
bluefin/iso_files/my_listthat's not what I am asking
oh i mean
we can move it anywhere
shrug
I don't want to move it
I want to know if this is our canonical list of flatpaks we install on the ISO
which appears to be yes.
oh yeah
yeah yeah
sorry i didint understand
no problemo, I just wanna blow away the old list so we don't confuse ourselves
yup!
fixing the signing thing
I'm going to blow away all the old ISO stuff now
sgtm
Guess I need to update my flatpak lists... Lol
Does secureboot MOK enrollment work?
just needs a postinstall script to be written
havent tested it yet
@M2
@j0rge signing is done
Does readymade have a post install scripts dir?
where's the issue filed at?
NICE!
nowhere, need to file it
ill file it rn on titanoboa
put mvp label on it
yup,
/usr/share/readymade/postinstall.dok the bluefin github workflows dir is way more sane, one iso yaml instead of like 4
yes 😭
if theres a reusable-iso thing remove it too
Then literally copy the script to there.... That's an sh script without the shebang
I got it already
yup!
#!/usr/bin/env bash bam boom
that script is a lorax template thingWe had 4 so we could selectively turn on and off them. Like when GTS broke
yeah
Replace the first 2 lines with a shebang. Make sure it's executable.
And see what that does
its running in a chroot so maybe it will blow up
but its a very lax chroot
Does it mount in /sys/firmware/EFI?
mounts everything
our current set
Then you'll probably be fine
looking crazy good
@madomado iso sizes
that's still gigantic huh
yup
at least its a huge improvement
Are we doing DX ISOs?
i dont think we should
are you using erofs
but we can
squashfs
probably some inefficiency somewhere
so your live system uncompressed is around like 25GB?
about 15GB
The readymade post install happens inside the chroot?
yup
yes
15…!?
its huge
our images are 17GBs uncompressed
?
i mean
thats why i said about
its probably bigger
i dont remember
need to run titanoboa locally again
hm
I think we should. Turning on DX sucks as a UX, it's 2 reboots
OR .... we can put the user in the docker group ahead of time so dx is only one reboot. 😄
bbl
Yeah. Plus it's an additional 4GB or whatever download beforehand
if it's 5GB compressed, then I think it'd be around 25GB uncompressed…
how much flatpaks are you shipping dude…
runtimes are huge
we can make it smaller tho
Do we still have Thunderbird installed by default?
I hardcode the docker group to 252 and the incus groups. We could include that on all of our images to make the change
i think so - doesnt seem like it actually
Let's do it
FYI - don't know if it changes this
I'm switching Docker over to using systemd-sysusers
We removed tbird for space
We should add it back tho
i think this is the best solution for that
I think hardcode makes more sense. Since we want the groups to be static?
Unless systemd-sysusers can dynamically add users to a group
systemd-sysusers is the upstream solution for this stuff
And cuts our iso count in half
Worth it
Upstream won't want to (nor should they) hardcode GIDs
They currently hardcode a lot of GIDs
GitLab
group · f42 · fedora / Fedora Atomic Desktops / ci-test · GitLab
Experimental project. See details in the README. Expect rebase and push --force.
I'm talking about the Docker upstream
The RPM package
I'm saying we hardcode it
GitHub
m2os/build_files/server-packages.sh at 29a6e83cb58d490dd1899b940cc5...
Contribute to m2Giles/m2os development by creating an account on GitHub.
This isn't stopping what I'm doing - I was just saying the Docker RPM will be switching over to sysusers in the near future, to it might change how we implement things here
Oh
If we pre make the group I believe sysusers no-ops
With m you could add all wheel group members to another group... Interesting
No read that wrong.
Does that mean we won't need to deal with groups to enable users to use docker?
@madomado since ur commiting directly to the main tree, i sent you a PR, if you can plis take a look cuz then there wont be as many conflicts whenever ur done doing what ur doin
yeah I'm reviewing
ill fix the ummm
conflict
It means we won't need the hacky "copy from /usr/lib/group to /etc/group" when adding the user to the docker group
Ah good enough
LES GOU
this is also happening by default because rechunk does this
conflict fixed
Rechunk does do this, but it doesn't work on an existing system since /etc/group isn't updated with the image updates
check the suggestion
holy shit i didnt know that existed
the more you know™️
if it's not extend_from_iter then it's extend
it's something on the lines of that I don't remember exactly
nah seems to not work damn
i wonder if just extends works then
it's probably extend() then
where can I get the new installer ISOs?
https://download.projectbluefin.io/bluefin-live-stable.iso
we arent building beta anymore
Looks like lots of good stuff is happening in here.
I'm assuming we can just write a user creation script for folks using KDE (and other DEs)?
@tulip what's the tldr on the erofs compression?
very VERY fast but it increases ISO sizes a lot
entirely implemented, its just a toggle
oh I thought it was the other way around.
ok let's optimize for ISO size then and leave it like it is, was thinking if they got smaller we could add tbird and pinta, but we can save that for new yafti
i feel like the best solution for KDE is using taidan to be entirely honest
while they dont make
gnome-initial-setup kde edition™️
oh yup!
the ISOs can shrink a lot stillwait you just said the opposite tho
i know, but we can shrink them with other things
I'll check into KDE OOBE. Last I heard, they have someone working on it.
the entire process is kinda inefficient still
even w/ squashfs
ok so not a today problem, gotcha
yup!
just don't use flatpaks smh
what else would we use?
rpm-ostree install 200 packages LES GOOO
they are kidding but yeahI think I'm the only flatpak hater in this chat but yeah
what the sigma what do you mean ur a flatpak hater
flatpak the GOAT
in every way
having rpms in immutable is a bit not practical so I guess that's a good tradeoff
I agree that we should remove RPMs from the world.
ok so ... back home now
you'd be surprised how shit rpms are considering what other solutions we have
nah nah nah
naaahh
hell yeah! the pr that adds signing and some other thingis is almost done
ok
so waiting on a merge and release or titanoboa only?
yup, merge + release after the PR is done
then we should have proper nvidia support all over the board
and signing
ok so SB and the autolaunch outta do it?
It looks shitty without the logos but that's not a blocker for today
SB is already done for bluefin in specific, pretty sure we dont need to do anything
oh ok
autolaunch is a single
cp statementalright so I'll just go over each feature
ok let's bust it out then while we wait for a merge?
or we got other things in flight?
i just need to actually fix the thing thats gonna be merged
they really should, is there an issue tracker anywhere for this?
btw the logo is an issue on your images
KDE Linux has an issue for that somewhere
interesting
whats the thing?
we just replace
fedora-logosor actually
the icon should be
fedora-logo-icon unless you modified [distro] icon
in the tomlits fedora-logo-icon yeah on the thing
on the image
at least it should be
a…!?
dw ill evangelize you on the cloud native ways at some point
https://github.com/search?q=org%3Aublue-os%20fedora-logo-icon&type=code
does everyone except bluefin set this lol
interesting
seems like it
ill try to fix this on the ISO directly and see if it works
if so ill just patch our bluefin-logos package
@madomado ur stuff is good to go now
i think
i had to go with a for_each cuz iterating and doing those things was making the borrow checker super mad
GitLab
Use for Out-of-Box Experience (#38) · Issues · Plasma / Welcome C...
plasma-welcome already has a large portion of what we need for a great first-run experience. We need this in place for KDE to be a viable...
looks like they're working on it
ah
cc @Noel
k
thank god they are working on it
tbh I think we can just stick to the anaconda web installer for everything else for now
kill it with fire!!!!
lol good luck with that
yeah it was a pain to get anaconda working at all on the ISO
is it difficult to setup?
no
it just doesn't work
oh...
oh so anaconda just doesn't work on the live ISO?
the reason we're switching is because anaconda is a no op and they're prioritizing package fedora and not bootc
it works but the flatpaks are a problem
oh ok so it would be easier to add a users page to readymade than to use anaconda?
1 billion percent
i do think thats like out of scope for readymade tho
hm
no idea
thats why id like to just use taidan
yes for sure, anaconda isn't going to work for us
that's why we're moving to this
you can't add custom pages; but what you can do is to add a script that launches a dialogue for that (not sure if that works tbh ww…)
so like
kdialogew
very convenient™️
brilliant move
bru
I mean we are gonna run this on KDE so
but yeah the problem is that we don't get network connection, timezone, etc
the meowing
- use taidan for OOBE KDE stuff
- use that kdialog nastiness
- add conditional page for users on readymade
we only need the users page for
bazzite-deck if we go with taidanlet's finish what we have now tho
yupyup
ok so
the PR thing is done
GitHub
feat: add bootc_target_imgref, bootc_sigpolicy and bootc_kargs by t...
This adds support for signed bootc images, custom bootc kargs via the config file, and custom imgrefs for target if necessary
Fixes: ublue-os/titanoboa#76
thius
signing is good
we need autolaunch and...?
SB?
autolaunch + dock
i think also the logo
+ the logo wherever that is
This logo thing is confusing to me
you fixed the logo, lets just bump it all
no idea either
its some icon that is misnamed or something
the test is not passing
true
i dont remember why...
maybe i missed serde(default)?
oh maybe we can make a aurora one even though the username doesn't work we would be able to see if the logo works?
oh thank you! didnt see it
oh yup
ctrl + c, ctrl + v workflows
cc @imbev @ledif @inffy we're closer now, now's the time to start thinking about your github workflow.
@madomado tests passed LES GOOO
god bless they are doing this
last commit being 4 months ago is sad
oh that thing is 1 billion percent on maintenance mode
yeah
rewrite!!!!!
alright, merged!
now we just need one of those RPM things
0.12.2
building new ISOs!
ugh wonderful the thing just canceled itself cuz i ran another iso build
wtf is wrong with this cancel culture
so real.
yeah don't do that
the more you know™️
DISCO
DOCK AND AUTOLAUNCH
@tulip ok I think this is good enough for a forum call for testing
wdyt?
YO THE DOCK TOO???
yep
LES GOOO i was expecting needing to patch the livesys package
sounds pretty damn good to me
oh let me check to see if it stays on the dock
the nvidia ISOs and the custom partitioning probably are kinda finky
funky*
yeah I'll mention that
oh
no, the autolaunch works
when I close it it gets removed
ok the dock icon gets removed then?
yeah I'll update the ticket
ok ill write a patch thing
shouldnt take too long
third party repos page is still there
ugh no idea what to do about this at this point
we had this off before
@tulip ok can we turn off the
-live in the filename? At least for gts is that hard?yup should be a 2 line change
ok we can roll with what we have now
very nice.
Secureboot? Key enrollment?
secureboot should work, havent tested it. havent added key enrollment yet
ill just mark this as a draft
Has anyone tested this on physical hardware yet?
yup
i have
installation works perfectly fine
ill actually test this once more with my laptop
can you add an issue for key enrollment?
@Robert thank you for the rename :)
Universal Blue
Call for Testing: New Bluefin ISOs!
Alright everyone, thanks to the folks at Fyra Labs we are now pretty close to fixing one of the worst parts of experience, installation! Readymade will be the installer that we will be using for the forseeable future since they are working diligently on good bootc support and that’s important to us. We need some other things for better KDE s...
@madomado @Cappy Ishihara why is this the default? 😭
@j0rge
dock is working on latest images
ill PR a change that makes the readymade icon show up.
woo!
burning a stick for some bare metal
@j0rge
found out the logo issues!
we really didnt have the
fedora-logo-icon pixmap at all
any logo but that one pls. 😄
LOL ok
the stock dino is fine
and then for lts we'll use achillobator
🍣
man they are either super small or have bad contrast with this
dang
!? didn't know tbh
theres a
liveinst.desktop symlink that makes it appear two times too
LOLits super small but it looks cute
I'm sure we can grow her later.
LOL yup.
#63: emcc device doesn't have /dev/disk/by-path/, bug in lsblk-rs
#64: rdm bug
#65: that's not how it's supposed to look, will fix
oh interesting
emmc drivers are painful
drives
LMFAO yeah...
can't type a
eMMC sucks
apple
abacaxi
https://pastebin.com/hBKd8PTi @madomado
Pastebin
~ ❯ readymade(process:8884): GLib-CRITICAL **: 23:38:09.813: g_va...
Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.
huh.
makes me so happy
its working amazingly well
you gotta try it on a VM
should pull from accent colour portal
also jade can we use taidan? i think it needs to be on terra :P
wanted to add it to aurora
so we could use readymade there too
yeah as far as i know it needs a bit more work, lots of polish
and we're still playing with compositors
@lea has been heading it up
of course
hell yeaj
TL;DR: so the installer doesn't show in the installed environment (fedoraism), the symlink is also a fedoraism because of the live-scripts package
distro people be like
i just removed both i dont carw
yeah, it's a really silly mechanism, but it's needed for compat with livesys-scripts
I gotta play with these test isos lol
kinda curious how well it works
i just removed all this shit from the ISO, it should just launch the thingie and thats it
livesys has a ton of anaconda stuff in it
yuuup
well, if it works it works
quite well! installed bluefin at least like 20 times testing this thing
oh wow
that's great!
can i have a better screenshot
a good one that will go in the announcement
no.
just kidding ilk get a better one soon
i don't have vm software and have the 42 upgrade about to start
eh okay laptop #3
oh oki thanks tulip
oh oops
GitHub
fix desktop file by nothingneko · Pull Request #66 · FyraLabs/rea...
computer -> device per MOS
yup its fixed now
it was a misconfiguration on our part
we need to freshen up the ISO
lemme take a screenshot of my working one
based PR
i might also make brazilian portuguese translations for it
here you go
@jade
perfect
i can get a bit of the wallpaper in it
mm
thank you
mmm?
i can put another wallpaper in it
that was a statement not a question
i like to surround screenshots with a little bit of wallpaper
more screenshots
this is working so amazing
im so happy
have you tried the installation
its actually super fast
i have on um, not on bluefun
yeye
i will after i finish my marketing assignments
they're due in
uhhh
its great anyways i guess LOL
3 hours
it's easy it's a discussion about omnichannel marketing
this one is crazy easy i mean look at this
i don't even need to talk that much about marketing
tulip
you should tease the new wallpaper
just like how you showed a snippet
like crop it so you don't spoil it but it'll be like OOOOOOOOOOHHHH
basically just change the wallpaper and run it locally and just screenshot it hahah
this is looking so good
ya'll slayed the past few days
like this is the first time I am like "we are now ready to make USB sticks"
midst of drafting again jorge
this was the final piece
this + the new yafti + flatpost
bam then we'll be done
:)
Is it possible to edit the intro text?
you know I gotta be more dramatic than that
intro text on rdm or?
this part
and I think in ultramarine you launch a browser to open the docs.
@tulip omg, for GA we pin the offline docs opening in Papers right on the doc too.
nope, we don't open a browser here
Like ship the offline documentation.
we do tho
on the installation page
Actually, Yelp is a tyre fire we have to get rid of
TRUE
there are thingis that you can click to open a browser
YES
we should just change the yelp icon to "open bluefin.pdf in Papers"
yeah on the last page
just not this page
filing
oh yup
put it as a MVP
ill add the icon and stuff
to make it look fancy
I am making the pdf already
we just gotta copy it
remember those custom ublue icons
for the updates n stuff
that but for the docs
yeah
omg wait
on the live ISO we should just pin the existing doc menu
filing that
we might want to just modify that one to open the PDF
I agree.
i wanted to actually just include the PDF straight into the image
but we can do both independently
right
/usr/share/doc/ublue-os/bluefin.pdfi think thats the right place yeah
it's building in github now we just need bluefin to slurp it
I will file both now
this is gonna be such a cohesive experience this is gonna be amazing 😭
omg the PDF will also open with the fancy GNOME PDF thing
its gonna look awesome
see how much faster we can go with a finished loop?
yep.
it does look awesome try it
GitHub
Update pdf.yml · ublue-os/bluefin-docs@6570ca9
Bluefin Documentation. Contribute to ublue-os/bluefin-docs development by creating an account on GitHub.
yes!!! i've used it a lot
nice.
testing TPM2 unlocking and the MOK enrollment now
already have the mok script here locally
ok TPM2 unlock isnt enableable yet
GitHub
Pin Community and Documentation icons in the live session · Issue ...
The live session could be WAY more useful if the docs and community icons were just on the dock for the live session. Gives people stuff to do while it's installing, and then will be in the men...
this is what we want
oh hell yeah
that's gonna look so good
@j0rge
what if we create a board
for all these ISO related issues
cuz this stuff is hard to track cross repo
and you saw how well the board workflow worked at least for me during the bluefin LTS prototyping :)
k
just prioritize on what you wanna see done first and ill do it
i just love doing anything related to this
on it
pinning this
on chromium
YOYOYOYOOY
@j0rge @M2 WORKED FIRST TRY
LES GOOOO
LETS GO
ok how's that look tulip?
perfect
I arranged the issue in order, just go top to bottom, right to left
ill need help from someone that knows podman better than i do on this one https://github.com/ublue-os/titanoboa/issues/80
GitHub
Install fails if system isn't online · Issue #80 · ublue-os/titan...
Pretty sure this is file in bootc upstream but tracking here. If you're not on wifi bootc fails trying to fetch ghcr.io/ublue-os/blah.
yeah let's start posting the calls for help in the main ublue dev we need everyone
should be just a single line change on titanoboa, then a single line on the readymade config
there's actions work, package work, all sorts of work, not just readymade so we can scale out
yes.
and the renovate one will need to be next since we need to make a test bucket and then just watch the fyra repo
we have some rules for renovate for pinning the package versions on dnf
so itll pin, then autobuild/merge, then
on: merge_queue do a new build w/ cloudflare stuff
bam.
just mentioning like
i was working on the ultramarine bootc stuff because of this
we can reuse all the workflows that bluefin does for um bootc
oh you get to go in the weekly report!
yay!
theres just some package on um bootc thats breaking the builds
but like, once we are done with bluefin um is mostly done too
workflows and stuff for um bootc are working great
if any of you get this reference i will be so shocked it's so goddamn niche
you know what it's too niche
Bootsy Collins
Apple Music
I'd Rather Be With You by Bootsy Collins on Apple Music
Song · 1976 · Duration 4:58
it's that bootsy collins song
which itself is a reference to flock 24
lol
was the branding fixed on the install progress page?
when someone asked the difference between collin's bootc and bootsy collins
so thats what this gif meant holy shit https://github.com/bootc-dev/bootc/discussions/1194#discussioncomment-12653268
GitHub
What is the name of the little seal on the logo? · bootc-dev bootc...
I dont think anyone really thought of a name for the little nice seal there, if not, id like to propose a name for it: "Bootsy"! It would be just so sweet to have that name as it sounds c...
stable needs a bump
then an iso refresh
but its fixed now
YEP
colin surname bootc
yep
fun fact it's a law that you have to listen to bootsy collins while working on bootc
like how you have to listen to jpop while working on coreboot
.
listening to this all the time
just gonna drop the playlist ww
current song is weeb shit
also don't ask why i have the polish cow song on the playlist
So right now main blocker is podman tries to check registry for image even though it's local.
yup
that thing you said ummm
about using oci-archive
is probably what we wanna do
its actually skopeo doing that
inside
bootc install
my guess is that going with oci-archive or oci might make it not check the remote registry
i already fixed a blocker related to this, too. we can use bootc_target_imgref to make it use the containers-storage transport once on the installed systemWhere you embed the container. Change it to just be a tarball on root of the directory.
Then for the bootc install you specify two flags
yup, 99% of this is already wired up
also hm, maybe not having it be a tarball would be best?
It's going to be a tarball if it's an archive
You could do dir:/ but I would be surprised if that works out better
we can just implement all this stuff in titanoboa then make it selectable on gh actions
@Gerblesh send your gh handle or website i need to credit you
mado says you helped on rdm
how's titanoboa going
yup he the goat
found him
well, does it actually... install properly?
working sb
installer
everything
it just works
yes
has it been tested on non bluefin
not yet
should work
should
Gnome only
lol
its just bootc
it should work cuz its just that
I'll try testing on my image
works in um
its all the same installation procedure
tomorrow, maybe
tulip can you add a KDE and Bazzite column
one for each
so we can start tracking those
aye, do you wanna be credited as your handle or first name
in...?
I haven't updated the website in a bit so it's kinda fugly but yeah
ultramarine, um
UltraMarine linuxI mean like
THIS IS THE WAY
custom images
I am convinced this is the way
lol
depends on what the "norm" would be but first name would work if that's what everyone else is using
anyways, this is fucking rad
can't wait to use this
whatever is comfy, first name is what i try to do but truly anything works
add the bluefin live ISO workflow to ur thing and change it all from bluefin to ur tghin
you need a bucket somewhere
good job chat
yeah
why sob reaction
:v
whar
"this... is a bucket"
there we go thanks everyone
oh nice is this the github release?
ehrm achschually the custom partitioning type doesnt have bootc support yet 🤓
fyra weekly report
i'm doing them again
next week we'll be done
just bringing in the curse
sob
Since it auto logs in.... We could have some sort qemu script to make sure it auto boots?
yup.
someone needs to wire it up
but absolutely yeah
explain, I don't get it
imagine a QEMU check that checks if the ISO actually boots before publishing it to the r2 bucket
this is genuinely cool
itll run a VM and see if it gets a GNOME session going
maybe a webhook or something
are you able to, like, change the images shown in the installer
not yet
HELL YESS!!!
ADD THAT
to the board I mean
fuck man
I wanted glorp installer
yupyup!
you can change the logo
fedora-logo-iconnot enough
ya ur cooked right now
we need to figure that out for bluefin too
I need random glorp gifs playing while I install my system
it's peak
commit to the joke
wrote the QEMU check issue
Probably cloud-init could be used for testing. Just something is needed to basically phone home after it boots
yup!
funny thing is that we could actually do this, wow
damn
@Noel @Zeglius need you guys to tell me the bazzite requirements for this
scope creep as much as you want
we need to write it down to actually know what to do in the future
Make a user (and make sure they are added to correct groups)
Custom Partitioning (no idea if bootc install nukes the ESP)
custom partitioning is like 50% done, just needs the copying part for bootc support
making an user is an issue cuz we'd need a new page
https://github.com/ublue-os/bazzite/blob/main/installer/lorax_templates/set_default_user.tmpl
This is what Bazzite has as a fallback
GitHub
bazzite/installer/lorax_templates/set_default_user.tmpl at main · ...
Bazzite makes gaming and everyday use smoother and simpler across desktop PCs, handhelds, tablets, and home theater PCs. - ublue-os/bazzite
we could pull a steam deck and make a user without a password
ideally we could use an OOBE thing
yeah this
maybe this specifically for the deck images?
cc @Kyle Gospo
Like it's shitty to require a user to figure out how to change name and home....
Also plasma is significantly more popular on Bazzite right now. So this is actually an issue
but yeah I agree ideally we either have a proper OOBE or integrate the stuff into the installer
user creation happens after install in the rdm designs
that is my design intention for readymade and taidan, to do exclusively oem style installs
plasma is fine, we can use like taidan or something
the problem is specifically the deck images
(since it's coming up taidan runs predesktop btw)
TULIP I AM COOKING
yeah we'd probably do a postinstall script for that or something, though tbh I'm not a big fan of the way valve handles it
cook it. 😈
like why cant we do oobes?
on
-deck?
is it cuz of steam?YO so like.
idk I'm not too familiar with bazzite deck
we neeeeeeeeeeeed that user creation shiiiiiit
yeah that probably will be a thing eventually
conditionally shown via the config and whatever
but like
we also need partitioning.
oh nice so it runs independent of desktop, sorta like the windows OEM thing
for Readymade
almost done
the partitioning GUI is there, we just need to wire it up with bootc
we need that secure boot enrollment.
one thing to worry about is on screen keyboard support/controller support
scroll up a bit
@Noel
the user creation for KDE is actually the big thing.
taidan probably can fix that for the desktop images, deck is the major problem here
I mean.
osk is being worked on w/ taidan, we probably can re-use those efforts for readymade on
-deck imageswould it be possible to just pop up like a terminal window with a gum tui thing that just creates a user on the newly put down disk?
controller support is a pain tho
no terminal, it would need to be one of those nasty
kdialog shit
and im sure people would skip itOR.
hold up.
idea
well...
bazzite bazzite as the default?
actually
@Noel ur too high come back tomorrow
bazzite bazzite for the deck images is fine.
because we do that today.
actually let's undo that
real
X_X
we have a keyboard now
no more nasty workaround
true
real users, real passwords
enforced
hell yeah
can you make kdialog pretty?
yes and it doesn't matter what desktop you go to after it
the ideal thing would be to use GNOME's or KDE's solution when those exist
but like
but intoxicated noel my goat... 😔
this fits in very well right now
also yeah
libmanette seems like the way to go for controller support
libmanette
The simple GObject game controller library.
i think cappy was working on early controller support in rdm+tdn
based
add things to the board now pls
we need to start writing these down
kyle wrote those
there's no way bazzite only wants 2 things lol
i think everything noel said is written down
LOL yeah.
how would a software keyboard work?
click on a part of the screen, the software keyboard will show up then you can touch on it to type whatever you want
@lea probably has better insight about this
oh derp, it's just the desktop keyboard lol
yup LOL
like, the gnome one, etc.
yea!
hahahah, killer
you gotta pick a osk though
use the desktop's for readymade yes
not my thing to think about
but for taidan we gotta pick one
bazzite gamer nerds will do it
or um nerds will because we need one for 2in1s and completeness
real
we should also get in touch with GNOME people to see if they know how to get controller support going
for readymade
i mean
only people we can talk to about this, right
just asking about libmanette in specific
oh i was confused yeah sure
gnome is uninvolved in rdm unless you don't pretend that gtk and gnome are seperate
they are practically not separate
yeah 😭
that's why i said don't pretend
yup
Alice Mikhaylenko (@alicem@blogs.gnome.org)
Steam Deck, HID, and libmanette adventures
Recently, I got a Steam Deck OLED. Obviously, one of the main reasons for that is to run a certain yet to be announced here emulation app on it, so I installed Bazzite instead of SteamOS, cleaned up the preinstalled junk and got a clean desktop along with the Steam session/gaming mode.
For the most part, it just works (in desktop mode, at least), but there was one problematic area: input.
Gamepad input
Gamepads in general are difficult. While you can write generic evdev code dealing with, say, keyboard input and be reasonably sure it will work with at least the majority of keyboards, that’s not the case for gamepads. Buttons will use random input codes. Gamepads will assign different input types for the same control. (for example, D-pad can be presented as 4 buttons, 2 hat axes or 2 absolute axes). Linux kernel includes specialized hid drivers for some gamepads which will work reasonably well out of the box, but in general a…
blogs.gnome.org
found a useful blogpost for the gamepad navigation stuff
if theres like anyone we can talk to about this its probably them
does anyone here have a potato pc? how's the startup time of rdm?
surprisingly good
i dont see anything wrong with it
is it basically instant? (~2s)
my dual core celeron laptop likes it
slow when ur on a slow storage medium tho
as expected
I see, that's os-probe
like 6-8s or something
if only you weren't in brazil, owen and i would be breaking down your front door with a brand new (used) z book
os-probe?
yeah
new fyra this week just dropped (for the first time)
https://blog.fyralabs.com/fyra-this-week-1/
Cats on a Keyboard
Hopping to It (4/14-4/20)
We're trying it again! Let's see what happened in the Fyra community this past week
feat @tulip and @Gerblesh hence why here
also funny foreshadowing in the post
You can very safely ignore this entire blog post
If you implement this as they did, it breaks Steam
And the deck has lizard mode if steam isn't running
In other words you can just not launch Steam for the live ISO
@tulip can you send me a lsblk of the device you own with the EMMC?
and
ls -alh /dev/disk/by-idoh, my
why is the "previous" button not as rounded as the "next" button
because it was just supposed to be a link style button
but it turns out those are a bad idea so they have a bigger hitbox now
:v it bugs me lol
No boarder or matching boarder? Looks like a link but bigger hit box?
looking fancy!
ill send you the output soon, just woke up
@M2 im thinking like, what if we just put titanoboa entirely in a single, privileged container?
kinda like bootc-image-builder?
it would be a lot more ergonomic at least LOL
that's what I was thinking is that this should be run like bootc-image-builder
sounds great to me!
it was just a justfile at first cuz it was a lot easier to debug
but this is starting to limit us LOL
my thoughts were have a devcontainer that includes podman-machine.
If you are inside a rootless devcontainer then it will auto setup a rootful podman-machine.
make sense?
ngl didnt really understand it
basically either expect that you are running in an already rootful container or spin out into a rootful podman machine
man i just wish xorriso wasnt that annoying to use
i wonder what even requiring rootful access on this entire thing
would be a lot easier to make it work w/ devcontainer stuff if we could make it non-rootful
that's why i'm thinking podman machine to alleviate that issue
OH I KNOW
@M2 we need some SUID binaries on the rootfs
thats why its requiring rootful access
like, the
podman save tarball thing doesnt SUID the binaries that should be SUID'd for us
so we manually add that, but managing any files with the SUID bits requires root
so xorriso fails, and anything else that modified the rootfs in its entirety requires rootfs
podman machine or some similar thing is most likely what we need :)it’s supposed to show that it’s not the suggested action but also not destructive
maybe make it gray instead? I feel like that might flow a bit better
your accent colour is gray
you guys are welcome to talk to the helium designer ping lains in the fyra server
whats up
oh based
libhelium is goated
@tulip should we switch back to nightly readymade?
i think so yeah
we needed those fixes super fast yesterday but they all should be on nightly now
right
ok finding
also @j0rge
our ISOs are smaller than an ubuntu ISO
while having like 3x the content LMFAO
the tons of flatpaks, and two rootfses on the image being smaller than an ubuntu ISO is kinda crazy
ok anything left to merge before I make new ISOs?
nope
should be fine
oh here's an easy one
when does the readymade-nightly timer go off?
we should just daily ISO build after
oh ur gonna cron it? LOL
lemme see
actually @jade @madomado do you guys know it?
it'd be like a cronned version of what we want
not perfect but good enough for now
@Robert unless you'd just rather do renovate and monitor it that way
I think cron is fine for us. We don't need a new ISO the moment a new version is released
ack
the renovate thing on non-nightly releases is important tho
omg
the new ISOs are all fixed up
yeah but we don't need it now now
nice.
a daily roll up will work
like, we kinda got the "I need a release now!!!!" huge parts outta the way
ok so @j0rge stable ISOs are mostly fixed, the bluefin logo isnt there yet cuz we'll need a stable bump + iso refresh
ill check if the latest ISOs are good
this is looking perfect tho
ok do we need to refresh latest?
ill check rn
10h ago
latest bluefin package was 14h ago
so the new latest ISO should be fine i think
that is a mado question and i honestly don't know where to check
but i'm fairly certain he's doing exams so i will try to find out
no worries we mostly fixed the issue already
we'll eventually move to renovate for this so theres no need
nightly does this
cc @j0rge
oh yeah I forgot libhelium was being used
every 24 hours
yeah the only interactive element on this page is the previous button which is why it's highlighted
oh
I've got most of the iso stuff to run rootless. Just working on the final build step
GitHub
EMMC storage media is filtered out · Issue #63 · FyraLabs/readymade
Couldnt find my laptop's internal storage cuz it is EMMC ant that seems to be filtered out by the installer somewhere
this is odd
we had trouble with it showing loop devs and partitions as seperate disks so it may be a side effect of fixing that?
mado mentionen that this is a bug on lsblk-rs
makes sense
Universal Blue
Call for Testing: New Bluefin ISOs!
Okay, I did some testing on the live system, and both my Intel A380 that drives my displays, and my Nvidia RTX A4500 that does CUDA work, worked. I ran my own CUDA tests, as well as GeekBench GPU tests (Intel, Nvidia), and everything worked as it should. I was not able to install, and created an issue on the readymade github. Finally, some fe...
ok so my takeaway from this is that we need to make the repart-handling code more resilient
theres a
FIXME: on the code there
ill try to address thisok let's just fire off a nightly ISO an hour after that, want me to PR that?
dunno, if you wanna PR it i'd be fine with it, i can do it too
on it
GitHub
v0.6.0: eMMC detection logic by madonuko · Pull Request #19 · Fyr...
resolves FyraLabs/readymade#63
@tulip did we want to remove on pull request?
yup, that thing makes it so ISO runs get canceled
we might want to up the concurrency there if you want the pull_request trigger
BASED
oh you made that!
whats a super basic use-case for this? i can test a super simple rust program on my laptop and try to get this working
like
can you run cargo test -- --show-output
ill be cloning it on my laptop
just a sec
and then try
.is_physical()
YEAH, IT'S ME ALL ALONG MUAHAHAHAAAhell yeah.
@madomado
dump from
cargo test -- --show-outputah, in blockdev.rs, the very end of the file
there's a test case
replace is_disk with is_physical
@tulip ok tested fresh iso, the secureboot stuff is awesome!
logos are still busted but this is expect right? until we gen new images.
and the enable-third-parties page is still on there.
@tulip where in here can I change the livecd schema? I think I can just add the 2 .desktop files to the dock, for community and docs. https://github.com/ublue-os/bluefin/blob/main/iso_files/configure_iso.sh
GitHub
bluefin/iso_files/configure_iso.sh at main · ublue-os/bluefin
The next generation Linux workstation, designed for reliability, performance, and sustainability. - ublue-os/bluefin
like how we do in the packages repo
https://github.com/ublue-os/bluefin/blob/b7d32a8b0d979e5c608aacf232420af29a0a118f/iso_files/configure_iso.sh#L92
like I feel like if I inline the gconf key here that should do the trick?
https://github.com/ublue-os/bluefin/pull/2402
here's what I'm thinking
Maybe don't pin thunderbird and VSCode?
Code shouldn't matter but an email client on the doc of an installer seems weird
Also dumb addition needed for Nvidia.
Since the kmods are signed with our keys and not Fedora's and we blacklist nouveau.... It's gonna be a pretty shitty experience.
Please add the keys to the EFI partition so we can enroll them before install
And if we're doing that, might as well include efi-shell and memtest
yeah those are just vestigial from the original one, if this works I'll go back and clean that up.
code is in there because the key is set even if you're not in dx, so when you reboot into dx that's how we magically put the vscode icon on the dock
oh we are kinda cooked there i think
we might want to straight up use the fedora keys or something
cuz dunno if we can even do it
sorry where do i do this?
We don't have their keys so we can't sign our kmods with them.
i mean like
cant we use nvidia from fedora?
we could do a post-processing thing
:P dunno
YUP!!! you did it right!
thats exactly what i was about to suggest
we do also need the readymade entry there
yup, til we generate new bluefin stable images
oh I saw the mangling you were doing before so I thought that's how you did it
you did it perfect i think
dunno if its the right entry, gotta review properly
but this is great
k
we might want to just not disable nouveau then maybe?
we can add the EFI keys to the boot partition tho
do you know how we could fetch these?
are they fedora packages?
Both are in fedora repos
we do need just the
.efi things directly tho
maybe i can install the fedora package on a container then copy the thing from /bootYepp. They are in the packages
dnf download maybe...They are not signed however.
cant do much about it, right?
dunno
We could sign them. But again only helpful for those that have already enrolled our keys
ok so:
- add EFI keys to /boot
- add efi-shell and memtest
- enable-third-parties page still exists
- help mado with eMMC troubleshooting
- review jorge's thing
But realistically right now our Nvidia ISO is pretty much pointless if you don't have our keys already loaded
right now, we can just remove the kargs and done! :)
add EFI keys to /boot
add efi-shell and memtest
First 2 are nice to haves. Mostly just remove the kargs for Nvidia blacklisting nouveau
put these on the board please
writing it right now
we might just want to have a configurable "this image supports nvidia" toggle on titanoboa
Actually yeah, just remove the blacklist. And we'll check if we blacklisted them in modprobe as well
that adds the blacklist if you have support n stuff
@tulip when we're done today let's spin off a stable build so we can be ready for next week.
oh also a thing
we need to disable
bootloader-update on the ISOs
cuz thats bootupd
it doesnt work because we dont mount /boot, but if we did we'd get the same errors that fedora is getting on f42 ISOs
ok @j0rge found out why the gnome-initial-setup thing is still showing the third party repos parthot!
the
/usr/share entry gets loaded before the one we have there
then gnome-initial-setup stops trying to read itoooh good catch
although we are kinda cooked, cuz i think that configuration comes from some package
we might want to just do that directly on the image build instead
ok fosstorrents is good to go, all we need to do is poke them when we have things ready and the automation kicks in
bazzite already has over 7k ISO downloads from there
NICE
I haven't been following this thread too closely lately, but is the readymade stuff on main? I want to try it for vauxite
working for bluefin
we have live ISOs for bluefin now w/ readymade and evwrything
The only thing missing for other DEs is the user creation. Gnome has a firstboot wizard to create the user, but KDE and potentially the others don't.
There's Taiden, another app by Fyra, but we've not investigated too much yet
theres a board for all the missing things
taidan would be the best solution but stuff like bazzite-deck cant use it
also why this? @Kyle Gospo
Probably joystick/touch navigation won't be implemented yet
Or low resolution scaling not being good enough yet
bazzite deck could
yup not yet
if it runs before steam it'd be fine
theres some reason why that wasnt a thing
Probably this
It's touch screen
Controller or not it's navigable
feel like i'm drifting from my bazzite homies 😭
hi kyle
Hi
taidan should be fine on a touch screen
what was the concern?
yeah taidan is good then?
i do wish we could habe controller suppoer
support*
its had as hell to do it
hard*
sorry typing on my phone
we could in theory it would just be a LOT of work
yup
i'm not an eng so i don't get to make big decisions like that ww
i mean
nor kde or gnome have support for controllers on their main things right
gnome-shell and plasmashell
Why not adding a default user then?
that can also be done but its kinda nasty
We already have preset username/password in anaconda
Yeah but is the most braindead solution
@Kyle Gospo said someyhing aboit that
we could just create the bazzite:bazzite user in a postscript
yup
Mm I wonder if can be done in the image itself. Like a pre populated etc/group passwd
i do like taidan as a solution a lot more
What
just scuffed
and wouldn't handle homedir creation
Cmon is a good idea, a tmpfile with an etc/group passwd
So if the user is dumb enough, can nuke the file and done :v
when are you guys getting a starboard
maybe a shartboard too
i also wanr that
No
I don't want a starboard filled with glorp pls
the sheer number of gloros
schmitboard
glorps
would be so good
if you didn't figure it out already btw shartboard is the opposite of starboard, starboard being for good messages, shartboard being for dreadful ones
glorp
glorp harder
glorp more
glorp faster
glorping is now legal
next person to say glorp has to implement the android kernel in bluefin
YOU TOOK SO LONG TO NOTICE
and we will make sure it works
I barely check anything daily tbf
The glorp badge is for anyone who is being naughty. All they can see is the general channel.
hey noel
do you uh
know what you just did
Being on the cusp of greatness?
omg noel my goat
do it then
why did it reply to that
bluefin but android kernel
you won't /j
@tulip was gonna spin gts here in a minute, you think we're done with the image for the day?
i think so!
Too late, I'm already porting it to the Nintendo switch.
okay jorge stupid question what is bluefin gts (vs normal bluefin)
bluefin fedora-1
hey we're actually doing this if you're interested in the world's most cursed steam deck
port it to the Nintendo 3ds and we will be proud
someone got arch working on a 3ds
Always a moving target.
who says we can't get bluefin working :clueless:
deadass??
yeah cappy is porting um
bruh
lawsuit incoming :letsgo: /j
we will share our ports with you guys if you ever want
"Bluefin, now on your Switch"
Bluefin on Chromebooks when
achieved a long time ago
bootc on chromebooks is the target
I'm running my own image on a chromebook
very smooth
other than running it off of a microSD but wedonttalkaboutthat
stock firmware or with coreboot
because i'm talking about the former
RW_LEGACY
edk2
:dread:
i mean fair but still
i thought mrchromebox killed his rw legacy stuff
let's move this to general
sure
if you're gonna
use SDL
don't rock the boat
Instead of using Cage with Taidan. Why not Sway which implements wlr-unstable-layer and use one of the virtual keyboards?
cage is deprecated isnt it?
we were planning to switch to sway or something
yeah sway would be better
Oh I remember, I forgot to write this down
but the intro text, the stuff below "Welcome to Bluefin"
how can we edit that?
Appears that Simon ser has pushed some commits to cage and released a new build 7 months ago
shoukd be just a config. i can try to add that
hopefully ofc, that also means no translations for it
hmm
I enabled stable-daily ISOs
that way we can just use dailies
OH we can edit the .po files directly!!!
and that will have readymade nightly in it, etc.
oh yup! should work fine :)
can u setup the skeleton and I'll edit the text?
https://github.com/ublue-os/bluefin/actions/runs/14583917985
toolip look what I did!
hell yeah! LOL
i think it should work fine, unless they are baking that into the binary, which wouldnt be too far fetched
well, at least we'll have the words figured out
then we can put it wherever
just put the file up in hackmd and I'll edit it
@tulip oh, it's still
stable on the liveCD when I try to install itwhere?
the oci url\
so the install failed cuz of it, right?
y
nice ok
should be a quick-ish fix
its because we cant get the final image OCI url on the image-info.json file, so it just puts whatever is the "flavor" there
the stable-daily flavor is... stable
ah nuts
lemme change them back
discourse is so awesome. 😄
did people report more cool things on discourse?
no I just reverted the "we now have dailies!" edits
im gonna sleep rn but ill try to fix this tomorrow
might wanna write an issue about making the image-info file more reliable
I'll write it tomorrow haha
good night calling it too
is_part sorry
like do i change the test name?
kwin
no uhh
there's a .filter()
whats keeping you guys from using the actual dwsktop session?
or compositor dunno
aight ill check it out
Go to bed 😉
nothing, but that makes it a bit harder to maintain
go to sleep
we've tried kwin, labwc, and cage
it's 1am for you tulip
get in your soft, comfortable bed right now
its not 1am
its 10pm :)
UTC-3
did google fucking lie to me
sleeping earlier hell yeah
yes
hucking fell
is that hardcoded on taidan? id like to test it out
LMFAO google's ai mode bring the dumbest piece of shit ever
fr
ai mode is literally somewhat worse gemini
"the Gemini we have at home"
Bro
Wait no nvm
Computers were a mistake
Not sure if secure boot testing is needed still, but worked fine for me
every one counts!
nice.
@Noel soon you will be free
and then now that you have your soul back. NOW we can play halo and shoot kyle
it's not hardcoded on taidan itself but it's in the service file
ublue x fyra bedwars nite when?
(i love bedwars)
OMG YES
its a bit of a pain to integrate but probably worthwhile. You can get 90% there with evdev though
because left joystick, dpad, and abxy are mostly standard
except on dualsense goddamnit
so I'd use evdev first. Code is 150 lines anyway.
https://github.com/hhd-dev/hhd/blob/4adbb8c0faebe502f9b57346b1681956a998756f/src/hhd/plugins/overlay/controllers.py
actually have a better one
GitHub
GitHub - hhd-dev/jkbd: JKBD: Joystick to Keyboard service. Converts...
JKBD: Joystick to Keyboard service. Converts joysticks into a keyboard device. - hhd-dev/jkbd
this is a full desktop mode service
this would be so based
i want an ublue minecraft server....
need a modpack server... :p
Didn't play since that update with the elytras :p
well guess what i have wanted (fyra mc server, we had one at one point)
you haven’t played minecraft in 10 years?
real
elytras came in 1.9 which came out 10 years ago
along with new combat
:huh:
I feel old
my first version in java was 1.16
from 2020
my first version overall was pi edition 9 years ago
My first version was 1.16
Beta
1.6 I guess
my first was 1.17
but
never java
always bedrock edition
:letsgo:
about the installer
is it possible for it to detect the amount of RAM installed and suggest configuring a swap partition/swap file?
ex. if it detects I have 4gb RAM it'll suggest it
future is dynamic swap
honestly i'm against swap partitions because of zram and swapping being bad for disks but if y'all want it we could do it
i did dynamic swap for deck images
systemd-repart doesn't support it directly, but we could manually impl that
noice
should make that a gh issue
I'll file it
repo link?
GitHub
Issues · FyraLabs/readymade
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
GitHub
Detect RAM and suggest swap usage · Issue #70 · FyraLabs/readymade
Basically, based on how much RAM is detected, the installer can suggest setup of a swap partition/swap file to the user. For example, if I have 4GB RAM detected, it'll ask me. If I have 32GB RA...
zram doe
interesting idea tho
CPU overhead doe
which is a problem on low end systems
ex. Celeron Chromebooks
very low prio problem for most images
ex: ur thing that ur using
no
the Chromebook i have has an Intel N100
also
isnt it
e.g. in english instead of ex:?not ex:
never seen anyone use ex in english
it's ex.
yeah.
apparently it's Korean :clueless:
I find it simpler to use "ex."
i use it all the time in portuguese
instead of 'e.g.'
(mainly since it's faster to type lol)
saving 0.0001secs per word typed so i can blow it all away gambling
real
it's not really an issue on my worst computers
and the wear that swap puts on emmc is not worth the trade imo
it's minor but eh
i see it every once in a while
y'all install to the eMMC?
the eMMC in Chromebooks does not have enough capacity imo
yeah ultramarine is small
a fresh install is around 6-8gb
I just have a lot of data ig :clueless:
on the lowest end chromebooks with shitty cpus and 16gb storage you can install xfce edition and still have 11gb left
we could also decrease swappiness but idk how much effect that'll have
which isn't a lot but it's better than no computer
isn't it like
60 by default
hehe funny time
yeah
??
ultramarine finally hit the microsoft store
after a week of fighting with them over "linux"
ofc you pull that shit lmao
tf does it even do
i do kinda wish you guys had better containers... might try to help w/ that eventually
we don't do swap on ultramarine by default
we don't suggest it either
no I mean the app
what's wrong with our containers
on Ms store
ultramarine inside windows
it's wsl
wsl 😭
fedora and arch announced on the same day as us and we're the first to hit the store
breh
I'm also not suggesting we have swap by default
I'm suggesting making a suggestion to the user during an install
they don't have to have swap
yes and i'm suggesting back that we shouldn't suggest swap on emmc
sure
if it detects the target disk is an eMMC, it can just skip suggesting to the user
we could also make this 40 be default if the user chooses to have swap
i dont think anyone here wants thst to be completely honest with you
maybe bazzite
for hibernation
yeah i'd push to disable the behavior entirely in ultramarine i want to push users towards zram only
tbf this isn't really that important of an issue/enhancement, but it could be useful to those that prefer/want hibernation as tulip here said
Constantly lol
https://github.com/containers/podman/issues/25950
currently blocked with using a devcontainer to provide a vm for building ISOs in a rootless context
damn... i have my fedora install on a poor 256gb microSD
it works surprisingly well
Oh lol
Yeah I just have to many a lot of Chromebooks
very good catch tho
But running in a privileged dev container I got the ISO to build
I have an NVME 1tb lying around tho
I'm thinking about buying an adapter to USBC so I can ditch the microSD
Ah yeah those are fun
You should
I'm thinkin
NICE!!!!!
yours is also slightly more green!
Unfortunately it seems to be stuck at this screen
wow i didn't know we had that screen
it looks kinda good tbh
oh shit what?
@madomado huh i wonder why that happened
maybe lsblk-rs crapped itself?
logs?
I don't have the logs but I ran it from the terminal and there were warnings/errors for the language selection screen but once I moved to the next screen (the one in the image) there was nothing printed to the terminal while it was hanging
Well get to test out the live ISO to fix grub....
Rufus resulted in an unbootable image.
Grub couldn't load it's files
interesting, it worked when i tried it
the live ISO
Trying fedora media writer
OH nvm i used fedora media writer on windows
rufus does some crazy mangling tho
I think it formatted the disk as NTFS and we do not have the NTFS driver on grub
oh that should be a quick fix then!
I think when you do grub-mkimage it only pulls in some of the modules
try
READYMADE_LOG=traceos-prober is never executed
!?
ah… this is some sort of race conditions
try this
we also need a page for locale and an after script to apply the locale for vconsole
i wrote down the after script thing
i feel like readymade shouldnt implement the locale screen
taidan or gnome-initial-setup can set up this for the live ISO
well, we need the locale before user creation
we need to keep readymade as slim as possible
yeah that happens in taidan
thats why anaconda turned out so shit
the language selection happens in readymade
and then taidan installs the locales
cant do that on bootc tho
hmmm
don't you just
dnf install them?
no
not like it's ostree?
you effectively have no dnf
hah?
on runtime you dont have any of this
its just the image, and DONE theres no runtime modification of the system packages n stuff
at least on bluefin we deprecated it completely
no package installs?
at all?
nope
at all
i was thinking we just make the post scripts for taidan the same way we do readymade
w/ info from taidan
then taidan doesnt need to do shit by itself
each image has its own way of getting stuff like language settings going and whatever
that exists already
so thats perfect
thats all we need
GitHub
taidan/scripts/oobe.sh at master · Ultramarine-Linux/taidan
Out-Of-Box-Experience (OOBE) and Welcome App. Contribute to Ultramarine-Linux/taidan development by creating an account on GitHub.
this is the script for us
this shouldnt be upstream if you can do it
can do
taidan isn't in terra yet anyway
yupyup!
im guessing you guys just made it specifically for um for now
and thats fine LOL i get ya
ill try to help out with those after the installer hurdles
im glad so many people reported issues and stuff for the installer
this is great
readymade is gonna be amazing whenever you guys fully deploy it on um
yup
GitHub
Allow usage on Live systems · Issue #62 · Ultramarine-Linux/taidan
This would be particularly useful because it would allow a generic setup experience for multiple desktop environments even on live ISO systems, it'll help a lot on troubleshooting issues and se...
something like this probably requires an rfc
ill be honest i dont know where to write that
yeah this is kinda awkward because we usually write rfcs in our internal outline
i can write it on discord and you paste it there or something
here?
could be a good idea?
sgtm
LOL
i think "discussions" is what you want
https://outline.fyralabs.com/doc/taidan-v1-lg5Qywh5AG
can you edit this
Outline
A modern team knowledge base for your internal documentation, product specs, support answers, meeting notes, onboarding, & more…
nvm you can't
ok maybe I'll open discussions later
issue found
huh interesting
@ledif is that on the latest readymade build?
(latest nightly)
I gave him the latest build
yes
It's on the binary that was sent in discord
something is very wrong then
maybe
tree /dev/diskOh lol I didn't add a disk to my vm
that's not really the reason
just try to run
tree /dev/disk -l in the vm firstThis is after adding a disk
heh win95 in win95
this looks normal…?
man you should make a CMD-like PS1 thing
i've seen something somewhere like that
ah…?
bug in readymade?
I thought it could be a bug in lsblk
huh.
wasnt it os-probe being weird?
can you also
RUST_BACKTRACE=full
also run mount and lsblkthe problem is probably from here
so either
/run/initramfs/live or / is not mounted from /devoh yeah
I really should get back to work though, bbl
its an overlayfs thing
right
LiveOS_rootfs is mounted to /
so the assumption is falsebut like
are you guy's live ISOs made differently than this?
the
dmsquash-live module in dracut specifies all thiswe haven't been testing rdm in our isos lately
oh!
fair
mainly because well
tired
need rest
it seemed perfectly good for ur ISOs
real
real real real
maybe something like this would fix it?
so this yields
/dev/sr0 on /run/initramfs/live type iso9660 (ro,relatime,nojoliet,check=s,map=n,blocksize=2048,iocharset=utf8)
fixed
didn't test but most likely fixedcc @ledif whenever you can come back
just realize this is gigantic
12.67 MB
its the images i think
i wonder if you can load external images via the gtk::image thing
248.1 KiB
huh.
no… it's a background css thingy
cant you use gtk::picture/image tho?
which is why the progress page is hard to customise
lea made the ui, so I believe lea came to the conclusion that it's not possible
huh.
but I just realized a hack
maybe a stylistic choice then
we can make it point to somewhere like
file:///tmp/readymade/bento1.png and then we symlink the image files to there during runtime
will try to impl tmr
actually not tmr
no time
next weekwhy not just
file:///usr/share/pixmaps/readymade-*.png
:clueless:should distro customisations go there?
usually yeah i think so
alright then
based
then we put like
readymade-panel1, readymade-panel2, readymade-panel3it's called bento but yeah
and change it accordingly on downstreams
yup
(including UM)
well need to verify if it works tho
thats like the general name for the UI object thing we are talking about?
yup!
it seems fine right
well… I don't know
ill try this on a VM
but lea calls it bento
so it'
it's bento
fair
id call it a card
very boring
smh it's a bento
oishii bento
What are the images for?
time to bikeshed how this ui element is called just like what happened in 999 other technical discussions in the linux community
its those cool images we can have on the install page
we are discussing on how we can un-hardcode them
(so we can put custom branding!)
Oh, as long as we can change the links and text we're good
those seem to be less of a problem tbh
yeah I have the code locally already
dunno where the translations come from but they seem to be overriteable at runtime
Then let's just do those
forgot to push
thanks for reminding
We need to publish isos first so the branding can wait if need be
Like we can just push the MVP
The logos should work now?
place sounds wrong?
shouldnt it be readymade?
sent a suggestion
LMAO
uh
uhhhh
hey that's why the funny
what doc is that in
Outline
A modern team knowledge base for your internal documentation, product specs, support answers, meeting notes, onboarding, & more…
wip
ah
just writing out everything I think of
okay comment on the rerun thing
you comment on the outline
or both
okay sure i'll levae it there
to be completely honest i think they have very different usecases
having them be explicitly different will be a lot better for maintenance in the future too
yeah
i also mostly disagree with having the ability to rerun taidan but for bazzite's usecase where they do system tweaks in oobe instead of just apps and normal settings like UM — having a welcome app where you can toggle those tweaks may be needed
why would you rerun taidan tho?
bazzite wanted that as an option to reconfigure
we can use yafti for the system tweak thingsies
oh reconfigure the system?
but that doesnt make sense? huh
the design intention with taidan is that it would run one time and everything done in it could be undone in the settings app or app store
id just remove the "app store" part
that thing is so much fucking trouble that its best to have a separate thing specifically for that
thats what yafti was
we have some apps in taidan it was initially a browser picker only but then we decided to showcase common apps
having this "three step" architecture thing wouold be best tbh
live ISO (taidan -> readymade) -> host (taidan -> (something similar to yafti))
"three app" not three step whatever
taidan already fixes a LOT!
KDE and others dont have OOBEs at all, and taidan is amazing for this
I'm thinking like
app management should be done afterwards, either via another app, or whatever else
we might need a welcoming app
yeah thats what im talkin about
just for a11y and i18n configurations
instead of baking them into taidan directly
why tho?
no that doesn't make sense what
this stuff is super important like, first thing you do when you start up ur PC you want those
imagine navigating through taidan w/o knowing english
that's usually done once first time and then never touched again, and you can just open settings to change it
why make another app
it should be done by the welcoming app
and then taidan runs after the setup
what is a welcoming app in this scenario then
and taidan further setups the stuff
thats effectively just a taidan step then you know
just a11y and i18n settings
because that's the most duplicated part between rdm and tdn
but what does a separate binary/project do for that?
i mean, you need this stuff
maybe readymade can just not show the language part if the locale is configured through taidan or something
im so confused now
honestly that should be how this is done
also uhh
okay let me explain design intentions now
readymade's i18n everything is only intended to make the installer readable and navigable and should not have any bearing on the installed system
taidan's i18n stuff needs to be more fleshed out so you can actually use your system
yeah as you can see I should eep I'm mixing up rdm and tdn
readymade is also designed with the assumption that it will be the only thing running, just like taidan
additionally, everything configured in taidan should be reconfigurable later without opening taidan, everything should be able to be done in the settings app or app store
this is how i intended it to be done
we need a library that rdm and tdn both rely on tbh
so we can factor out stuff
and then we can get rid of the eww gnome deps
i thought about this
cc @madomado
I don't think system configuration needs to happen before installation imo
in the live environment and installed system
"system configuration" would be mostly just creating a new user and setting locales
cuz new people will be confused by it at least during installation
network connections n stuff
maybe we just set locales and do wifi and keep the normal liveuser user instead of creating one
yup.
the liveuser user is a given by the livesys-scripts
but we dont need that if we use taidan
w/ that idea
it also reconfigures some shit that shouldnt be configured btw
i think it'll be somewhat confusing to create a user in the live env then making a different one on the installed system
nono i mean like
the live environment would not have a customizeable user
it would be just
liveuser w/o a password
like it does right now
taidan would just drop you there w/ that useroh okay maybe
so launch into taidan, config internet and locale, drop into desktop?
or launch readymade from the cage/sway/whatever session taidan runs on
just imagine how gourgeous that experience can be on a handheld for example
would be nice on a handhed tbh
yes!!!
okay i see what you're cooking
my idea is that readymade would pass those configurations to the newly installed system
like the keyboard layout, those things
so the taidan on the installed system wouldnt need to do much
maybe just a file on
/var/lib/taidan-installation.json
that has all the things that the user configured on the live environment
no direct coupling between them thokinda defeats the purpose of OEM style installs hmm....
nope
if you do OEM installs taidan will just run normally
w/o any override
the goal was for every install to be an OEM style install
i mean, if we already are running a live installer theres no reason not to persist the stuff from the live installer
"regular" OOBE would just run if there werent any overrides from the installer
hmm
i'm off the beleif that nothing from the live env should persist
i feel like that will just be annoying for the end-user
but that can also be just configured
by.. not having the override
wonder what mado thinks
and @lea
cc @madomado
this stuff seems just amazing
and not too shit to maintain
imagine someone got their steam deck or something and is going through this process
Persisting form the live environment would make sense to a new user
click click click install click again write ur username done
i think that the installer environment and installed system should be as distinct as possible but that also combines with my beleif that installers and live envs should be different things don't get me started
but i wanna make sure you guys get what you want too
the user doesnt need to know that there even exists a live environment
its just an extension of their new system
"oh i need to click these things before i get bazzite ok"
you know something i just thought of
a page in taidan being like "Who are you installing this system for?" "Myself" or "Someone Else"
myself -> configs from the live env persist, someone else -> they dont persist
either taidan or rdm would do it dunno
oh that sounds nice
this can and will be a top notch experience this is amazing
Fyra labs saving the Linux desktop 🔥 😭
teehee
this is the actual fucking way to go imho
android has a similar experience, windows also does
but we can refine this a lot
It absofuckinglutely is
cc @Kyle Rankin @Kyle Gospo @inffy @j0rge would like you guys to also comment on this thread, specifically starting from this message
pinged you guys in specific cuz you guys have different views than i do about this stuff
I don't have an opinion I'm only focused on mvp
and the clock
clock?
oh yeah
oh the website thing is done pretty much
bluefin just needs readymade, what we have now is perfect the bugs just need to be fixed.
yup!
ww, they are cards, I mean the layout of the cards is a “bento layout”
my kid has a bento box
it looks like that from the top
that makes sense
I think this is a good idea
nice!
I should prolly read through the git log and catch myself up, at least it looks like a lot has happened while I was gone ww
All of it looks great though
I think there is the two different options, some like that the onstaller does eveything vs installer+welcome app.
I kind of prefer the simple installer rhat set locales/keyboard layout/lamguage and network and then the rest would be done after installation
@SFaulken
Mostly sounds good.
I might be able to use some of it
got another computer
BLOUEfIN
bloo
never actually daily driven bluefin
or aurora
BLUEFIN USES BASH?!?
and installs starship from brew (it is in terra)
we wont include it in-image we are leaving the base image super clean
fair
we source as many things from brew as we can
im noticing that
it's my mac workflow pretty much
all of my cli stuff is in brew and all of my desktop stuff is either normal applications or app store apps
but in this case it's 100% centrally managed apps which is so nice
apps should not be allowed to manage their own updates!!!!
heh that's not how we roll here
why brew?
because it's awesome
what's the alternative also
it doesn't make as much sense for a distro like UM because we have normal RPMs but for bluefin where that's quickly becoming less of an option it's great
i havent really put in many hours on bluefin so i should probably play with this more
especially since we're doing so much collab with y'all now
gotta add a bluefin and aurora machine to my fleet of non-um machines
(on the modern side it's 2 windows-only machines, 4 dual booted windows/um machines, a quad booted um/windows/bazzite/macos machine, and a mint machine for testing owen's free geek script)
Two more laptops won’t hurt :clueless:
get them goodies
i have ~60 laptops
I know 😭
That’s why I said that heh
it’s less of a game of how many laptops do i need
and more of a game of how many does it take for my mom to hold an intervention
My parents already that I am crazy for owning 3 laptops
if i had 60
My family gave up
I have shelves and shelves full of laptops
occasionally my sister insists that you only need one computer overall and i’m like well what about the different architectures and she’s like who cares then i start explaining my work and she stops arguing
"WTF WHY DID MY LUNCHBOX PR DISAPPEARED"
"WTF WHERE'S MY PR"
"WTF WHERE'S THE GIT BRANCH"
"???"
"oh I'm in the taidan repo it should be in readymade"
oh its already not an option
if you use pure image mode you just cant install rpms
at all
blud does NOT see the difference
did you see it? no? see
ahhh
based\
Real.
oh tulip
the text/links on the installer, the ones that say ultramarine, etc.
I can't find an issue is it still an issue?
i think we can change most of them
ill try to do it, but we should write an issue
there's a pr
ack, thanks!
I think that's basically all the cosmetic things?
aaaaaaaaaaaaa
ok it just can't access the filesystem
is yaru no longer default or is this a readymade moment?
readymade moment cuz the css can only talk to elements inside the gschemas bundle
cant we just replace the gschemas thing?
like i've seen tons of these scattered about
no chance we can load it on runtime?
if so we can just fix that on the ISO building itself
on the installed system for reference, i know rdm can't be themed
i mean like
isnt there any glib function that allows us to load gschemas after compilation?
instead of bundling them inside the binary?
im completely fine running
glib-compile-schemas on the ISO building scripti think that makes sense
but i just wanna make sure we're on the same page
when i booted into bluefin after install the default theme was adwaita across the board
yeah it's all libadwaita
we change the mono font, that's about it these days
I don't know how to do that asd;lfkjas;
its ok someone might know it
mfw the docs for the css is terrible
for gtk
real
I hate gnu
I blame gnu
everyone should just switch to msdos
i mean its not GNU anymore tho
gnu is still stuck maintaining their coreutils, breaking glibc and not-maintaining xorriso
it's basically gnu
/hj
if people are still calling Linux GNU/Linux then yeah gtk is gnu
i mean, it isnt gnu linux anymore
its just linux
some people still do]
so yeah
08[9 e4i;sro h/dgyeqwar[o';G? VHBZCy9[8awi g;p vgq2w93agv n
losers do
sorry the cat just uhh meowed on the keyboard
(me)
someone joked that since msft wouldn’t let us put linux in our name on the microsoft store we should’ve put “Ultramarine GNU/“ as our name
LMFAO
real tho, linux is trademarked
“Ultramarine systemd+” was my retort
"Ultramarine" would have been perfect
ib4 ubuntu takes GNU/ out of them this year
and they cant put linux on it
GitHub
Readymade errors out at os-prober screen · Issue #91 · ublue-os/t...
Framework Laptop 13 DIY Edition (AMD Ryzen™ AI 300 Series) Bluefin: https://download.projectbluefin.io/bluefin-live-stable-amd64.iso Okay, booted up, stuck at waiting for os-prober. Running journal...
any word on this one?
we can move issues to the readymade repo if need be
mmhm that should be moved to the readymade repo
i think @madomado already saw this
maybe its fixed on a PR already
yeah if we need to give one of them perms to move issues to the readymade repo we can do that, whatever makes it easier
so we'll have fresh images tomorrow night, then we can do another round of testing?
mmmhm!
the offline ISO thing isnt fixed yet tho
at least the dinos will be working hahaha
true!
it got merged into bootc we just need a release or carry a patch right?
nope we dont even need that
what do we need?
the bootc change was just making that the default
we just need the
bootc install command to have --skip-fetch-check on itoh!
on older bootcs (pre 1.1.7)
let's do that then?
thats why it worked just fine on most of my bluefin LTS testing
we're running out of time. 😄
ill PR a change to readymade just allowing us to have a
bootc_args thing
so we can do bootc_args = ["--skip-fetch-check"]
there you goyeah
man we need this weblate thing for like the docs
yes!!! that cool framework we are using does all this stuff for us
there are docs about docusaurus integration w/ crowdin (similar to weblate)
i just havent touched that in a while
yeah maybe this summer I can look into it
ok so the images will fire off tomorrow night so if we can land critical things before than that'd be ideal
fixed
will merge
then I can spend all sunday/monday testing.
oh awesome, thank you!
that and the bootc thing were the last 2 major issues right?
@madomado this is beautiful btw
I mean other than missing functionality, I mean as capable of being the default on the ISO, MVP disk blaster.
that "device busy" thing is a problem
oh is the os-prober and the device busy two seperate things?
my friend tried installing bluefin w/ windows on another drive and there was that "device busy" error all the time
mmhm
not dual boot, just another drive
right
did it find the other drive and try to mount it or something?
for whatever reason repart seems to be trying to open up the other drive
this still needs a toooon of debugging tho
ok after you land the other fixes update the issue and I'll update the thread, that way we can get as many people testing as possible.
we'll call this round 2 or whatever so everyone can try again
most of the issues talked about on the thing are done pretty much
let's update the board then
the os-prober and repart thing aren't on there
mmhm they should
they are on the readymade repo
instead of titanoboa
ok then make one tracking issue in titanoboa for "MVP" or something
then just link to the few bugs, os-prober, repart, and then the links that still point to ultramarine
that's the three items we need right? bootc online is already a card on there so we got that one.
Just trying to determine which list of bugs is preventing shipping.
because like changing the artwork etc can all happen later.
then the links that still point to ultramarinethat one already has a PR, just needs merging afaik
bootc online is already a card on there so we got that one.PR up too! then we just need to update our configuration on bluefin til we update to bootc 1.1.7 ill make a tracking issue anyways tho
its this
oh I see, this is baller!!!!
also fixes the links!
ok so everything is inflight except the repart stuff, ok. Nice.
yupyup!
looks like the images are there too so we can just swap those out?
theres a
FIXME: on the repart code LOL
sadly i dont think soI am deathly afraid that coreOS will swap for F42 on tuesday
we need to figure out something in glib that allows us to import new images n stuff
like, one more week would be perfect lol
ok, the stock images are fine, no problemo
OH! @lea didnt you make the UI? maybe you know if theres a way to load schemas on runtime?
I am going to put toy dinosaurs in my kid's bento box and take a picture, should make for nice bling for the next update I post on the forums, heh.
Gio.Resource.load
Reference for Gio.Resource.load
oh shit @j0rge found out how to do runtime stuff
pr incoming to readymade
LET'S GOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOO
include! -> build time
load + register is what we waant for runtime
hell yeah!
We're cooking! We are so back, it's so over!
WE ARE SO BACK
it WAS so over. RAAAHHHH 🦅
I'm asking the artist to make a harpy eagle version of bluefin
LES GOOOO
everything that person does is amazing
this thing eats monkeys right out of trees.
what.
that thing is brutal
just an absurd animal, largest eagle in the americas
I think the phillipine eagle is larger tho
LMFAO it looks so mad
and then also a terror bird version
my commit has a
babe in it
nice shout out to your girlfriend!
LOL YEAH
girlfriends
that's what it's called on the store
holy crap
my favourite scary bird is the cassowary aka "murder bird"
yeah I saw one recently
ooo
that's the closet we'll get to a real bluefin
always wanted to see one in person
inb4 jorge retires from computers and has a farm full of evil birds including a cassowary named bluefin — like kyle's goat farm
DHAWDFHAHAH
oh man thats awesome
i forget that they walk like pigeons
it's so goofy to see something that could tear you limb from limb walk like that
search for "shoebill stork" if you want some nightmare fuel
looks cool- oh no
greyfin
It’s this bird fr
For readymade.
What state is everything at the conclusion of the installer? Is everything still mounted and pressing restart triggers everything being torn down?
For titanoboa. I've made a devcontainer that has all of the x86 stuff in it.
It can simultaneously be the default builder and the base of the action if we wanted to.
just a single bootc mount thing after its done
/run/bootc/mounts is still mounted after readymade finishes installation
thats a thing on bootc install directly, maybe we should handle that?So can I chroot in to run useradd?
nope, the postinstall scripts run on a new container
New container?
and its as if you were chrooting to
/sysroot
kinda like podman run --rootfs
try running podman run --rm -it --rootfs /sysrootOkay. So that's a chroot.
thats the environment, but with /usr and some folders from the host mounted
i thought you'd manually do it
yup the postinstalls run in a chroot/container thing
Okay. So I can also do the chroot manually.
Now to debate putting in arch-chroot or just use podman's --rootfs
i mean, yeah
you do need a way to know when the installation is done tho
I'm assuming when I'm on the last page it's done
how come you know when the user is on the last page tho
also why not just use the postinstall scripts?
I want to do useradd. Since we don't have that first boot for KDE/cosmic
Guys, just do a tmpfile for passwd :v
oh y ou need to add this directly to a fancy place on
/ostree
thats why i was pushing for taidan cuz doing this on the ISO is a lot of trouble
useradd wont work even if you chroot cuz the rootfs isnt anything like what we get post-bootWould something like this work?
oh shit yeah it absolutely would!
you do need to find out what the checksum is
although readymade is already doing some sandboxing so nspawn wont work
Ah good, a rabbit hole less to deal with then
looking to help get titanoboa in the image template or ready to use for other things. I'm gonna try knocking out some issues
@j0rge @madomado
external gschemas WORK
LES GOOOOOOOOOOOOOOOO
i just copied the damn centos wallpaper directly cuz that was the first thing i found
this is basically all thats needed
but this handling is unsafe cuz it crashes it if the file isnt there
I'm going to put it in the release announcement
unsolicited suggestion
making the logo 20px instead of 25 makes it look more in line with the other icons (20 vs 25)
Going to airport but sgtm
18–20 looks like the sweet spot
stuff happened
btw thanks tulip
a lot of stuff happened
amazing
we're so close!
we also have our weekly builds running tonight so we were motivated. 😄
it's small but I managed to contribute something: https://github.com/ublue-os/titanoboa/pull/93
GitHub
Use
uname -m instead of arch because it's more portable by erik...I was trying to work with titanoboa on manjaro (an arch derivative) and I found out via a lot of explosions that the arch command doesn't exist by default on it. Since it is the same as una...
every little bit helps, thank you!
I'll tell you guys how to migrate while I go buy my lunch
first of all you need to have
/usr/share/readymade/bento/{1,2,3,4}{,-dark}
currently it doesn't work because well gresources
for that I haven't really looked too much into tulip's pr but theirs work so I'll look into how that can be combined
these will need to have the background images for the progress pagecheck out my pr
we can just override the gresources directly now
your pr is merged btw
I just haven't looked into it
I found no "rm -rf /" so I merged /hj
will look while walking
second you need translation files
or you don't
because you can just use ours if you don't care to change the title/description of the bento cards
(the distro name is replaced automatically)
to do that, copy
po/ into /usr/share/readymade/bento/
third of all new breaking change dropped for the config file
in readymade.toml you must now specify the bentos
check out our toml for ref (you can probably just copy)
templates/ultramarine.toml
EOF
forgot the CSS doesn't work
will revert
also the disk thingy page is truly fixed now I thinkLOL based
i just tried to handle that part as safe as possible
sohuldnt that be like a global override? (like for everything)
dunno
shrug
yes, will impl
tulip I PRed this to packages repo - I dunno anything about a package version bump rpm thing tho
that's temporarily the uh workaround I guess
ok so like 99.9% of the things are done then pretty much!
no worries ill just rebuild it whatever
i can bump it for you too
its just so the copr gets happy
look how good this looks
look at the little baby, so smol
ITS GREAT 😭 LOVELY
not tuna
banned
petition to rename aurora to saba
What does Saba mean
it's a kind of fish
Ah
Well I hate fish
Sooooo
NOT SIGMA
no bluefin for you
smh
That’s fine
- global LL override
- revert CSS
- testing?
the biggest problem is that "device is busy" thing on repart
i have no idea how to fix it tbh
need to get a reproducible thing to get it working
the MBR header also fixes BIOS if we can figure that out
right I think I found a very sketchy way
rerun the readymade subprocess
what does that fix doe?
under the assumption that the funny device is busy bug is a fluke in repart, if we rerun the subprocess then it probably might work the next time?
it is yeye
sometimes its a fluke
cuz windows makes it crap itself
I cooked so hard this week
Rename Ucore to Albacore
wha-
also I haven't tested my code
I'll leave that to everyone gotta prep for my tests
Thanks, we'll spin new things!
another kind of tuna
it’s the kind we get the most in the US
https://en.wikipedia.org/wiki/Albacore?wprov=sfti1
Albacore
The albacore (Thunnus alalunga), known also as the longfin tuna, is a species of tuna of the order Scombriformes. It is found in temperate and tropical waters across the globe in the epipelagic and mesopelagic zones. There are six distinct stocks known globally in the Atlantic, Pacific, and Indian oceans, as well as the Mediterranean Sea. The al...
for mostly ready to eat applications and otherwise cooked applications
the japanese say otherwise
So regarding the user setup after installation, rn is only planned for gnome, right?
Thinking of maybe adding something similar to the livesys package, some default user able to run whatever if no regular user is detected
From there we can have something that wipes the runtime state or something
But needs to be baked into the image
Not for readymade or anything, just writing here for the record
that'd be dope
ok so how can I check the freshness of the readymade nightly?
I'm going to spin new iso's tonight and everything was already merged
KDE is working on this. They see it as known gap between them and gnome.
I have no idea where other DEs are at but for something like sway it would never be a thing.
Gnome simply already has this done.
I mean in general, should be agnostic
But eh, if they gonna do the work for me, fine
Bluefin is MVP
Least effort needed
User setup will be needed in the future simply because you could only install GNOME without it.
But let's not even think about it until Bluefin is situated
we'll have ISOs in about 1h, waiting on weekly builds to finish, which failed last nite so we're building those now
then new ISOs, then I'll update the thread.
then I'll test all day
then the priority after major bugfixes should be the KDE stuff next IMO, but up to ya'll
sounds good to me
the bluefin website PR is good to go for whenever we merge the "promote live ISOs to the default ones" PR
so bluefin, except for the branding, is pretty much done
yeah, I'll do that tonight
as soon as we can prove the old ISOs are useless.
mostly it's the systemd-repart that concerns me
everything else is gravy
just tell me whenever you do it cuz i need to change a single(!) string on the website PR
mado was working on that actually
oh just add it to the draft PR
@j0rge
maybe the retry stuff will be enough? what do you think? do we have a bingo card?
lets see!
repart is kind of annoying for whatever reason with that
yeah this will be the major goal tonight
the ideal thing would be having someone with a windows dualboot (separate drive) test this
ok I'll say that
like, we need someone to confirm that readymade WONT break if windows is anywhere near
cuz repart failed very consistenly on my friend's machine since he has windows on it
ok so your friend is key then. 😄
but the partitioning worked for whatever reason
it just showed as a failure
hopefully this fix mado was talking about makes it so readymade wont get mad then
ok building ISOs now
going to get caffeine bbiab
ok new ISOs uploaded
backing up the old ones
ok the doc icons on the dock didn't make it in
readymade fails with:
https://github.com/ublue-os/titanoboa/issues/94
ok this is what we're stuck on hopefully it means we can just respin ISOs and not images?
Yeah, it's complaining about the live iso toml, so should be something needs changing in the configure_iso.sh script
oh!
thats a new bug
thats for the text on the installation screen :)
I reopened the pinned docs thing too, those didn't make it to the live session, that was fixed before I think?
oh those arent there yet
did anyone implement it yet?
i didnt do it at least
yes I PRed it
oh!
does readymade show up on the dock?
yes
okok
it just crashes silently, I got the error when trying to run readymade in a terminal
so I think it's just the name of the .desktop files?
that part yeah
lemme see what are the right names
we need to change something in the config
i dont know whats the right format but it shouldnt be too much of a pain to figure out
ill download the ISO and try to get it going
did the format change?
nope, a new entry was added
We just need to add this to the TOML
And optionally desc, link and icon
so we can add the custom URLs on the installation screen
k standing by then we can push new ones
ok backups of old ISOs complete, we're good, just need that config file and we can try again
my thing downloaded
lemme just spin up a vm and get a working config real quick
ok here are the branding assets
do we wanna do these now or later?
im fixing up the config first jsut to make sure its gonna work
I filed an issue for the bento images now I can't find it lol
@j0rge we can open arbitrary things on the installation page
i.e:
file:///usr/share/doc/bluefin-docs.pdf
.yeah!
we don't have the pdf yet but we can do that later
yupyup!
also, it'll be pinned on the doc anyway
ill just make it so all of these will be ummm
online things
it looks really good when both of those icons are pinned on the doc
i wonder whats making them not show up
it's fine that can be for later
can you run
gsettings get org.gnome.shell favorite-apps if you're in the live session?
discourse.desktop documentation.desktop are both in /usr/share/apps it should be workingoh lemme see
dang i think we got a regression
huh the schemas didnt change
@j0rge
huh
look in github I added them there
unless they need to be somewhere else?
@madomado dang
i wonder what broke this
ill try to find out
but yeah i got a functioning config
yeah let's just focus on that
the icons are just icons nbd
ok I was able to get other things merged around the place
ok fixed the toolboxes too, how we doing on that config file?
nice!
the config is done its fine
im actually figuring out the branding now
understood.
i dont know exactly what to put on the lil texts there
on it.
we can also set different icons, too
oh do we have the text for the other screens somewhere?
fixing this now
i think we can do overrides somewhere
not super sure
lemme jsut change the icons on this one
left suggestions
i dont thnink you clicked on "start review"
oops
ok got it
though we can force this
ok what's next?
ur other thing didnt get applied
it's ok this is fine, I deleted the old one
we can fine tune after we verify it works
yeye
okok so
wait
list:
- branding on the installation page
- figuring out what the hell happened with that bug i pinged mado earlier
- fixing the documentation icons on the dock
- adding offline docs to the image/ISO
yeah
matches the board
@tulip ok should I make new ISOs now?
i think readymade is just straight up not working right now
fails with this
kk
i think i found the issue
trying to build readymade and see if it works
heres the tracking issue https://github.com/FyraLabs/readymade/issues/80
GitHub
Regression from
0.12.2 on InstallationType handling · Issue #80 ...Seems to be failing due to some mishandling of the InstallationType enum, not particularly sure where/when that got changed but this is readymade-nightly, ebe8fdd at the moment. 0: Readymade subpro...
and that crashes the whole thing?
mmmhm
seems to just be missing a
rename_all = "lowercase" somewhereI already pushed the button so at least we can see the bento config working
@j0rge this is starting to look so pretty
nice nice!
my windows is super wide
Looks like a VM issue? Where are your top bar buttons to shutdown etc?
Maybe your VM isn't 16:9?
I tried resizing it, I'll check again when I remake it
heh i think this is an issue with the latest readymade builds
but the crasher is there so I can confirm it at least
im doing a git bisect to figure out what broke it
OH
@j0rge the window got super big cuz we put way too much text on the bentos.
the ones I just edited?
this in specific
ok I'll PR a fix now
done
OMG i somewhat fixed it
now i got a new error
Progress!
wrote down what i found on the same issue
ok so
@madomado this makes the installer actually get to the install screen
but it seems to hang there, no logs, no nothing
ill work on the offline docs for bluefin cuz this is going to take a while for me to actually figure out
or wallpapers
omg true
that one then
cuz that one is very close
standing by for bluefin pdf after tho
yeah
that's next week. 😄
GitHub
Unattended install? · Issue #81 · FyraLabs/readymade
Hi, does it support Kickstart or any other means of unattended installation?
anaconda/fedora contributor apparently
taking the next 24h off btw
I messed up the type conversion
fixed
OEMs…? interesting…
readymade nightly is now commit based
so whenever you commit within 10 minutes an update will be built and pushed
That's awesome! Thank you!
NICE
ok so we are close then!
latest readymade build should work fine
beautiful, thank you!
this definitely introducted a regression
on the readymade subprocess part
stdin seems to not get written to
trying to revert your changes and see if it works
its this part in specific
@tulip is it worth for me to respin latest and build ISOs?
nope :(
latest readymade commit is cooked
but shouldnt be too bad of a fix, ill revert those thingies
it's fine, the hostname issue is worse
@tulip we should have an updated ISO here soon even though it'll still be busted
sgtm!
mew I'm back
whatdup
pinged you on a few things, but i sidnt experiment much w/ might have happened
is this from git bisect or something
does this work after revert
I'm confusion
got it on git blame
this seems to be pretty much what is going on
stdin being piped makes it so serde cant stop reading it
so the subprocess hangs forever before doing anything
havent experimented with it, no idea
probably would work
no code regarding the behaviour with stdin changed in this commit though
I fixed the issue
turns out the stdin was dropped so yeah
ok firing off another fresh set! https://github.com/ublue-os/bluefin/actions/runs/14733101666
GitHub
Build ISOs (Live) · ublue-os/bluefin@dc9d395
The next generation Linux workstation, designed for reliability, performance, and sustainability. - Build ISOs (Live) · ublue-os/bluefin@dc9d395
bad timing, rdm is broken rn
oh ok I'll cancel
do that after 20m
okey
ww
the nightly build is still in ci
no worries, they set it to rebuild the rpm on commit
github, pls compile faster :p
GitHub
bump: nightly · terrapkg/packages@d493a3e
Monorepo for Terra Packages. Contribute to terrapkg/packages development by creating an account on GitHub.
yup LOL
figured that was going on
i wonder what the fix was
can now
more testing
TUNAAAA
also we screwed up the bentos on the latest confjg
gotta change some stuff
no tuna bento for you
only california rolls
rip....
both are great.
it worked (this is a failure from
postinstall/reinstall_kernel so yeah)
i do love california roll
i’m a sucker for crab stick
based based
why does it say Instawwation wucky 😭
en-OWO locale.
please tell me this is not a real thing
its just as real as glorpfetch.
which was nuked from the repo
JUST as real
also hi person named after flower
based right
no
based and flowerpilled
nop
whats the current issue with the installer though
latest commit should work completely fine
fails at installing the kernel from what i can see
nope
misconfiguration from mado's part
fair
ill actually test this just to see
ah k
also i need to change the rpm spec so we can override the stuff easier
excited to ship this to aurora soon
cuz currently if we put stuff on
/usr/share/readymade we r kinda cooked (conflicts w/ the readymade package)maybe i can already install my new framework with it when that drops lol
its real
it will not being going anywhere
because catgirl company ™️
device error is still there
it shows it tried the three times. This is on a single disk nuc with a version of bluefin on it already
huh.
is it the same error?
i wonder if it would work if we manually run repart
added what I could of the error
interesting.
so it's reproducible…?
but only for some people
I only just put out the call for more testing so still waiting on more reports
but that was on my test NUC
man jorge what is your desk like
⃤
YouTube
SIX CONSOLES, TEN COMPUTERS, BUNCH A FUCKING WIRES, SOUND PADS, SPE...
streamdeck, analyzer, fuckin... moderator stance, sound filter, six monitors, pedals, drums, A GAMING GUITAR, CONTROLLERS, SIX CONSOLES, TEN COMPUTERS, BUNCH A FUCKING WIRES, SOUND PADS, SPEAKERS, ANTENNAE, SATELLITE, DOG, DOG CAM, CAT CAM, CAT TREE, CAT THIS, and a buncha dogshit dud
i have 60 laptops but about 30 are under my bed
trackball mouse maximalist respect ✊
which 2in1 is that
slide
Heh, I'm pretty sure I have that exact same trackball and mouse plugged into my computer right now.
i am in need of an extra laptop /j
i switch between mice but i'm on a razer basilisk rn, it's ergonomic and relatively light
oh where are you in the world
Logitech MX Ego, and a SteelSeries 300 Sensei or somesuch.
if you're in the midwestern us, bet
? me?
yes you who needs a puter
I pretty much use the trackball exclusively, the mouse is only for gaming.
the trackball talk makes me wanna whip my magic trackpad back out and start dailying that again
does California count
:clueless:
i know a guy
meaning...?
a guy who owns far more laptops than me
laptop god
true
lol
where he at though
LA
(btw we should prob move to dms or smthn)
yeah probably
noice
brb actually
to change the subject again, here's the last two revs of my desk setup, i wanna bring the google home back but that spot was best for ethernet for the pi and the short ass charger for the pi and i kinda need the pi rn
hmm
hey should we add gparted to the live ISO so that we can clean up a disk before running readymade?
this reminds me
does it support manual partitioning
I don't think so
if you mean readymade?
whatever's the installer program
I forgot
I think eventually
nice
I'm assuming rn is just full disk format
which also begs the question
what's the disk layout in that case?
yeah
not sure offhand
on um it's 500mb /boot/efi, 1gb /boot, remainder btrfs
Last time I tested on Ultramarine yes, it'll just pull an installed GUI disk program (including gparted). I'm not sure if the bluefin config is different but the code should be all there
alright we'll work on that next after we slay this accursed busy device thing
Including gparted in the Live ISO would be best imo (we are also considering this)
thinking of the way VanillaOS's installer handles manual partitioning
did someone find rdm working
latest
everyone having the repart issue?
I'll announce a new round on the discord
fail for one guy but he was on the wrong nvidia card
ah…!?
Universal Blue
Call for Testing: New Bluefin ISOs!
Ok! ISOs refreshed, time for round #2! We know the links are still placeholders, don’t worry about that, please focus on: Installation on bare metal if possible Check and test secureboot Nvidia people please report in, live session should be accelerated If you have another disk in your PC we think we got the os-prober thing, please confirm.
new report!
@wolfyreload in here!
Sorry, didn't know about this thread
oh don't worry, appreciate the testing!
The new live ISO image is looking really promising so far, like 95% there imo
yeah it's so fast
Failure, just a new vm created in virt-manager with the unversioned Fedora defaults and an 80gb vhd. Selected entire disk.
no BIOS support
v
@j0rge mind putting this on the thread too?
we got this like 4 times now 😭
no idea how to even fix it to be completely honest
Understandable, probably at least needs to throw a warning about it though.
yup.
Rebuilt with EFI and giving it another shot
it seems to have installed perfectly fine on ur vm
its just failing to install the bootloader cuz it doesnt support BIOS
I assume all the
No localization for id: stuff is just placeholder for nowyup
should be a quick fix but we havent done it yet
the images are also a quick-ish fix
I would say don't
Don't fix it
It's either a user with a misconfigured vm, a user with a misconfigured uefi, or a user with hardware so ancient we wouldn't want them to be able to use our shit anyway
i think we should have a page or something
just to say "YO you cant use BIOS"
Yep
But don't fix it
real btw
That's what we do over here.
Zero BIOS support, Not gonna happen, and if you do manage to hack your way around it, no support.
based tbh
legacy hardware support just gets harder and harder as software moves on.
Yea, I wasn't at all arguing for support, just a warning that it wasn't.
may be smart to keep the anaconda images as bluefin-legacy for advanced use cases and bios users
?
nope
anaconda is trash and supporting bios is annoying
Would it be possible to block the install from starting if the system is not using EFI? I know you can check it with
[ -d /sys/firmware/efi ] && echo UEFI || echo BIOSand readymade does have support for bios generally just not bios+bootc
its the repart formatting
where? which issue is it?
bootc install's "bootloader install" step fails to run because our formatting config in repart doesnt have a hybrid MBR format, so it fails on BIOS systems
i dont know if theres an issue for thisGitHub
Virtual Machine failed because of missing BIOS support on repart co...
Installing Bluefin ISO fails in VM. Installing in a VM on my daily-driver Bluefin install. 0: �[91mReadymade subprocess failed�[0m Location: �[35msrc/backend/install.rs�[0m:�[35m190�[0m �[96mNote�[...
found it! ill just add more contexts
Oh
we don't support bios systems we should just say that
ok
yup LOL
we can put that on the website
bios really only going to be seen on defaults for virt-manager
yeah
Btw tested the GTS image too. No issues. Updated my report https://universal-blue.discourse.group/t/call-for-testing-new-bluefin-isos/7981/81 Is there anything else I can test?
ok so assuming gts worked just fine because your virt-manager switched over to uefi?
Seems there's no partition management yet?
And yea installs ran fine after switching it to uefi
This can be reproduced by just running bootupctl backend install -vvvv with the generic image argument, itll try to install the BIOS versionso the issue should probably be "say bios isn't supported"?
oh i just added a few details on what is going on
we can/should just rename the issue to that yeah
So I'm thinking since we have backups of the anaconda one
add gts and the new ISOs now?
that repart issue is way too prominent still :(
yeah but no one is testing it
our problem is we're not casting a wide enough net of testers
i think best way to figure out a fix is having a nice, reproducible way to get this bug to happen all the time
i just dont know exactly what causes it
cuz like most of the issues found in testing are resolved now afaik (or at least catalogued on the readymade repo)
Which?
it's a… bug that I don't know how and why it happens, it doesn't affect anything but it's just there
that one is cuz we did something wrong
on the bluefin config
nothing to worry about
ah ok
ill just need to change a thing in the readymade package
also an update on the repart issue
it works
like, it makes the partitions
but it fails because "device is busy" after that
this is the closest thing we have to it
GitHub
Failed installation leaves partitions on disk · Issue #86 · FyraL...
After the installation fails it leaves these partitions on the disk, there's no way for me to remove them when trying to do a reinstallation.
can we just do #1 to make it simpler?
can someone list the mounts
lsblk
its the normal mounts, the
/tmp/* stuff
for /dev/sda1, /dev/sda2, /dev/sda3
it just fails after the last step
when repart tells the kernel to re-read the kernel partition table!?????
yup.
so why does it happen so inconsistently consistently
i honestly think this is an issue on repart itself
maybe we can check if the output has "Telling the kernel to reread the partition table" and not fail when that shows up
cuz im pretty sure the partitions work when we get to that message
oh shit @madomado repart has json output
we can parse that then maybe?
yes
we use that json output
oh we already do?
yeah
hold
shit
we never logged the stdout when it fails
mmm?
but like it shows up fine on the logs
that the device is busy blablabla
yeah that's stderr
we don't capture stderr in the subprocess
and of course there are no docs on the json format
a;sdlkfjzpxcvupoqwe
wonderful LMFAO
maybe just dump literally everything to the logs?
@wolfyreload that totally worked!
blew away the partitions
oooo what?
rebooted, install went fine
what
Ok, install of bluefin-live-stable-amd64.iso went off without a hitch on my hardware
wait
what
so what's the
issue
zjcvpioq[wet
install, reboot, setup, desktop, no glitches.
that's on a beelink n100 based NUC
wow nice!
uh
so does that mean
no one can repro the issue now
doesnt mean anything like
the repart one
lemme test the GTS image
its just sometimes the device blows up
ok idea
@j0rge do you still get the repart issue if you try to re-install it now?
like start over now that it's installed?
I can try it
btw
[main 04c6382] feat: record the repart output when fail
oh ok so we should refresh ISOs then just in case it blows up?
also @madomado what are the args we send to repart?
do we need that mount into
/run/readymade?wait what mount
wasnt there a mount on
/run/readymade?…?
lemme find it
no?
I don't remember
i think in some part of the code theres some modifications to the repart templates
GitHub
readymade/src/backend/install.rs at main · FyraLabs/readymade
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
i wonder if that fucks it up
well if that fucked it up then it would fuck it up consistently
not like now
true
maybe the
--offline mode is weird?no
only --offline works
ok so
im pretty sure this repart issue thing is going to happen consistently with windows dual boots
ill try to install windows on a vm and repro this issue
GTS image installs fine on the same hardware. Secure boot works with both images (other than needing to do the MOK enroll on first boot)
at least on my hardware.
So tl;dr
WORKSFORME
that is in a single drive, single boot configuration, just to be clear.huh
maybe I am just disproportionally running into it?
What are you doing to it to make it angry?
did you call it's mom fat or something?
I would love it if it was just me because then we could ship. 😄
Oh, I certainly get that.
sorry your software didn't fall down and eat itself when I tried to install it.
I hate it when I install something and it just works.
ping if someone can repro with the new rdm build
well! reinstalling the nuc with the old rdm build worked fine
@tulip can we fix the bento box strings? I'm thinking to add the GTS ISOs on the website
oh yup yup
like, we need to get more testing, I'll just add a note that they're in beta on the website
then that'll at least get us more eyeballs
but we can keep the stable ones anaconda on the website, but it would at least move us forward
then I'll make a list of the old ISOs on the website and we can do "if this install fails try these"
Oh this is a fun one, if you boot the ISO using ventoy and you accidently click on ventoy drive instead of the drive that you meant to install the ISO, then it nukes my USB and then fails with the error that you were getting jOrge 😂
can you file that one on the readymade repo?
I found a few as well
oh what?
wait so thats good! we have a more-reproducible way of getting that error LOL
GitHub
GitHub - FyraLabs/readymade: Install ready-made distribution images!
Install ready-made distribution images! Contribute to FyraLabs/readymade development by creating an account on GitHub.
so it doesn't detect the root drive properly? want a lsblk
@Cappy Ishihara @madomado would any of you be against not having
/usr/share/readymade be a part of the readymade package?no
each consumer (including you guys) have like different defaults and stuff
ack
so lets just %ghost that?
ill PR it
usr share doc
hmm?
we can put the example in /usr/share/doc/readymade
no need
just leave it be i think
Here are the
lsblk results when booting Bluefin using Ventoy Grub2 boot option
And it has "Ultra USB 3.0 (sda)" as one of the install locations which is the USB device that I'm installing fromwhats the lsblk when you do the funny thing?
cursed how it's not mounted as /
very cursed
looks the same
attached the full log which looks just like jOrge's log (I think)
Weirdly after shutdown and checking the USB on another device, it's definately writen the new partitions
use the latest rdm build to get the json output maybe?
i can send it to you @wolfyreload
also
this means we can finally report to systemd with a good repro
@wolfyreload
curl -fsSL https://github.com/magic-wormhole/magic-wormhole.rs/releases/download/0.7.6/magic-wormhole-cli-x86_64-unknown-linux-gnu.tgz | tar xzvf -oh lemme spin new ISOs with the updated rdm
tulip if you have anything to merge
now is the time
nothing
should be fine
This looks like malware 😂 looking at the name
./wormhole-rs receive 50-orlando-revenge
LOL its wormhole-rs its really goodGitHub
GitHub - magic-wormhole/magic-wormhole.rs: Rust implementation of M...
Rust implementation of Magic Wormhole, with new features and enhancements - magic-wormhole/magic-wormhole.rs
Probably better than copying the output to a file and then doing
cat /path | fpaste 😂oh i was gonna send you the latest readymade build
.
omg true
LOL
btw dnf in croc
Everyone just install this to zap logs to each other: https://flathub.org/apps/app.drey.Warp
i think we dont have it on the ISO
thats why i sent the wormhole-rs one
that thing is amazing tho
maybe we should for testing heh
I added gparted last run too
pretty good! thinking about defaulting to that for custom partitioning n stuff later
I mean it should be as simple as adding like a 1mb empty MBR partition
how do we do it with repart 😭
not super familiar with repart tbh
https://www.freedesktop.org/software/systemd/man/latest/repart.d.html#
I don't see a type
Additionally, I think we shouldn't make a hybrid partition table. Ideally, would be a different config.
I also think we really shouldn't support bios. The only use case that people will run into are poorly configured VMs.
I think we should be uefi only as well
Hm?
I don't see a bios boot type
Ah, why do we need one?
I think not supporting BIOS is a feature
That's my view as well.
Not supporting BIOS is absolutely a feature.
how are we confused did someone say we should do this?
I mean, I'm sure "somebody" did, but who is "somebody"
i just asked how we could do that
but its not something we should do yeah
https://docs.projectbluefin.io/downloads/
ok readyish
Btw the "runbook" link on the bluefin downloads page links to pager duty documentation, guessing it's just a placeholder for now?
Another VM install succesfull. Might test on a bare metal this weekend
Just did a full bluefin install on baremetal, worked absolutely fine 😄
thanks for the sponsor btw
@tulip told me to do it
@inffy, nope worked for me
yeah started to work now, dunno what was the issue
fluke
where's your sponsor info we need to put you on our donation pages
the heck just happened
thank you wtf
I put you in the donation page for bluefin too
how's that?
perfect.
also i think robert has a gh sponsors now
oh nvm, my bad
i think i confused him with somebody else
GitHub
chore: get rid of the install script entirely by tulilirockz · Pul...
This will let downstreams create packages for their custom stuff under /usr/share/readymade, as we dont need
the templates or anything that the install script does really
i was thinking we do that, then make a separate package for ultramarine branding/translations n stuff
(and bluefin would make its own package for their thingies)
we too have a sponsors but please sponsor mado <3
I did. 😄
lol nice
mado deserves everything he's so cracked man
failed
I think you can keep install.sh, you can just modify the file lists in the spec
oh! i just saw the "config-ultramarine" subpackage!
so you guys already made the split thing
yeah
ok so it should be fine then?
no, uh
I forgot to remove some files from the main package
some files are listed in both
i think the main package just needs to not have the /usr/share/readymade thing
yes
and /etc/readymade.toml
okok ill change my pr to just that
that seems to already not be there
ah ok
done!
this package split is what we need for the bento boxes?
I'm ready to ship gts ISOs asap if we can get the strings fixed
yup!!!
Ready!
ill just make a new package for shared bluefin readymade configs
should be quick!
im on my gf rn but ill be back home soon
Kk
Universal Blue
Call for Testing: New Bluefin ISOs!
When I open and close the installer the console prints: Errors [ResolverError(Reference(Variable { id: “distro” }))]. at /github/home/rpmbuild/BUILD/readymade-nightly-20250430.b8adbe7-build/readymade-b8adbe782eeb67a456714fd1ee66340e6b9bf04a/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/i18n-embed-0.15.4/src/fluent.rs:170 followed ...
I don't think we've seen this one yet?
ok @j0rge this is fine now
the descriptions
love the little mado still there
heh yeah
omg i closed the loop!
@j0rge @jade
close.
i love No localization for id: "page-installation-docs"!
this is actually going to look so gourgeous
heh yeah didnt override that
this looks pretty good imho
definitely not perfect tho
looks great
there might have some way to make the css like this a bit more
i wanted it to be kinda like just some crops so it doesnt look like distinct images
WOW
THIS IS SO BLINGED OUT
cc @j0rge @Kyle Gospo @Niklas ⚡
holy crap that looks so cool
so heliumed
it looks gourgeous when the screen is maximized
like man this is amazing
im just fixing the
po stuff firstBtw do you mind if I make a bluefin Youtube video showing off the new installer or are there more things you want to fix in the installer still?
after the branding is done and is on the ISOs, i think it should be fine yeah!
if you wanna showcase everything you might want to wait for this pull request: https://github.com/FyraLabs/readymade/pull/88
then we'll get dualboot/partitioning support
tpm2 still needs a bit of wiring but everything looks great
I might wait for that, then I can do a full on guide
Granted... the new installer is so simple that it probably doesn't need a video guide 😂
i just think the dual boot / custom partitioning part will need a guide :)
its a bit confusing for new peeps
That's true, I will need to put in a disclaimer for that, that it's an unsupported configuration
at least for bluefin
the default layout is more than enough for 99.9% of peeps on bluefin
I'll probably need to make a video on how to remove Windows without re-installing bluefin/bazzite/aurora as that's quite a common scenario
we include gparted on the ISOs now btw
I saw. Now I can recommend just using bluefin and not linux mint for fixing stuff 🙂
ok so
https://github.com/ublue-os/packages/pull/403 and related PRs are all up
GitHub
feat: add bluefin readymade configs for live ISOs by tulilirockz ·...
This will let us auto configure a ton of stuff on the live ISOs, also makes for a good template for other configurations
tagged on that pr
we just need a merging spree and the ISOs are done
Incredibly based
need to also add that new package to the copr
Fishing for acks?
fishing for acks.
Is there any way to get that nice user creation on first boot that you have in Gnome/Bluefin in KDE?
GitHub
GitHub - Ultramarine-Linux/taidan: Out-Of-Box-Experience (OOBE) and...
Out-Of-Box-Experience (OOBE) and Welcome App. Contribute to Ultramarine-Linux/taidan development by creating an account on GitHub.
we are planning on using that!
no screenshots on the repo tho
it looks just like readymade
but works like gnome-inital-setup
Oh I got you at least 2 of those
Oh nvm 💀
lets just get bluefin out
and ill make an aurora branding package thing
after aurora and bluefin we can go to bazzite
order of difficulty to support LOL
Aurora will be interesting with KDE
honestly shouldnt be hard
the live ISO will be probably just fine
hard part is taidan
Yeah I'm more worried about the oobe
it needs to be on terra, and get the un-hardcoding treatment we did to readymade
at least one thing i can think of is the gresources fix
even if it wouldn't be on terra, couldn't we just explicitly install it from the repo
yup!
its just that theres a lot of ultramarine stuff hardcoded still
right
i also like how renovate approves itself, its so fascinating to watch 😭
taidan needs some porting from rdm
big renovation
in addition I want to make the pages runtime-defined instead of hard coded
we heliumed the ublue
Oooh wow
Looks great
idk why but custom is failing with device or resource busy too
fail on this line
that sounds based
new ISOs spinnin
I'm confusion
what's up
DEVICE OR RESOURCE BUSY
AAAAAAAAAAAAAAAAA
it's the new "unable to decrypt message"
on the plus side, we got a bunch of other stuff fixed
omg I think it's working
wat.
I'm using UEFI????
Hi there, just wanted to say thanks and congrats to the new isos, been testing on 8th gen i5 laptop and an old 2nd gen i5 laptop, two old gaming desktops with Nvidia cards in, all on bluefin:latest and bluefin-nvidia:latest, all working great, been updating since this morning without issues and rebased to nvidia
so on 8th gen i5 laptop ive got windows on primary ssd and bluefin on second ssd, but after a reboot on bluefin at 15:08 today I cant seem boot into bluefin, any thoughts on what might caused the grub to be inactive or removed?
if you can't boot into bluefin what are you booting into
overwritten with the windows bootloader?
could be, there was windows updates installed and checkdisk before I got into windows
well you've answered your question probably
could be that the windows 11 update nuked the grub for fedora
might need to just reinstall bluefin but otherwise all good on my en testing so far
just tested bluefin latest iso on my end
I'm confusion
@Steven Terblanche theres a chance that the readymade build you had was broken
because sometimes that thing didnt install
/boot for whatever reason
have you seen the new bento branding?
@j0rgeoh yeah
also i think the images (1, 2, 3 ,4) are just straight up not necessary 😭
hmm?
i just edited the gresources thing
those image files under
/usr/share/readymade/bento/[1-4]* dont seem to do anythingthey do nothing
I wanted them to do the thing
but uhh
doesn't work
so yeah
i ended up just making a gresource on the bluefin branding package thing
Are there new ISOs for another round of testing?
I gave up
I'll let someone else figure out
just making sure that the latest ISOs even have readymade on them
use another codec tulip 😭
yup.
its the gnome thing
the default gnome recording tool
Uh, the one I just downloaded seems to not
wut
oh yup yeah they dont have it LMFAO oh no
ahhh ok. but on my other 3 machines bluefin boots fine
also
sudo dnf install -y readymadeI clicked the action, how is it not on the image?
there was a conflict w/ the branding package cuz the readymade-nightly package we have on terra doesnt have the fix i made yesterday
ah ok
ill just do some fuckery on the install script for now
well ready to respin!
nono just a sec
else it wont install it
k
also entirely unrelated but this system is so much faster than the one I was using to test it on yesterday 😂
were you experiencing that "device is busy" thing on the other system?
No, but I was only doing it in VM
btw the custom bootc pr thingy
why da binary
nyes
noice.
but windows and bluefin are not on the same disk
yeah it should work fine
i think it was an installation failure tho
like on bluefin
windows bootloader is still there but fedora is gone from the bios
@tulip ok ready?
oh i just pushed the button :)
lets see if its ok
ah ok
even better
isnt there a way to repair the boot with the live iso maybe?
i dont think-
maybe?
actually yeah you can probably do that
otherwise ill reinstall, no probs
but how will I do that tho? like regenerate the grub
basically boot into the live ISO, then open a terminal and:
just change sda to the device you have bluefin on
@wolfyreload maybe a good thing to write down
thanks, will try asap as I am in windows now for work things , got another windows update pending so fingers crossed lol
yeah mine is sda
nice.
so this all should work fine
Is that for manual partitioning and installing without the UI?
this is for re-installing grub if windows blew it up
ha ha ha, I read
bootupctl backend install as bootc backend install. That is much simplier than the other way's I've seen people do itI see ubuntu actually installs their grub bootloader inside the windows boot manager directory , from like the 25.04 install I tested yesterday
but the way bluefin does updates and the whole rebase and docker thing is just insane , always come back to bluefin
like I main bluefin on all my machines here at home, even use it daily for remote work as well from home, just wanted brag and say thanks 🙂
its great im a big fan :)
gaming machines also, smoother than bazzite to me
@j0rge its looking so gourgeous on the latest build
oh my god
yes!!!!
ok one last adjustment on the bentos, pan the camera "up" a bit
sounds good!
GitHub
Switching to Readymade · Issue #66 · ublue-os/titanoboa
Todo Fyra rm hardcoded branding move Readymade to Terra Split config into packages so you can install readymade-config-ublue or whatever UBlue add Readymade to image config enable GNOME OOBE Relate...
or just follow this since those highlight the cool dinos
like if we can get the trike in there
but don't block anything
am I ready to test gts?
i think so!
can you adjust the website PR to only add GTS?
okok
leave the filename alone for now we can fix all the filenames later
5.8GB on bluefin:gts
beautiful
so getting this , in the bluefin latest live iso now
sudo mount /dev/sda3 /mnt
sudo mount /dev/sda2 /mnt/boot
sudo mount /dev/sda1 /mnt/boot/efi
sudo bootupctl backend install -vvvv --auto --write-uuid --update-firmware --device /dev/sda /mnt
mount: /var/mnt/boot/efi: can't read superblock on /dev/sda1.
dmesg(1) may have more information after failed mount system call.
[TRACE bootupd] executing cli
error: boot data installation failed: failed to install, invalid re-install attempted: /mnt/boot/bootupd-state.json already exists
huh.
so maybe rm -f /mnt/boot/bootupd-state.json
sudo mount /dev/sda3 /mnt
sudo mount /dev/sda2 /mnt/boot
sudo mount /dev/sda1 /mnt/boot/efi
sudo bootupctl backend install -vvvv --auto --write-uuid --update-firmware --device /dev/sda /mnt
mount: /var/mnt/boot/efi: can't read superblock on /dev/sda1.
dmesg(1) may have more information after failed mount system call.
[TRACE bootupd] executing cli
[INFO bootupd::bootupd] System boot method: EFI
[DEBUG bootupd::efi] Found metadata grub2-efi-ia32-1:2.12-28.fc42.x86_64,grub2-efi-x64-1:2.12-28.fc42.x86_64,shim-ia32-15.8-3.x86_64,shim-x64-15.8-3.x86_64
mount: /var/mnt/boot/efi: can't read superblock on /dev/sda1.
dmesg(1) may have more information after failed mount system call.
[DEBUG bootupd::efi] Unmounting
error: boot data installation failed: installing component EFI: Failed to mount "/dev/disk/by-partlabel/EFI-SYSTEM"
ok ur thing is cooked
might want to straight up reinstall
so did a file system repair now in gnome disks on /dev/sda1
and ran sudo mount /dev/sda3 /mnt
sudo mount /dev/sda2 /mnt/boot
sudo mount /dev/sda1 /mnt/boot/efi
sudo bootupctl backend install -vvvv --auto --write-uuid --update-firmware --device /dev/sda /mnt
mount: /var/mnt/boot/efi: can't read superblock on /dev/sda1.
then got all of this
your /dev/sda1… might be fried?
yeah repaired the filesystem now, rebooting
[TRACE bootupd] executing cli
[INFO bootupd::bootupd] System boot method: EFI
[DEBUG bootupd::efi] Found metadata grub2-efi-ia32-1:2.12-28.fc42.x86_64,grub2-efi-x64-1:2.12-28.fc42.x86_64,shim-ia32-15.8-3.x86_64,shim-x64-15.8-3.x86_64
[DEBUG bootupd::efi] Reusing existing "/mnt/boot/efi"
[DEBUG bootupd::efi] Get product name: Fedora
[DEBUG bootupd::efi] Creating new EFI boot entry using 'Fedora ' efibootmgr: Warning : Boot0004 has same label Fedora BootCurrent: 0001 Timeout: 2 seconds BootOrder: 0005,0004,0002,0000,0003,0001 Boot0000* Linux Firmware Updater HD(1,GPT,cd9489ef-409f-4378-92ca-29834f82983d,0x800,0x12c000)/\EFI\fedora\shimx64.efi5c00660077007500700064007800360034002e006500660069000000 Boot0001* UEFI 7A63CC174A1F04BC PciRoot(0x0)/Pci(0x14,0x0)/USB(12,0)/USB(1,0)/HD(2,MBR,0x1d4d273e,0x3a88000,0x10000)/\EFI\Boot\BootX64.efi{auto_created_boot_option} Boot0002* Windows Boot Manager HD(1,GPT,8a1a3abd-0b75-4343-b00c-14c879e37bea,0x800,0x32000)/\EFI\Microsoft\Boot\bootmgfw.efi57494e444f5753000100000088000000780000004200430044004f0042004a004500430054003d007b00390064006500610038003600320063002d0035006300640064002d0034006500370030002d0061006300630031002d006600330032006200330034003400640034003700390035007d00000000000100000010000000040000007fff0400 Boot0003* UEFI PM991 NVMe Samsung 256GB S50ANF0N125716 1 PciRoot(0x0)/Pci(0x1d,0x4)/Pci(0x0,0x0)/NVMe(0x1,00-00-00-00-00-00-00-00)/HD(1,GPT,8a1a3abd-0b75-4343-b00c-14c879e37bea,0x800,0x32000)/\EFI\Boot\BootX64.efi{auto_created_boot_option} Boot0004* Fedora HD(1,GPT,9f40a58d-d6ba-486b-a202-2096956f1cfd,0x800,0x100000)/\EFI\fedora\shimx64.efi Boot0005* Fedora HD(1,GPT,9f40a58d-d6ba-486b-a202-2096956f1cfd,0x800,0x100000)/\EFI\fedora\shimx64.efi [INFO bootupd::bootupd] Installed EFI grub2-efi-ia32-1:2.12-28.fc42.x86_64,grub2-efi-x64-1:2.12-28.fc42.x86_64,shim-ia32-15.8-3.x86_64,shim-x64-15.8-3.x86_64 [DEBUG bootupd::grubconfigs] root_dev=72 boot_dev=2050 Installed: grub.cfg [DEBUG bootupd::grubconfigs] vendordir="fedora" Installed: "fedora/grub.cfg" [DEBUG bootupd::efi] Unmounting seems to have worked
[DEBUG bootupd::efi] Creating new EFI boot entry using 'Fedora ' efibootmgr: Warning : Boot0004 has same label Fedora BootCurrent: 0001 Timeout: 2 seconds BootOrder: 0005,0004,0002,0000,0003,0001 Boot0000* Linux Firmware Updater HD(1,GPT,cd9489ef-409f-4378-92ca-29834f82983d,0x800,0x12c000)/\EFI\fedora\shimx64.efi5c00660077007500700064007800360034002e006500660069000000 Boot0001* UEFI 7A63CC174A1F04BC PciRoot(0x0)/Pci(0x14,0x0)/USB(12,0)/USB(1,0)/HD(2,MBR,0x1d4d273e,0x3a88000,0x10000)/\EFI\Boot\BootX64.efi{auto_created_boot_option} Boot0002* Windows Boot Manager HD(1,GPT,8a1a3abd-0b75-4343-b00c-14c879e37bea,0x800,0x32000)/\EFI\Microsoft\Boot\bootmgfw.efi57494e444f5753000100000088000000780000004200430044004f0042004a004500430054003d007b00390064006500610038003600320063002d0035006300640064002d0034006500370030002d0061006300630031002d006600330032006200330034003400640034003700390035007d00000000000100000010000000040000007fff0400 Boot0003* UEFI PM991 NVMe Samsung 256GB S50ANF0N125716 1 PciRoot(0x0)/Pci(0x1d,0x4)/Pci(0x0,0x0)/NVMe(0x1,00-00-00-00-00-00-00-00)/HD(1,GPT,8a1a3abd-0b75-4343-b00c-14c879e37bea,0x800,0x32000)/\EFI\Boot\BootX64.efi{auto_created_boot_option} Boot0004* Fedora HD(1,GPT,9f40a58d-d6ba-486b-a202-2096956f1cfd,0x800,0x100000)/\EFI\fedora\shimx64.efi Boot0005* Fedora HD(1,GPT,9f40a58d-d6ba-486b-a202-2096956f1cfd,0x800,0x100000)/\EFI\fedora\shimx64.efi [INFO bootupd::bootupd] Installed EFI grub2-efi-ia32-1:2.12-28.fc42.x86_64,grub2-efi-x64-1:2.12-28.fc42.x86_64,shim-ia32-15.8-3.x86_64,shim-x64-15.8-3.x86_64 [DEBUG bootupd::grubconfigs] root_dev=72 boot_dev=2050 Installed: grub.cfg [DEBUG bootupd::grubconfigs] vendordir="fedora" Installed: "fedora/grub.cfg" [DEBUG bootupd::efi] Unmounting seems to have worked
nice.
wow tulip this looks amazing yeah let's ship GTS ISOs
hahahaha
LES GOOOO
bentos are rocking and rolling
working amazing too
loop is closed there
yep tested here
new firefox looking swank
we can change the images however we want
also suggestion
you may wanna suggest wiping the disk with gnome disks and trying again if installation fails
it solved a lot of my issues early on when i installed this copy of bluefin
yeah tulip after you land the picker let's work on the wording
there's things we need to add there
we can also drop the DX toggle in the file picker
https://github.com/ublue-os/titanoboa/issues/96 string updates
https://github.com/ublue-os/titanoboa/issues/98 ok I split this one out so it's more clear
just an update, this worked, booted straight into bluefin
secure boot working as well
nice!
/dev/sda1 filesystem had to be repaired
cc @wolfyreload
It would be nice if ready made could also create users
But that's beyond the mvp we're going for
yeah we'll need that for dx
On a handheld rebooting multiple times, especially having the first boot be straight to the desktop, isn't a great experience
We need a default user though
if anyone has repart skills
It can be blank, here's the cool part
During install we have a live ISO environment, and I can enable the stock virtual keyboard during that environment
why would taidan not work? would that be bad for you guys?
The problem is if we were going to go with Taiden, once you reboot your back into a normal bazzite flow
Which means steam provides the keyboard
And if we're booting straight to the desktop on the first time so Taiden can work
You need to log into steam to have a keyboard, which requires a keyboard
oh i dont think you'd need that, right? we can make taidan boot directly to the steam gamemode
Now you're stranded
We can't do that because you need a user for a game mode to work
whatever, we can figure this out
Just assume there's no future for Taiden on bazzite
It's not happening
ye, so it would be: taidan -> new user
No that would require a keyboard
We won't have one after initial install
OH
Not until steam is logged in
OOOH understood
didnt know
but like how do you expect readymade to have the keyboard then?
Yeah it's fine, this is why I'm so strict on MVP
We can worry about this after Bluefin works
mmmhm
Caribou/Maliit
its already working, we are gonna ship GTS ISOs w/ it now
I'll just turn them on for the live ISO, same way we turn off steam
no chance we couldnt use that on taidan?
like im just pushing for it cuz it seems like the best solution
but whatever works
No chance
Thank you :). I'll play around with that in a VM.
Wdym it can be blank?
We'll have a keyboard during install
Users can just type a password and username
LiveISO magic
Ah in readymade
Kk
Yes
i was thinking about also adding the keyboard layout to the kargs
so luks will get the right keyboard layout right away
i think the bluefin usecase is like 99.9% done
KDE and bazzite in general will need some work
hi kyle
no taidan in ublue?
no idea
bazzite itself will absolutely 100% need the username/password thing
Just won't work for bazzite deck
i do think that it would be nice for desktop KDE tho
I think the only requirement is just a11y support right?
the on-screen keyboard part
No it's genuinely just a bad experience to have to reboot three times to install
And I would need to make some crazy workarounds to make the stock keyboards work on the first boot after install
you dont need to reboot after taidan tho
why?
It's not something I'd even be interested in attempting
I mean why 3 times
thats a big problemo yeah
Install, reboot, boot to desktop for Taiden, now you have no on screen keyboard because steam isn't logged in and you can't log into steam because you don't have an on-screen keyboard. You're now fucked
Then reboot again after finding a keyboard to finish the install
Then you're brought to game mode, it updates, you reboot again
Extremely bad experience
The installer needs to do the user, there's just no other option
I'm confused about the steam part
Steam will not give you a virtual keyboard until it's been logged into
you need to be logged in to have the steam keyboard
anyway, jade insists on not adding the user mechanism into rdm
ye
You cannot boot to Game Mode where the option to log into steam with a keyboard is presented because you don't yet have a user
Chicken and egg
is steam keyboard the only… keyboard?
Yes
i tihnk we can get away with not having it for everything that is not bazzite-deck
alright then
this is very weird btw i wonder why that is a thing
?
steam keyboard being the only keyboard on bazzite-deck
uh
well I'm kinda ignorant on the bazzite stuff soooo yeah
same tbh
if anyone can repro #59, please send the json
Ok so, can this thing create a separate EFI partition?
custom partitioning
yup
Maybe we can improve dual boot compatibility thanks to that
the partitioning is done on gparted
or whatever else
Had someone getting their install nuked by windows like rn
By simply switching to it briefly
Oh so not by systemd-repart?
for the custom partitioning yeah
its literally just gparted
or gnome disks and whatever
Uff
honestly thats the best solution we could come up with
bilvet sucks very much and making a partitioning ui sucks
No I mean like, at least if there is enough unpartitioned space, systemd-repart that space
its not that easy tho
if you can figure it out it would be awesome
Wdym is not that easy?
how do you even start do to that
I mean probably won't be able to avoid gparted
Start?
like, how would you find empty space, then repart that space with the layout we want
?
like how do you implement that
Isn't that what systemd-repart already does?
i dont think so
It places the partitions in the free space
we wipe the disks n stuff with it
afaik it just expands partitions
Yeah because you use the --wipe flag if I remember
that's
dualboot (wip and will take quite a while to get remotely close to finish), not custom (the thing that I'm trying to add support right now)
Oh kk
blocked?
oh no..
for dualboot, uhhh lea is kinda not here and I think there was code that didn't get pushed
invalid argument again
yeah
OH
maybe its the EFI-SYSTEM thing
wait
is it not EFI-SYSTEM?
i mean, the /boot/efi partition needs to be EFI-SYSTEM
is it that?
I think so…?
I regret giving 8 GiBs of ram to bluefin
why tho LOL
restore and saving taking forever
seems to be correct
also 8 GiBs is too much and that makes me unable to open readymade code and the vm at the same time
wheres the label tho
bruh you need more ram
label?
it's 16 already
mkfs.fat -L EFI-SYSTEMI didn't anticipate
oh
oh wait that's needed?
but repart
oh yeah
I forgot
about that
LOL
Forgor
-L is invalid arg
?
ok it's -n
i forgor
nice yeah thats it
needs to be
-F32 or something i dont rememberyeah I did
I'm not forgetting -F32 for the rest of my life
also bootupd is goated
you should use that for ur grub installs on regular fedora
it feels brokey tho
its just kinda crap cuz of the hardcoding everywhere
but i think fedora already uses it for updating bootloaders
its surely a lot easier than grub2-mkconfig
what
nah cap
what what?
nahhhh cappp
no goddamn way
whatcha talking about
grub2-mkconfig should be easier
I'm not convinced
at least on bootc we dont have that option
but bootupd just plops the file into boot, w/o any modifications to the grub config or anything
they just use that grub2-static config from coreOS
so no fiddling with the configs
well
updating is also neater
how should I put this
bootupctl adopt-and-updatefedora is creating a solution that only they need because it's addressing an issue that they've created
yup, fedora skill issue tbh
systemd-boot would be so goated if we could use it...
(i think we cant do it on atomic rn)
because ideally the workflow goes like installing grub, write /etc/default/grub, grub-mkconfig -o /boot/grub/grub.cfg, and then grub-install
the reason why this doesn't work on Fedora is uhhhhhhhhhhhh… random weird stuff to get shim going on, and then some random idk uh
fedora has a nasssty grub implementation
blscfg
right
no idea why they have that thing
I don't understand like what's going on
glad they are moving to UKIs
w/ bootupd?
both
whatcha doing then
I just
let fedora cook their stuff
I don
I don't understand how it works
blsconfig is ridiculous that thing shouldnt exist
bootupd effectively just does
cp /usr/share/bootupd/* /bootUKIs are not really solving the issue imo
less grub fuckery
and more bootloader agnostic
the main fuckery is blscfg
so systemd-boot
and I like
yeah
they're not getting rid of that
gonna keep bringing up that readymade is designed primarily for oem-style installs, if you want to create a generic bazzite:bazzite user you could do so in a post script, otherwise an oobe app is needed
it boots into kwin with only taidan loaded
we can provide a keyboard
Yeah that's not going to work
We'll have to fork it then or just not use it
they want the steam keyboard
I don't know what that is and I don't know why but yeah
steam keyboard IS good
not important for now
why
we can do the same thing in taidan
but
how do you even press buttons they just evaporate?????
conversation for later
You can't because I uninstall them
controller navigable
They conflict with the steam keyboard, only the live ISO for the initial install will have them
Which means the live ISO during install needs to be the thing that creates the user
oh golly gee it's fun with conflicts time!
it's hot steam the funny steam keyboard be rising in cool air
smh
I imagine bluefin dx will also want the user made during install
Because they need custom groupings and all sorts of other things
And modifying The Gnome user Creator is out of the question
jade please argue in another thread
yes discussion for later or move to the gc
that one can be figured out
bluefin as a whole doesnt need that stuff
also the taidan ci passed btw
let's just please finish the mvp
move to fyra server
ya'll are constantly arguing about things down the road and the basic installation doesn't work
also
let's fix the basic things before expanding scope
ugh wonderful
it's literally named EFI-SYSTEM
@madomado
just put
bootupctl-brokey -vvvv as an argument so you get the verbose bootupd logs
also you can try manually installing the thing with bootupctl and see what is going on
sudo bootupctl backend install -vvvv --auto --write-uuid --update-firmware --device /dev/sda /mntwhy don't you ship it ootb
thats bazzite
oh wdym
just change the things accordingly
thats already on fedora tho
what is it trying to umount
im guessing
/boot/efi and /boot
it seems to have installed everything fine thowell
what if you use strace there
also, just making sure: you are on UEFI right?
yes
how can I install strace tho
i think... dnf?
oh it works?
ok
it does cuz theres an overlay on the ISO
doesn't help
nothing special right above
error:
seriously
it just says error right at Unmountingh my god 😭
i remember running into this i just dont know how we fixed it
you know
at this point this is a fucking bug from bootupd I swear
oh it is
always have been
let them fix it then
I don't fucking care
that's not a me issue
I'm off
thank you
okok
ill try to get it working eventually
anyway, uh
maybe you edited the templates or something?
or something something EFI-SYSTEM?
mm?
how did you even figure out EFI-SYSTEM was hardcoded
i honestly dont remember
i just saw that install function and skimmed through it
then saw "HARDCODED_EFI_THING"
you should start documenting your thought process /s
i really should tho
this stuff helps a lot
I have one suggestion
let's rebase to Arch Linux
needs fixes on bootupd if we wanna use bootc on that
funnily enough
you should be by now aware of the fact that I don't like immutability
I have the most controversial opinions in this community
most WRONG opinions :clueless:
RAAAAHH 🦅 🦅 🦅 🦅
shut
sponsorship revoked
noooooooooo
NO more money
I'm now broke ；；
👿 🍤
shrimp
what does immutability have anything to do with the installer?
what shrimp
we should be talking about this on another thread yeah LOL
>bootc
a
the LTS isos are coming in hot
fireee
🔥
alright, they're live!
we'll see how often device busy shows up now
oh boyyyyyy
2024: unable to decrypt messages
2025: devices or resources busy
readymade now blocks sleeping
I get the appeal for readymade for bluefin. Live ISO, a gig or two less ISO size. But like, anaconda works and we fixed all of the issues except the weird errors? Where device active you also seem to have?
And the dual booting edge cases are also fixed/known?
mallit has gotten better recently. IMO its better than that hot garbage
ive been using it on the Z13
lowkey its quite usable and at least its consistent
steam keyboard on my x1 launches whenever it wants does whatever it wants
uses the wrong place on the screen
like for the stuff a keyboard its needed, i find mallit better by now. When we had the previous discussion mallit was a lot worse
there's no flatpaks for anaconda installs
good, stop the flatpakening smh
It's a requirement
just install everything using zips and appimages
skill issue
Our apps are flatpaks
btw, readymade now checks if the device is mounted on the confirmation screen
Though I'm way more stuck than you are
But the current iso installs flatpaks?
Even if it's a hack dunno
Is it working for you? Ours broke
Bazzite is using the 41 ISO
It's currently broken on F42
So no f42 ISOs still?
We need a new iso soon due to the updater issues
GitHub
Add tests for 42 by noelmiller · Pull Request #161 · JasonN3/buil...
42 currently doesn't have tests. We can likely drop F40 tests once it goes out of support.
This is the current PR. I have not had the energy to work on Anaconda stuff since a lot of the momentum has been focused on the live ISOs.
I'd prefer to find a solution with Readymade. If we can get the KDE stuff sorted out, a lot of the functionality we need is there.
There are too many blockers now
You'll just end up rushing it
Bluefin has less they can intro it
I need help then. Jason has not seemed terribly interested in working on this. I'm sure if we come up with a solution to pass the tests and restore the functionality for post scripts, he would be more inclined to merge it.
What's the problem right now?
Other than the tests
Post scripts don't run post installation as they are supposed to. At least that's what I'm guessing is happening. The ISO will be created, but it won't actually switch you to the proper upstream place to get updated with rpm-ostree.
Even the MVP was a struggle I'm not sure what you should do
Sounds like we are pivoting back to anaconda.
I'm looking at the PR
The PR is just adding F42 as a build target and in turn adding it to the test suite.
I'm not ill unGA first
Valid. I'll be talking with Jade and Owen today and see what we can figure out.
Once this "Device is busy" error is figured out, the installer will be perfect.
Then if Bazzite have the appetite for switching away from Anaconda, then we can tackle the requirements.
Then if Bazzite have the appetite for switching away from Anaconda, then we can tackle the requirements.
I'm not going to recouple back to fedora I'm trying to move away from it
For anaconda right now, I am forcing the installer to be F41 and then embedding whatever image. It appears to work
But I do not like this at all
Centos? Or is the ultimate goal using something like Wolfi?
Both
We need an upstream that supports bootc first class
the funny part is that we are literally developing an entire new liveISO/installer thing
and a 10y.o method to build this stuff is struggling just as much
i am personally 1 billion percent against moving back to anaconda. we are gonna struggle a bit with the new installer but we can at least try to fix it
anaconda doesnt work, thats the thing. it has never worked and doesnt support anything we need properly
needs to be worked on, but we literally came from a "this is an idea" to "this is live right now" in like 2 months
So one of the issues is the default flatpaks not installed right? Could you add them to the end of Taidan or whatever that first setup script is?
its the anacona isos thst are screwed up
live ISOs work fine
foe that
We're not deploying anything while there are blockers, we're not struggling through anything
yeah its ok
@jade what's the correct aspect ratio for all the bento boxes?
y:x
1:4
1:2
1:2
wait wtf
wait that doesn't make sense
i think its the opposite
ok turns out I've no idea how
attach[] works1:2
1:4
1:4
x:y
1:4
1:2
1:2
yeah
width, height
as you may have seen I've definitely studied for my linear algebra class
no goddamn person on this planet puts y:x instead of x:y except for people who knows matrices
YO
@madomado
got the device is busy error
Gist
device-is-busy-report
device-is-busy-report. GitHub Gist: instantly share code, notes, and snippets.
got that error in tiffin now
weird
there are two errors here, theres a second file w/ a different error
first one i got when i installed the thing with encryption
second one happened w/o encryption, overriding the first one
Is there a way to see what is making the device busy?
Like lsof or something?
Like lsof or something?
oh shit true!
@madomado
same report as the first one i sent
tiffin is getting these device or resource busy errors
this is repart?
after repart
here
its on tiffin
Gist
device-is-busy-report
device-is-busy-report. GitHub Gist: instantly share code, notes, and snippets.
repart
there are two reports there
theres a device is busy one and one that happened on tiffin
oh
KOKUCHO10
real
LMFAO nice
thats also concerning btw we might want to hide this stuff on the logs
I'm dead
turns out repart just doesn't provide any json output when it fails
you mentioned that it was a regression
that probably isn't the case…
I just checked the master branch and none of the last 10~20 commits would've caused this
the tiffin issue
also the error makes no sense to me… repart exited successfully (so they already synced the disks), and then we synced the disks again
and then we umounted them
ah shit
umount -Rl is the issue
that issues a lazy umount, so the disk could still be in use after the umount command has exited
ok that fixed the first error
we still don't know why uh repart fails@tulip🌷 can you see if there's anything good in the gnome os installer wrt systemd-repart? https://gitlab.gnome.org/p3732/os-installer
maybe we can look around at other installers using systemd-repart?
it seems to only happen on bluefin tho
well if the repart thing is an issue we would've encounted that during um testings
actually that makes no sense, ignore what I've said
What's the plans for UM to use Readymade? Do you have a sort of timescale when you expect to be shipping it? Or are you already doing so?
timescale is unfortunate not in American's dictionaries so we will ship it when dual boot and custom support can rival anaconda
um42 in the near future is the goal
at least as a preview
without dual boot and custom
This seems to have a GUI for a first boot setup.
However, it's basically just runnign anaconda again?
hasn't been touched in 10 months?
No idea. But timothee added it to some of the base images
Genuine question: are they masochist?
Like Jesus we will never get rid of anaconda :dispair:
But I noticed it in the cosmic image timothee makes.
Also the cosmic image pulls in a huge portion of KDE because it's using ark as the archiver.....
we used to use this
well it's designed and made by the exact same people so what do you expect
Either test { $distro } from this installer or start the installation now. You can always return to this screen by selecting "Installer" in the menu.can I get some feedback on the wording
"from this installer" is confusing
what do the buttons say
did you put a shortcut to the installer on the desktop?
does gnome have desktop icons?
desktop icons?
is there a desktop icon for the installer
yes
on the dock or dash or I don't even know anymore
so no
ah
does gnome support desktop icons?
I thought it doesn't?
i think they are hidden by default
if you put a desktop icon you can remove the second sentence and simplify the first one
then change the "Try" to "Test {$distro}"
and pull it to the right
@jade thoughts?
hold on lemme rename
*rewrite
no
gnome does not support it without extensions
Try { $distro } from this drive, or start the installation now.
You can return to this screen by selecting "Install to Device" in the menu.
hold on playing with wording
i would tend towards preferring test
but regardless, same word
on the preface and the button
test implies needing to achieve something, try only implies trying it
this is just shorter but it shiuld be good
also swap "computer" for "device" per MOS
nothing guarantees the string will match the desktop file
in fact the desktop file is not translated
why not?
what
you can put localized names in a desktop file
look at any gnome app
then do it?
ask lea to do that with weblate
gonna @lea and link this to her
I'm dumb I think I bugged pretty hard
hm
Localized values for keys | Desktop Entry Specification
Keys with type localestring and iconstring may be postfixed by [LOCALE], where LOCALE is the locale type of the entry. LOCALE must be of the form lang_COUNTRY.ENCODING@MODIFIER, where _COUNTRY, .ENCODING, and @MODIFIER may be omitted. If a postfixed key occurs, the same key must be also present with…
okay looked into this
weblate doesn't support the desktop file format
however, what a lot of projects do is utilize their existing i18n system (such as gettext), then have a script which generates the localized entries
honestly you know what
I don't see a point of going back to gettext or the
lang_COUNTRY format
it has happened again
LMFAO
oh my god 😭
fix: this is not taidan
fix: this is not readymade
What's the status of the Live ISO for Aurora and Taidan? Do we want to merge this PR? https://github.com/ublue-os/aurora/pull/365
GitHub
feat: add taidan as first boot OOBE experience by tulilirockz · Pu...
This allows us to actually use Readymade as our installer on Aurora :)
Okay I started setting up the workflow and I'm testing it on my fork: https://github.com/ublue-os/aurora/compare/main...ledif:aurora:main?expand=1. Ran into a few issues.
The workflow is pushing the ISO to my personal cloudflare bucket if anyone wants to try https://pub-969fbc86b5f24e4d81c6d022e8fd8dde.r2.dev/readymade-aurora-stable-x86_64.iso
oh you need the repart configs
check out the
bluefin-readymade-config spec
it should just work fine after thatDid we merge the aurora config package
Cant remember
It's still open and marked as a draft. Is it ready?
needs graphics
Is anyone working on the graphics/bento images or should I give it a shot
if you want, i think niklas was busy we talked about them few days ag
its wasnt anything major, 3 pics or something right
Well I can do it but I'll just crop the default Aurora wallpaper to be the correct size in various places
I guess we can use aurorar logo and two other from the wallpapers?
or smething
Looks like it's just three sections in both light mode and dark mode.
yea
I don't know what to do for light/dark so I just made them the same
isnt there a light version of that wallpaper?
or nah
nope
atleast i don't think so
just put it as the dark mode version for both then
i think itll look pretty good
or maybe put that other wallpaper too
can we make it to start in dark mode by default?
i mean
can you make aurora start in dark mode by default?
oh i think we have a dark style already
so it should be fine
Yeah it was already dark mode when I launched it from the iso
we can help with taidan stuff too at some point is currently finals week(s)
totally not dying inside
in case of concern, your full name is on this.
im aware it’s even the version without my legal name
Nope
@ledif do you have those graphics somewhere? then we can drop them into the PR from inffy here https://github.com/ublue-os/packages/pull/418
GitHub
feat: add aurora readymade config by inffy · Pull Request #418 · ...
wip
needs graphics etc etc
or the ones you made
i think they look fine
Yeah I dropped them in this channel so inffy could add them to his PR
i'll add them now, missed them
Good luck with your finals!!!
https://github.com/ublue-os/packages/pull/418 are we missing something else?
GitHub
feat: add aurora readymade config by inffy · Pull Request #418 · ...
wip
needs graphics etc etc
looked in the bluefin package and i think we have everything
Okay since the aurora readymade config isn't merged yet I'm testing with bluefin's config. Seems to be working so far
heh bluerora
lemme ack and put the thing on the COPR
man i hope this actually installs fine
we need taidan to get the oobe stuff tho :(
honestly this is a very good looking installer even on KDE
no rounded corners but thats KDE being KDE
heh
Arghh it failed https://paste.centos.org/view/e64e67b0
does it say what failed?
BIOS
you need to run on UEFI
Yeah I put the link above
Hmm, okay
ah yeah
UEFI only
we dont have the hybrid MBR/GPT layout as a repart template
it someone figures that out then BIOS will work but honestly i dont think we should support that
also, @ledif the aurora thing is merged
lemme add it to the COPR
Okay this is with UEFI
no bios
hell yeah!!!!
it wont boot cuz you dont have an user on the installed aurora system :clueless:
you can manually add it by editing
/etc/passwd by doing some fuckery on the live environment
vi /mnt/mountedrootfs/ostree/deploy/default/deploy/*/etc/passwd
and /etc/shadowfails for me as well:
https://paste.centos.org/view/eeb08fbd
0: IO error for operation on /usr/share/readymade/repart-cfgs/bootcwholedisk: No such file or directory (os error 2)
ledif isoyou need the aurora-readymade-config package
you can install it w/ dnf on the live ISO
its super cursed but it works
aurora bad edition ™️
Yeah sorry I haven't updated my fork yet so you still have the old ISO. Let me push it
works
that looks amazing!
noice!
im guessing we should just crop it a bit more
looking awesome tho
this looks fire
kinda expected the pictures to be all screwed up when I resized the window, but it's not
omg
I wish it was a little bit more verbose
had an issue with closing the luks mapper device:
this is from journalctl
journalctl _EXE=/usr/bin/readymade
https://paste.centos.org/view/79d7f41e
cc @madomado
oh we need to clean up that mount
hmmm
that one is done by bootc itself
Oh there's an aurora test iso now?
me being in the minority that hates rounded corners 😉
a bit undercooked
the branding isnt there on the ISO ledif sent
you also need to manually create an user w/ a funky workaround
do you know what we need to do to integrate taidan?
i am honestly not sure where i could start
nothing yet
im just getting the damn branding working on bluefin
then we need copr handling on the ublue-builder container
then taidan on ublue-os/staging
ah no worries
then we just install taidan on aurora
i am not in a rush
if this works you guys can straight up copy it 1-1
perfect
Was able to get anaconda-webui to work with titanoboa
Still has work to do for flatpaks/secureboot and stuff
oh wow what?
how did you do that?
With titanoboa, create live iso of bluefin-stable
install
anaconda-webui anaconda-live
add the ostreecontainer with appropiate values.
run liveinstoh yup.
the postinstall scripts afaik arent that absurdly hard to do
the hard part is just the flatpaks
yepp
Flatpaks are baked in the iso right?
they are, its just that anaconda has a weird ass way of getting them working
they "are" but I haven't configured how to get them copied over
Ok, tmpfiles and /usr/share/factory
oh i just remembered that theres a lorax template they use to copy the flaptaks
flatpaks*
not putting them in the image tho
they need to be copied to the installed system
its a lot harder than that sadly
You are overcomplicating with lorax
not trying to use lorax
And regardless, all you do is to have a pristine base image with everything baked in
Like you want the flatpaks, use the container image for that why relying in something so obtuse as anaconda
If is going to be more troublesome
we aint gonna put it on the image its a big nono
It's going to end up in the iso anyways,
bro. calm down
right now, we embed them into the iso
we are trying to have a method to get them into the installed system from that embedded repo
cc @Robert
Solve for our fedora-logos issue. Use the bluefin copr
https://copr.fedorainfracloud.org/coprs/ublue-os/bluefin/
@j0rge
sup
and then because aurora doesn't have to use this... no name conflict
i think we might just want to do that yeah
I've been saying this for like 3 days. 😦
thought you meant that you wanted to put stuff on
ublue-os/packagesIt doesn't matter where it lives, it just needs to work
@tulip🌷 i think what I'm seeing is put new
fedora-logos package in packages/bluefin, build it in the bluefin COPR then it's not a conflict with aurora or bazzite or anything else...
and if aurora needs the same treatment, I can create an aurora repo... but right now sounds like bluefin is our priority?ok so: the code is already there on the ISO building stuff now
now we just need
fedora-logos (not really fedora-logos, just the anaconda branding there actually)Keep it simple
tell me what you need from me.
I thought we used to have a
https://github.com/ublue-os/packages/bluefin location which was going to build packages in https://copr.fedorainfracloud.org/coprs/ublue-os/bluefin/packages/i think its all good to go
i PRd the branding, gonna add it to the bluefin copr and test the ISO
https://github.com/ublue-os/packages/pull/440
(wip)
it's beagle time!
@Aurora Maintainer if you guys can, please send me some branding for your ISOs too
cropped like these:https://github.com/tulilirockz/packages/tree/packages-bluerora-logos/bluefin/fedora-logos/src/anaconda
I uploaded some here https://github.com/ublue-os/aurora/pull/447/commits/b07c8fd139d9c8a82da0c40a522f3849baa10cb7
GitHub
feat(installer): add branding for anaconda ISOs by NiHaiden · Pull...
This PR adds branding to the Anaconda Installation Experience, currently WIP.
oh nice thank you!
@bsherman if there isnt a copr for aurora yet, mind creating it?
oh there is!
ill just ask for perms there
👏
let's fugging go
granted
ill just test if this thing works with the bluefin ISO, then ill make an aurora package and put it on their copr too
thank you!
jesus christ it worked
YEs
LETS GO
i've been rebuilding this damn thing all day
😭
fedora-logos-100 -> fake fedora-logos package
also added copr handling to ublue-builder
so now we can have taidan
(aurora live ISOs will be good to use)
I had some issue like this
Spent maybe 5 hours on it
i did the same
I think it was something stupid as well
I gave it an epoch and made a stub that depended on my different named package
now it works
I'm trying to make a Live ISO from bazzite-dx and I'm getting this error for some reason after selecting the "Try Bazzite DX" option in grub
missing the secureboot keys
we need to add some handling specifically for bazzite
cuz we need the vanilla fedora kernel on it
I made some updates to titanoboa. The only external change to the build recipe is that it will always clean now.
Because we dump a rootfs out (and unless you change it) things go wrong if not starting from an empty directory.
Other than that, I left all of the awkward build call. I only have one shell usage of for a recipe argument and that was for variable assignment purposes.
I have some serious questions on why livesys and polkit are options? And their defaults. I'm guessing things kinda organically grew but I'm unsure why we are providing an option for them especially for livesys where we end up overriding a lot of the good stuff it does.
I'm also going to change the hook test to configure anaconda live mode since that is in repo. This is not me saying we are using that, but to have our CI test not fail out because a repo isn't available
oh the check is failing cuz the script is dumb
we can jsut remove the check script there
Livesys option shouldn't be a thing tbh. 99% of the times is going to be used
its installing readymade in a weird way
That's fine, but I'm going to change it to something that more of the tests can reliably run.
Is there a reason to have the polkit be an option?
Also, did you guys release the action or are people just using main?
Since I really want to remove some of these pointless options
oh we are using main directly
Since I really want to remove some of these pointless optionsburn them with fire! remove them all! literally everything on titanoboa was made jsut as a quick hack make it good as you deem fit
ok with both sets of ISOs, anaconda and readymade, the bluefin bucket is 123GB
big boy
it was double that for this month
but it's still like, well under 10 bucks
Hmm I thought I saw Noel or someone was able to make a Bazzite live ISO at some point. Am I misremembering that?
https://github.com/ublue-os/aurora/pull/456
this ready?
yeah @m2 is looking at this now
but for tomorrow we're going with just the refreshed anaconda ones
which are currently unbranded, PR above and related bluefin ones are to get the artwork/logos in there
then we fix up titanoboa and then generate live CDs, one set with anaconda and one set with readymade. Then we just autobuild all summer
I trying local iso build right now
you can make one; however, we have a key issue. MOK hasn't enrolled yet. I do not know if MOK will autorun if a key is present.
Solutions are either to use a generic kernel for bazzite installs (this is what happens right now) or turn off secureboot
no secureboot
on his VM
i think BIOS also works for the live environment
So anaconda 42 is broken
Bluefin is working for 42 with 41 ISOs?
I tested post scripts for Bazzite, so the Mok enroll did work.
@Kyle Gospo I'm thinking we need to use F41 to deploy F42 until we solve issues. I'm not sure where we even report this shit.
Why bother
I'm not using this shit any longer than I have to
live ISO tho
Let it stay broken
Let's just roll back to 41
yeah, Bazzite doesn't have live ISOs.
yeah we were testing them
remember?
oh I see.
it worked for you
yeah.
I do
not super sure why tho
I've not done the Live ISOs.
I'm talking about the existing ISOs using build-container-installer.
yeah
but I think it should work tomorrow for 42?
m2 is cooking so I'm waiting to see what happens
Bazzite just built 42 ISOs and they're broken
Anaconda fails to load
what's broken?
oh damn
Fundementally broken
yeah, wayland is crashing.
I'm gonna test one quick in a VM.
I mean, m2's close to the webui and everything
oh for live ISOs?
we should sync and compare notes
yeah
we can't use Web UI on KDE.
also we're conflating things
live CD is out of scope for now
we're just trying to get working installers, liveCD can come later
wait, the web UI using build-container-installer?
or titanoboa?
titanoboa
not sure m2 question, which is why we need to sync
m2 made a bluefin live CD with the anaconda-webui stuff
everything kinda works, just missing the flatpaks
like we had before w/ regular gtk anaconda
and we can always turn the webui off for KDE, but we'd still have a live CD with bazzite + gtk anaconda right?
why dont we just make f41 CDs w/ f42 container for the bazzite ISOs tho?
Yep, good plan
like we are doing on bluefin
then we figure out the live CD and replace it entirely
The F41 thing isn't an issue if we just hide the version number, the user won't know or care about that
if 42 is broken again. Set VERSION for jasonn3's action to F41
broken again yeah
yeah, that's what I was thinking.
like use 41 to install 42 but don't show version numbers
I cannot figure how to hide that number. I actually think its hardcoded into the anaconda package
the only people who will notice are in this channel. 😄
not sure if there is a way to do that.
oh, bummer
why is it so hard to install Linux?
man.
every.
single
cycle
anaconda has broken.
at least now we are fleeing from anaconda
so... a bit less to worry about
now you know what people mean when someone says "they're shipping their org chart."
I cannot wait to troubleshoot this as a red hatter.
anaconda can't be much better unless Red Hat holds it back to an ancient version.
I'll repeat what timothee told me
since they can't really say it sucks
"Everytime I look inside I find a bug."
How hard is to fork anaconda in copr?
I'll just link this 5 month old open PR https://github.com/rhinstaller/anaconda/pull/6056
GitHub
Support Flatpak preinstallation as part of a DNF install by owt...
Add new functionality for running Flatpak preinstallation after installing the base system. It&#39;s initially implemented only for the DNF payload.
Flatpak installation is done as an extra &am...
at least someone is driving it forward?
Yes kudos to them 😸
I built an f42 iso and it works
F42 Aurora?
yes
We'll see what happens, but it's installing fine in a secboot VM.
no wayland crash issue.
Could it be kernel related with the Legion go @Kyle Gospo ?
WAIT
did you only test on a handheld?
man dude how did you get the artwork in there?
We've replicated the problem on the Legion Go several times.
we need to copy what you do in aurorafin.
the fedora-logos package is all Bazzite does.
@inffy ^^^ ok we just need to find out what we're missing, I bet it's something silly
can someone run a diff between the two packages lol?
It must be something on the logo package
Cant think anything else
Where is the css file in the Package
https://github.com/ublue-os/bazzite/tree/main/spec_files/fedora-logos
in case y'all need to compare.
Ah Bluefin has it, Aurora does Not
yup, need t hat
The Spec File…
Why is it so huge lmfao
bluefin's artwork isn't working either btw, so if it has a thing aurora is missing that's likely not the issue. Ideally let's just match them all up lol
Okay, so I believe the issue is due to the size of the assets
Did you get Flatpak installs working in the live installer? 👀
not yet haven't figured out the incantation for that yet
interesting!
bazzites works but bluefin's doesn't
Bazzite on CentOS anyone? 😉
@M2 this ready? https://github.com/ublue-os/titanoboa/pull/104
No
Close though
I think I have an idea for flatpaks
@M2 any ideas on the image differences? like are the bazzite ones much smaller or something?
I'll check again. But these are still on the larger size for an ISO.
For Bazzite, they may want to consider having an "install" image or at the very least prune
like I wonder if we just need to pngcrush or something?
Honestly. Getting rid of all of the non default art would help.
We also have a shit ton of fonts. I think having equivalent coverage of Fedoras installer is goof enough
fonts in the installer?
I'm confused
do you mean titanoboa?
did we figure out how to remove the "test media" entry in grub?
sorry.... blending right now.
Bazzite when using Jasonn3 is still massive. I believe this is due to the oci dir not getting deduplicated.
For jasonn3 no. that is part of the templates
ok so I built a
bluefin:latest ISO and it shows Fedora 42 and is working in a VM
so I don't think we need to pin the version in aurorafin
but they couldn't get it to work in bazzitewe are pinned by digest. Was there a recent update?
https://github.com/ublue-os/bluefin/blob/2b362d0a37de395fc5470aad92238ec72788bad4/Justfile#L3
oh you pinned it last week
yes, and added a renovate action for updating it
All I did locally was run
just build-iso-ghcr bluefin latest
and that did the right thing
and it says F42, is that expected?
basically, tomorrow when coreos updates, akmods runs, ingests the new proper kernel, then renovate updates main, then renovate updates bluefin.
Then we switch the stable tag
and run the actions, we should be good right, F41 and F42 ISOs?
(I'm not making new GTS ISOs since we don't need them)Yes, the if block checks if it's LTS tag or not.
If not it uses the fedora version of the kernel. If it's LTS it uses F41 since that one seems to not break as often
If everything releases tomorrow. We build everything. The build ISO will be button presses but is lined up to build 42 for 42 and 41 for 41/lts
I prepped the PR for unpinning the kernels
it'd be hilarious if renovate rebuilds everything before we wake up
i got a bluefin
our fedora-silverblue.css is what's broken
@j0rge
the style sheet had an error. missing url
I forced merge
WHAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAT
I KNEW IT WAS A SMALL THING
@tulip🌷
tooooooooooooolip
it's over!
WE ARE DONE!
i also got flatpaks to install
in livecd land?
yes
super hacky
ok since you force merged we can build these tomorrow, then turn off the fucking workflows and close our eyes for 6 months.
look at that firefox there
yeah and tbird!
okay... need a better method. Everything is the wrong SELinux label
but appears to work?
LES GOOOOOOOOOOOOOO
FLATPAKS WORKED ON THE LIVE ISO?
yes, but need to rework it a little
sounds pretty damn good to me
Catching up......
Aurora is misssing the branding .css completely it seems.
Yepp. Copy bluefin's. Change your color. Profit
LES GOOOU
if the liveCD is reliable im up to just using that one instead of the default ISOs
i do know ventoy breaks it tho
who cares about ventoy.
I'm so bloody tired hearing about ventoy.
"your thing doesn't work with this third-party tool that you have no relationship with, and it's your fault"
we probably just wanna say thst ventoy is unsupported
Yes. You do.
also this is just an improvement over old bazzite ISOs
we do say it's unsupported
we might be able to just use that there too
https://github.com/ublue-os/packages/pull/463 this should do it right?
we have been since that entire thing with the binary blob started
I ain't got time for that
ugh i hate that
look at all the stuff I say is unsupported
I do not care.
so once sherman adds s76 and tuxedo akmods it's basically like, if you buy linux friendly hardware from a dedicated linux vendor, we will support you. You spend the money we will give you a great experience
like that's what we bring to the table, if you have the right kit we give you a legacy free experience.
the future has arrived for you.
and like, sorry asus and surface, dnf is here though so that shouldn't be a heavy lift for other folks. We just ain't got time to deal with that. We have larger fish.
you can only support what you can support, and test, if you want to offer anything resembling a coherent finished product.
anaconda looks great! Are we able to load the webui or are we stuck with the gtk?
no webui on jason's installer
but webui could be done on the live environment
You can run it. The version in repos on F41 is useless though. F42 versions can run post install scripts
So I have 2 strategies.
1. Minimal set of usable flatpaks with bulk in repo. Minimal set can be either a set list or confrigable
2. Rsync the working flatpaks to the install
cant we just symlink the flatpak directory?
pretty sure it works
No
we can also do bind mounting
The install repo is in a compressed format. I can try reconfiguring flatpak to to use a bare-repo instead to see if that would work
how do you generate the repo?
The /flatpak/repo is an archive-z2 repo. /var/lib/flatpak/repo is a bare-user
Ostree init
gts iso doesn't have the artwork
but at this point shrug
How?
Did the package not get rebuilt
dunno I am struggling today with this VM thing
Oh... I didn't change the version so it probably didn't rebuild
@tulip🌷 can you press a rebuild
whats up
omg maybe thats it?
ill press a rebuild on the bluefin copr
Yeah it doesn't look like it
ok to rebuild ISO?
Does anyone know how we can get rid of the Test and Install in the grub menu? It's set by default and it's a shitty one because it also does not work.
LOL
@M2
oh when it's at high rez it does this
and as I scale the installer it does that
@Aurora Maintainer I wonder if yours will do the same thing.
rclone copy r2:backups/bluefin-gts-x86_64.iso r2:bluefin
man rclone is just such an s tier tool. That's a serverside copy too, makes it nice and easy for me to do the iso and checksumwe need to use larger image likely
if you know... you know
hell yeah
ok yea but wasn't one of the problems that the images were too big?
they are, we will definitely need to remove unneeded packages. Like bazzite doesn't need libvirt and the offline steam tarball
or maybe it does need that tarball...
no I mean the pngs themselves
I know the image images are fat, I mean the artwork
oh on the webui there are no images
I mean on the ones I screenshotted
how they get cropped, the images for the previous gen.
let's just call your thing 2nd gen installer
but I rolled them back so it'd be a waste of effort.
since you appear to be actually finished.
yeah so I think what we do with the web front end is leave it because that's fine, but since it's like css we could do a help wanted ticket, and then have someone like submit a design. Would be something neat too because aurora would be the same way.
and like for bluefin it's like, the artwork is there already, it's a web design, totally different skillset than the usual installer stuff. This will be such a cool differentiator too.
like why haven't all installers just shipped with web ui's, that's so linux.
but also an opportunity for new people to play
I'm like "ok design nerds, you're turn, I'm taking a nap."
and you know kyle is going to slay with his installer.
Going to see the rsync method will work. I expect it to. But I don't like having these in a unusable repo
fixed selinux I believe meaning that the flatpaks on the ISO now have right perms. This then means you can copy them from the iso to install which is much faster than doing a flatpak install
okay. I'm "NOT" using anaconda directly for installing the flatpaks now. This is significantly faster on the rpm install step (and doesn't have the same gotcha's with openh264 and other flatpaks needing an online connection in order to install)
us using the repo method doesn't seem to make sense when we already have a properly configured and working flatpak install. Having two repos meaning having at least 2 copies of the flatpak runtime that even when compressed results in about a 400 MB larger ISO
This also means that branching to readymade or using anaconda is no different. Creating the repo was only for anaconda.
This means that titanoboa just makes a live-iso. The hook-script is where everything is configured.
GitHub
feat: leverage just dependency resolution by m2Giles · Pull Reques...
This reworks the justfile to accomplish 2 major goals.
First, remove internal just calls. This prevents embedding titanoboa easily as submodule into other projects since your root directory will tr...
sudo just HOOK_post_rootfs="./.github/workflows/ci_dummy_hook_postrootfs.sh" build ghcr.io/ublue-os/bluefin:latestDoes this also work for the non-live environments? Or is titanoboa only interested in live?
But that's awesome! WebUI with flatpaks!
Titanoboa is for making a live-iso.
Trying to understand what you are asking here.
yeah its for live isos
Sorry. Both does the rclone work in non-live environments?
And I guess following on, is this rcloning something specific to titanoboa, or something we can implement in the old ISO action
For the old ISO action there wouldn't be a desire to change it. The current method is robust and since you do not need a working flatpak in the non-live version you don't have the hit on doubled runtimes
But I wouldn't see why not. It's an anaconda post-script
I think mild confusion here
Titanoboa makes a live-iso.
Part of it is embedding usable flatpaks.
The method for configuring an installer is left to an external script that runs inside the new rootfs.
Since readymade is still cooking, I looked at using anaconda.
So I used the anaconda-live and anaconda-webui. But this could be a different installer if somebody chooses to do so.
My work to use anaconda is in the ci hook script which was previously just a place holder.
For that, I install anaconda. And make some configurations.
But it's mostly an example of how to do so and not part of the exposed action. Instead you would provide your own script to use with titanoboa.
GitHub
feat: leverage just dependency resolution by m2Giles · Pull Reques...
This reworks the justfile to accomplish 2 major goals.
First, remove internal just calls. This prevents embedding titanoboa easily as submodule into other projects since your root directory will tr...
@Niklas ⚡ did your ISO artwork come out ok?
I just realized the artwork is in the RPM
@M2 should I revert the artwork RPM so we get the solid sidebars and not the green cutoff things?
We only have the solid colour sidebars now
which i think is fine for now
until we figure out the other stuff
yeah that's what I want
we have this
https://discord.com/channels/1072614816579063828/1348452653805342942/1372017475989799015
ah yeah
ours is just solid blue
@M2 testing in a vm now, if we wanted to "break" them back to just being a blue sidebar, is there a simple way we can do that?
I don't mind breaking it and rebuilding ISOs, we have plenty of time
Yeah revert and rebuild package
Need larger artwork
yeah but I'd have to guess to cut and paste and do photo stuff
this one right?
Yepp
and I need to bump a version?
https://github.com/ublue-os/packages/pull/472/files
like this?
WAIT
they come out perfect!
no green thing!
WE ARE LIVE!
Lol
Revert the revert
I just closed it
it was like The Rock again
I GOT GREEN SMOKE, I GOT GREEN SMOKE
Thank God I was stubborn enough with "only liveCDs", seems it paid off
I agree, stubborn is good
i've been out of the loop because of finals what happened?
we're rolling with 2 sets of ISOs, anaconda and readymade.
and m2 is working on the livecd but that's a slow burn backstage
alrightie that makes sense
has anyone tried taidan on plasma yet?
it seems a bit undercooked right now
Time for Yet another installer?
Nope, I sort of bailed on messing with it, since it's going to be a massive pain in the ass to get the dependencies into :Factory to mess with it for Kalpa
and I'm not a huge fan of sticking GTK stuff into a Qt environment when I can avoid it.
im a huge chad so i just enabled networking on the copr build
perfectly valid.
:clueless:
YAI the GOAT!!!
pls no
You will take it and LIKE IT
yes it is sadly undercooked
it’s why 41 was so delayed
but we’re working on things
finals should be over for everyone now
AMA is tonight though you guys should pull up
user stuff needs to be in the installer
no reason to be trying taiden right now
On titanoboa I fixed up how we were doing the tarball extraction to keep permissions and file caps. So no more sketchy setuid loop step
GitHub
feat: leverage just dependency resolution by m2Giles · Pull Reques...
This reworks the justfile to accomplish 2 major goals.
First, remove internal just calls. This prevents embedding titanoboa easily as submodule into other projects since your root directory will tr...
Will be merging this
yeah man rev fast
i'll button press this which will upload an iso for you to look at
https://github.com/ublue-os/titanoboa/actions/runs/15079368003/job/42393307557
This should have a bluefin:stable image uploaded in about 20ish minutes
@M2 I think we forgot the aarch64 anaconda builds
Oh.
For LTS?
they 404.
yeah
lts and gdx
Okay
I'll take a look at fixing this either tonight or tomorrow.
The workflow for titanoboa to upload is also busted (path issue) guessing no one had actually ran that
So Bluefin LTS needs aarch64 workflows or modification to the current one to accept arch.
Is GDX getting built as an ISO?
From their gitlab? Yeah, it's really exciting!
LTS is beta so it's fine, the readymade aarch one works and that's a good one for VMs on macs anyway
like unless we need to fix it anyway that can wait vs, what you're concentrating on now
where is this project?
wuh
is this some animation or PoC or something?
🤤
I think it works pretty well?
WHAT THIS LOOKS FIRE
fixing now.
https://github.com/ublue-os/titanoboa/actions/runs/15085042093/artifacts/3144409961
bazzite-deck live-iso with anaconda-webui
Jason's build-container action can only run on amd64....
So for aarch64 bluefin-lts I'm making a titanoboa config (and since I want to make one as well)
There is an issue open, but it hasn’t been a priority.
I’m pretty burned out working on anything anaconda.
GitHub
Build ISOs (Live Anaconda) · ublue-os/bluefin@889670f
The next generation Linux workstation, designed for reliability, performance, and sustainability. - Build ISOs (Live Anaconda) · ublue-os/bluefin@889670f
@M2 I can make another bucket if you want a CDN place to push
it's easy just lmk what to call the folder
that would be helpful. The lts images will land in the current bucket
k
and the others will be github artifacts
yeah artifacts are too slow
okay.
testing is the bucket instead of bluefin
perms and everything should be fine
storage is so cheap so we might as well get CDNed testing ISOs, heh
rclone copy "${SOURCE_DIR}" R2:bluefin should be it right?
at the end of the workflow?
want me to PR it?changing that right now
k
https://github.com/ublue-os/bluefin/pull/2538
merged
It's building to push to the cloudflare bucket
nice
https://projectbluefin.dev/isoname.iso will be the final URL
oh what's nice about this setup
is "promoting" the test ISOs is a matter of rclone copy r2:testing r2:bluefinyepp
https://github.com/cmj2002/r2-dir-list
@Kyle Gospo we should think about this across the board.
so that
download.bazzite.gg and friends show a root dirGitHub
Build ISOs (Live Anaconda) · ublue-os/bluefin@7fd7d84
The next generation Linux workstation, designed for reliability, performance, and sustainability. - Build ISOs (Live Anaconda) · ublue-os/bluefin@7fd7d84
looks like access denied for the repo
sec
in the docs, reading
the token is set for the account that should work
Ok we'll need to set new secrets, sec
we don't have a way to test until we push do we
yeah
ok I don't want to mess with the prod keys but if this work we'll go back and consolidate
for that workflow
they'll push at artifacts to github at least
R2_ACCESS_KEY_ID_2025
R2_ENDPOINT_2025
R2_SECRET_ACCESS_KEY_2025
are all setnew keys?
yeah
so the prod keys need to be rotated anyway but we can do that later.
when we set it up we used user API tokens, we should be using account API tokens according to this doc so we'll redo them but that doesn't have to be today
okay added _2025 to the stuff
ok let's hope this is it, heh
hmm that should we working but same error
oh old window
@M2 I think I know why this error is here
sec
sorry this is all different in cloudflare
ok trying again, I had a carriage return I put in the secret key window and I was like "I wonder if that will matter"
it must have
retrying
ok we're good
trying an LTS one
what does RAM only in the grub option mean?
YEAH!
are we uefi only?
If you're not, you should be.
oh no, that's with uefi on
loads the iso into ram. Was useful for some testing
we are not.
Not sure why that failed. I don't see a typo
lemme try a new one
appears to be the main issue
the rest of the log is a bunch of python stuff, you want it?
no, guessing that the name is somehow not being embedded the same?
There's the problem, somehow lts got embedded
oh in the stable image?
yepp
heh, cool
Looking at the run now, it's not using the current version of titanoboa
https://github.com/ublue-os/bluefin/actions/runs/15087695578/job/42412840785#step:8:852
something didn't merge that I was expecting on titanoboa. Fixing that
also, amazing work on titanoboa
you really refined it a lot
i managed to fill up a 2TB drive with different image builds.... 💀
based
now you have a excuse to get a 4tb drive
another set of ISOs for anaconda-live are being built
still will have the issue with desktop icon i think
if our problem is the icon, then we're in a good spot hahaha
https://github.com/ublue-os/bluefin/actions/runs/15088811903/job/42414768559#step:8:810
pulling the right image this time
oooh I got time for one more test, then I'm out for the night
https://github.com/ublue-os/bluefin/actions/runs/15088811903/job/42414768630
is stable, will fixup the missing deps for lts
\/\h\o\m\e\/\r\u\n\n\e\r\/\w\o\r\k\/\_\a\c\t\i\o\n\s\/\u\b\l\u\e\-\o\s\/\t\i\t\a\n\o\b\o\a\/\m\a\i\n\/\o\u\t\p\u\t\.\i\s\o ]]
nicegithub and escaping strings
I know I just dig it with titanoboa in there
http://projectbluefin.dev/bluefin-stable-x86_64.iso
somehow this one is 1 GB smaller
probably because the container is the same 💀
hahaha
5.3 GB... we could probably nuke fonts to get under 5 GB
indeed
the u looks nice!
looks like it's working!!!
looking pretty good so far!
go skopeo go
DISCO
@M2 wow, you did it.
reboot and make sure flatpaks made it
yep
mine failed and I don't know why
BOOM SHACKALAKA
if we bump sushi and gnome firmware in flathub to 48 we save a runtime
why is sushi out of date did something replace it?
oh well, this is amazing!
so the flatpaks seem to have made it for me..... not sure why it failed (maybe secureboot enrollment?)
omg perfect 😭
oh
I didn't get an SB prompt
can you run
mokutilI blew it away already, trying gts will circle back
That's okay, secureboot is a "solved" problem for enrollment
so probably just a goof in the script
k
I'll take an item on that qr code
we don't want to send people to the fedora forum with this stuff it'd be pandemonium
no idea how to change that right now
it's ok it gives me something to do
gts is old anaconda, that's expected?
the artwork is right at least without the green
i think so, f41 doesnt have a functional anaconda-webui
(afaik)
right
wow this is perfect then
instead of 1 installer we have 3, very on brand
it has anaconda-webui, i choose not to install it because it's very, very underbaked
yeah leave this it's nice
cant we just build it on staging?
whatever tho
this is basically better than the GTS one, the best old one we can make
Yes, if we wanted to
and do webui in 42
I say we just leave it because I tried both and they both work
other thank some polish on the new one
oh top right corner
oh no fedora 41
need to add something to fix that
np
at least we can hack it a lot now that it works on titanoboa
pretty sure its fetching from some file somewhere
/etc/system-releasehow did you figure that out
i've never ever seen that file
i read the anaconda startup file
LTS shows as "CentOS Stream release 10 (Coughlan)'
heh
so it grabs everything before release as the Name. The number after release as the version
ok so this means the LTS one would be non-webui right?
correct
ok so that works out, next cycle everything rotates
we need to package it if we want it
I'm thinkin least amount of work at this point
old installer with the artwork for LTS/GTS, webui for stable, done.
im such a big fan of this theme
it looks so good
yeah
on the webui
that long blue bar
im sure theres a
.css file somewhere
how can i get those ISOs?
downloads page?projectbluefin.dev/isoname
different url
thanks!
we make the bar like this
That would be amazing
then it'll be the symbolic
u and the "Bluefin GTS"
ideally we can bury that version # somewhere, because it's handy
maybe in that hamburger menu thingThat's the pretty name, super easy for me to change
bluefin LTS just has
Bluefin LTS on its /etc/os-release
shrugyeah but that'll be on old installer anyway
trying a few
that one is nice because it's dark on the left, so the words would stand out better
oh nice, the terminal sbkey warning is there in the motd, that's nice to know
mokutil is here
did it enroll the key? or was it an after step?
no I had to do it by hand
found a docs issue fixing
same postinstall script as readymade, right?
basically the exact same
does lts have our akmod key in
/etc/pki/akmods/certs?nope
nothing akmods related on LTS at all rn
Okay so for LTS still need to fetch the key
@tulip🌷 builds for uupd are failing because the SPEC requires internet
oh lemme fix that
how do we fix that, do we need to pre-download the deps in the repo somehow?
its a pain
need to package all the golang-* stuff we use
fucking whatever
oh boy
more rpms
cc @SFaulken
:glorp:
builds are going
w/ internet
are we able to get builds with internet or no?
could we build in github actions and then just upload the final rpm?
would be annoying to have to package all of the golang deps
Golang deps can be vendored if it helps.
oh that would be much easier
I think it's as easy as
go mod vendor && git add/commit/pushis it just
go mod vendor
oh ok
do I need to specify a directory or does it just create oneIt just creates one at the root of the project named
vendorin COPR, yes
but overall, vendoring is the way to go
centos doesn't have libblockdev-btrfs meaning that anaconda cannot install onto btrfs when using lts as an ISO.
They said it couldn't be done
that doesn't look quite right
ooh all green
LMFAO oh no
hyperscale is apparently building it
no idea how to actually grab it
oh we can get it from their repos
sec lemme get it
you can install centos-hyperscale-release then isntall it from there
Doesn't seem to work
you sure? that looks perfect...
hmmm
isnt it centos-hyperscale-release??
https://sigs.centos.org/hyperscale/
going to bed. But that's the next blocker
where are you working on this? i might be able to help you out
while ur at bed
Pretty cursed
didnt you have a functional win95 installer?
readymade is literally just systemd-repart + bootc install
It kind of worked, but it couldn't create users. So I'm trying to to use taidan currently
taidan!!
go with ur cool installer! bootc install fixes so much stuff
Yeah if I can get readymade to work for the installer part, my custom installer should be fine. But I just want to focus on the OOBE currently
@tulip🌷 figured out the hyperscale. The repo was staying disabled even when installing it. We have libblockdev-btrfs now on lts
NICE!
anything for me to test rn?
yeah. Main one that is untested right now is the MOK enrollment
I think I fixed that
ok what iso should I pull?
is removing the fonts from titanoboa a matter of an exclude list?
I can figure out fonts if you want
its not an exclude, it's an uninstall
ok, I assume it's in one of these sh files
unless we want to purge them from images as well
like where do I insert the remove
so there's an open discussion for that that will take forever to resolve
but for the ISOs I mostly want a free gig back if I can get with a few dnf commands
yeah
also timothee has ideas on fonts I need to run by him
given X amount of time with timothee I always optimize for the more important things
awwww yeah!
https://github.com/ublue-os/bluefin/actions/runs/15096679228
kicked off a build to push to cloudflare
I can't find where we clean up the ISO
like where would I put the dnf remove google-noto-fonts-all
https://github.com/ublue-os/bluefin/blob/ae250f89ff1e0e52ea1cc7af0f2a1f001f643319/iso_files/configure_iso_anaconda.sh
we can do one of two things.
Remove it completely from images. That's removing it from packages.json.
Or, we only remove it from the iso and that's putting it in this hook
yeah only ISO for the space for now
right we'll get a gig back?
lts doesn't have it. But gts/stable does
its huge
yeah that's what I mean, it give me something useful to contribute. 😄
This will then be our font coverage with noto
I would put it under line 13 where it says configure the live environment.
The area you put is where anaconda configuration is occuring
do I need a explicit dnf clean at the end?
doesn't hurt, but I run autoremove and clean all during the cleanup step after this
make sure to put
|| true to handle if it failshttps://github.com/ublue-os/bluefin/pull/2543
look what copilot recommends
I don't want it to fail....
TIL
I mean, it's not the worst idea in the world, other than it's completely inappropriate in this context, and dnf would give you a fail message anyway if it died =P
yeah it's ci if it breaks i don't care
fail fast for me
in this case we are removing somehting. The fail here would be if its not present, which is the same desired endstate
got anything offhand that is huge on the image you wanna see gone?
most of the font sins are in dx
yeah
it looks like samba, opendyslexic-fonts are possibilities
I guess also gcc and brew setup
are we using
jetbrains-mono-fonts-all as our default terminal fontI recall tulip dropping brew already
yes we are.
we disable it, but it's still there on the iso
dropping brew would help a ton if it's not already removed, I know we didn't have it in the readymade, and it's just removing it via the rpm right?
oh ok I can add that
you think
ublue-brew will work in there lolyepp
wow that's 2 huge things
honestly get rid of the motd as well
nah it's too cute
like it's in every screenshot everyoen takes
well we need a different message since it talks about brew, is broken for rpm-ostree command that runs in it
but we can argue the small ones once we get these huge wins
oh I see.
damn .... ok, let's see what we get with these 2.
to see if it's worth it
oh drop yaru-theme
yeah
Nuke these folders
https://github.com/ublue-os/bluefin/tree/main/system_files/shared/usr/share/fonts
well not inter?
inter is a package now there's a seperate PR in packages
ok that's a good win for now, I don't trust my bash with anything more I can socialize some more slimmin'
LOL nice!!!! hell yea
you didnt need to remove ublue-brew or ublue-motd 😭
we'll figure it out tho
we don't configure brew and have the tarball on disk. The motd is broken since we don't boot with rpm-ostree
probably just need a different live-inst motd that doesn't have brew stuff in it
i think we get that error even without the motd
nope. it's caused by the motd
the ublue-brew tarball is interesting tho
we are checking if you are on a signed image in the motd
correction we are checking which image you are on in the motd
We are checking if you are on an old image
also we should make this stuff into hooks or something
cuz it would be cool for bazzite for example
bazzite uses this as well
it doesnt use our package tho
we didn't remove anything yet
it's a PR still
nah its all ok
oh I merged it...
lol
HE'S GONE FULL JORGE.
but's its only for anaconda-live
it's ok sherman is asleep
yolo merge all day
we're only removing this on the anaconda-live iso, not on the deployed image
yup!
also so far appears to have zero space savings. GTS is weirdly bigger than stable
Before removing those packages
Didn't we find out there's deduplication between the container and ISO anyway? So removing files from one probably wouldn't do much - all the fonts are still in containers-storage
yeah
that's what its looking like
yeah we don't need to remove that, we'll edit the config to not do the rpm-ostree check so that doesn't appear
i wonder how erofs looks like now
there was a proposal on fedora for switching to erofs
on isos
they did switch
oh what? really?
yeah on F42
it makes the ISOs so much bigger tho
even with their configs
they are actually bigger.... lol
waow, well done castro
i mean negligable difference
so I guess things are deduped?
but looks like if we want font savings we'll need to remove it from the image
yepp
ISO is purely additive.
I agree but let's not couple that to this work, that's a hairball
I mean we should do it
I'll PR a revert?
so we'll remove that, change the config key for motd
no, don't worry about it
k
just like removing most extra coprs too
if we back it out we don't need to change the motd tho?
i'm about to go for a run. We are not exposing these to anyone right now so I don't think this matters
yeah
but on the next set of changes, will remove the removal
I don't want to side quest, I thought I'd get an easy gig for you, but let's not distract from the main work, table for later
GitHub
Build ISOs (Live Anaconda) · ublue-os/bluefin@5bdeb19
The next generation Linux workstation, designed for reliability, performance, and sustainability. - Build ISOs (Live Anaconda) · ublue-os/bluefin@5bdeb19
like these are pushing to cloudflare right now
I'll try to meet with timothee soon and get a font recommendation
nod
our next main quest is making sure mok enrollment is working
waiting for green on the iso
using .dev for these is smart, let's keep this
double raptor but probably a gnome issue
dude the docs docs are there.
oh shit what? how did you open that
omg we need that on the thing too
on the readymade ISO
that's a bummer
probably some CSS thing
we can figure it out
@M2 MOK works!
Double raptop is because that's a zenity window showing.
works for me
On 42 clicking the raptor is also weird since the installer runs as a Firefox window
good enough
Okay. So css changes
So what's left?
I think we're good
GTS/LTS cosmetic issue with the sidebar
stable looking awesome
do we a screenshot of the new titlebar top part on the webui with the cleaned up name?
but IMO these are already better than what we have.
What happened to the other installer? So we back with Anaconda? Nice that it's still live ISOs though
we're making both
@M2 these are at least good to go to fix the download page
@M2 ok to copy the LTS ones over to the prod bucket?
or if you just want to switch LTS to prod bucket that's fine that's already way better
Sure
ok so the rclone glob is only useful for either copying the whole bucket or individual files
a subset is annoying
@M2 and for GTS/stable, how you feeling?
I'm feeling kinda good
Fuck it we're doing it live
that's what I wanted to hear
ok fuck it, kick the actions over, lemme run through a few stable ISOs runs
snaggin usb sticks
supposed to still look like that?
we should straight up just modify os-release on the ISO
@M2 ok so order of operations, turn off all existing anaconda workflows, adapt the new live one to push to prod, then push?
Mind if we cut a v0.0.1 in titanoboa so we can pin to an actual release?
cc @M2
do you think its good enough for that rn?
good idea no matter what
sgtm then
someone cut one
man this webui installer is nice people will love this
confirm mok fixed in stable
https://github.com/ublue-os/bluefin/pull/2544
Yes. All of the things that j0rge mentioned are on the bluefin side
Yes, this is the pretty_name of the ISO.
We can change this on bluefin side if we want this to be smaller. Personally I would drop the FROM portion and the word version.
You have to force merge. I made it so changes to the ISO files don't cause the image builds to kick off since they are a separate
@M2 anaconda live ISO seems to have the RAM option
you might want to remove that one
it just breaks all the time
Oh it was useful for me during testing
But I also have a stupid amount of ram
dinos are squished LOL
Lol
what if we just reuse the readymade icon?
or make a
ublue-like icon for the installer?
like the docs oneThat could work.
The one on the right is from pixmaps so we're not changing that one
But the one on the left we can definitely change
Sooo. Supposedly fedora switched to web UI in their livecds?
for Workstation i think
KDE doesn't have the webUI installer
Ok that makes sense
So they basically hardcoded the locale switching to gnome, great
Yepp. Locale/keyboard is hard coded to gnome right now. It literally just opens up the settings from gnome
the webui installer isn't probably going to work with anything other than gnome anytime soon, because of the unholy hacks they're doing to get mutter and firefox to play together to make it work.
it's something the Fedora KDE SIG is already wrestling with.
well it "works" if you can deal with the us only kb layout
yup.
The old installer is still available. Just don't install
anaconda-webuiIts me or lately is all firefox issues? having to setup an entire profile just to have a navless windows, now mutter ducktapping
Jesus christ
Also why cant just localectl in anaconda to switch the language?
no idea. Localectl would be the logical solution for me
I removed teh to Ram menuentry to avoid that confusion.
I'm getting errors in Titanoboa during the relabel step.
Is there anything specific needed beforehand?
Hmm. I'm just going to remove wget from my workflows. Maybe that file will never be added
i have seen that occur. I'm not really sure the best method to force a cleanup beforehand. But basically
chcon won't apply partial labels so if its still unlabeled after setfiles runs, it will bork out
just submodule method works!
LES GOOO :letsgo:
gotcha's are that you cannot set a variable directly, but instead can pass it as a environment variable
additionally, relative paths don't quite work due to being inside the "source" directory. Pass paths as absolute paths
but I think we saw that already in the action.yml
I may get those move stuff changed to and the sha256sum to be part of titanoboa. Basically what should the output name be, and whart directory should it be in
but other than that, this worked. Need to start making some flatpak lists now
and my own hook!
Thoughts on instead of doing the full system, to instead just do /usr, /var, /etc. Or at least the LHS
After removing wget usage, I don't have this. It's probably not something Titanoboa needs to worry about - just something general OCI builds should try and clean up.
No idea why the files are present. But they seem useless
No idea why the files are present. But they seem useless
I also don't understand why wget would put a lock file in / instead of /root/
But yeah makes sense. We basically expose an entry point for a person to run a script or many scripts that get passed as a string
I guess the only other thing to think about is if we should expose other hook points.
Accepting many scripts or flatpak lists would be nice!
It can accept one flatpak list.
For the many scripts that works right now. All it does it cat the file. So if you provide to files it cats both of them
So you can have a "base" configuration, adding things like the installers. Then image-specific files for changes per image
Yes, that's possible right now for the hook. Not possible right now for flatpak list
So for your hook.
"hook1 hook2" both would be cat together and passed into the container as one big script
That's a can of worms
The moment you have more than one, the entire thing will overcomplicate
I also can't really think of something other than providing a custom grub template.
Everything else can be done from inside the chroot
yup!
so we are done!
RiscV support 🙂
everything that we expose right now is good enough
i wonder if the aarch64 iso building method would work for riscv
literally everything else is the same on aarch64 and x86_64
I think in CI we should delete the rootfs after the squash step.
After the squash step we don't need the rootfs and essentially have another copy. I believe during ISO step we can run out of space because it makes a copy of everything in the iso-root directory.
This should only be in CI since we may want to inspect the rootfs otherwise
I don’t know if it is related, but I did select the “Try Bluefin (RAM only)” option from the grub menu.heh, we can get rid of this entry right?
Think M2s already done it. Maybe need to regenerate ISOs
ack
I'll refresh after the next set of work
things are looking quite nice right now!
yeah its gone if you refresh isos
ok running a set
https://github.com/ublue-os/titanoboa/commit/15cd3ca4d1f34f5c12dbc32bfe83a8ffb3611add
This seems to have broken the builds.
iso-organize step requires access to the rootfs for the vmlinuz
iso-organize step requires access to the rootfs for the vmlinuz
yepp. I just fixed it
https://github.com/ublue-os/titanoboa/commit/0835d3f5521c407f40b4e923a5f89982857b5881
the sudo call when invoking this is also making things weird for environment detection. The action.yml needs to explicitly have the CI environmental variable passed into it
Added this show config so we can acutally see what is being passed into the build call
Also failing in the process-grub-template step, needing access to os-release
I've pinned to 1d07ef51dbd6c5fb88b86bab9c41505ca11cd359 (removing the RAM grub entry), which works
Hmmm.... I'll take another look. Guessing ordering issue
Probably put the rootfs delete at the very, very end
Anaconda WebUI with Titanoboa works perfectly with Cosmic!
I'm strongly thinking we switch to these live isos as a default ASAP if we haven't already 🙂
Switching the keyboard layout doesn't work
Is hardcoded to open gnome keyboard settings 🫠
Ah, the deeplinks into settings
This is a Mandela effect moment, but I'm sure added rootfs removal back then
I'll check, there may have been one when we were doing internal just calls
There is for sure a clean step in the action.yml
https://github.com/ublue-os/titanoboa/blob/03413da6f4576edabb94da01b937ab20261e3067/Justfile#L361
No, it only deleted the image
so is it a link or is is calling a desktop entry? If its the latter we can probably do some sed magiv
A link
It's an anaconda:// link
Yeah, might be handled internally somehow(?)
Anyways if we run out of space in the runner, we can always do the funny :v
https://cdn.discordapp.com/attachments/1172596198788575322/1372001698503200788/image.png?ex=682c6fe7&is=682b1e67&hm=caebdd60a4800ec59c4766af13f40a6790215412ef514b43e7602f7e45e23416&
Cranking up that compression if we need to
https://github.com/rhinstaller/anaconda-webui/blob/f28f6ffda2ffcffc64aa33cb8766d6f3ac4234d7/src/components/localization/Keyboard.jsx#L226-L232
So this needs fixing
https://github.com/rhinstaller/anaconda-webui/blob/f28f6ffda2ffcffc64aa33cb8766d6f3ac4234d7/src/components/localization/Keyboard.jsx#L74C14-L140
... Somehow
Don't know a lot of JavaScript, but I'm guessing it is like that's the part where it is hardcoded to us?
Yeah, so gnome just redirects to settings.
Other DEs should still support changing the layouts within Anaconda WebUI itself, but it's broken. The second link is the "other DEs" handling
Ah, so it's the "is gnome?" detection that's broken
Okay looking at this, seems webui might be way easier to patch
Or at least for bazzite default password, really neat
honestly, we probably should do something funny like this.
Ye download free disk space wins 😂
Anyone have an image to hand? Could you check the other default config files for an "[Installation System]" TOML header and the different "type" values in
/etc/anaconda/profile.d/*.conf?
Should look something like
Oh, ignore that
Looking at the wrong line :facepalm:
This is how it's detecting the different DEsWe did on bluefin stable
And Lts and GTS but they're not webui
Right, to that might be ok,
Probably the desktop variant function
When you hover over the "change keyboard" button in the installer, does it show the anaconda-gnome-control-centre:// link?
Yup
Strange... The logic looks good in the installer.
Mmm I'm assuming I can open the firefox dev console there
Yep. F12
The profiles are still in anaconda and seem to affect anaconda's behavior
I've been looking through the webui code.
It says "set desktopVariant to the value of $XDG_CURRENT_DESKTOP. If that desktopVariant is GNOME, set isGnome=true. If isGnome, use the anaconda-gnome-control-center hardlink, else use built-in KB selection".
can we like override that handler
somehow
Unlikely at runtime. It'll be compiled and not really readable
Ok kind of weird, in my VM, the keyboardNonGnome function doesn't show up
im sure there is some fuckery we can do
Hmm. Do you know the version of webui you've got?
42.27.12
Likely we should build from upstream.
Upstream is on 35
Upstream uses an older version?
That function was added in v25.
Might be minified, so cannot search by name
If you search the dev console for the following string, is it there?
"More than one layout detected. Remove additional layouts"
Don't think is minified
OH! Added last week
https://github.com/rhinstaller/anaconda-webui/commit/09c196f8c6e71b931bd8e4e4d3a63b1e8c4c5a8d
So should be fixed in an upcoming release - v37
Oh nice
we can always put that on staging btw
Or we can. You know, grab the rpm directly
If I'm not mistaken, they packit there
Hell yeah
lol nice
Building on every commit it seems
Every commit. And probably every pr
Problem is ofc it's a copr, so we cannot pin that version, will be the bleeding edge
👀
id be 100% fine w that
https://github.com/rhinstaller/anaconda-webui/blob/main/packaging/anaconda-webui.spec.in
They have a spec file and everything
That in is suspicious
Probably they do templating with packit
Couldn't we build it and pin versions?
we can just periodically build the rawhide RPM
Also noticed they include the RPM Firefox on non-RHEL systems. Removing that may reduce the ISO size
Edit: Maybe not that simple. Uses non-webui Anaconda unless changing another firefox-ext script
we probably need to audit all thr things on the ISO just to make sure we arent including like 3 KDE runtimes
The thing is, now you would need to setup their firefox profile yourself with the flatpak firefox
Nah nah reading the wrapper for the shortcut, we better don't touch that thing for now, is a rabbit hole
Bluefin is pulling in free desktop 23.08 due to thunderbird. That's the only pretty bad one.
Aurora is in worse shape due to the combo of gnome and KDE apps and a couple of runtimes
we can drop tbird on the ISO can't we?
oh nm it'll be in the payload
Yeah I looked that and noped pretty quickly.
They are going to have to solve this for rhel/centOS stream so it won't hurt to wait.
For fedora, the Firefox package isn't bad.
we have a few G47 ones in there but it's not that much of a diff between gnome runtimes
For gnome 47 vs 48, it more depends on what is the base it's built off of. If both are 24.08 it's minimal.
Any flatpak remove is immediate savings; however, it will not be copied to the install.
It is not embedded into the container, but is instead apart of the live-iso
oh
well we can always can it if the runtimes aren't getting updated
Thundebird is weirdly on 23.08. I'm considering moving back to evolution right now and seeing if I can get the flatpak to talk to gnome's EDS.
Evolution is ugly.... But it works and is usually getting small but frequent updates
This is confirmed fixed and removes rootfs fixing out of space issues I've seen (not to say we probably should still consider steps to help this out)
We definitely should cut a release now so we can pin to a tag.
I'm around
Tag a release for titanoboa and I think we are in a good spot for calling it v0.1.
I think remaining changes are more ergonomic than anything else at this point.
I think our consideration now is mostly on the side of installers.
I'm interested in us packaging the webui and it appears to be getting active changes right now. For bluefin being gnome, we also have the more complete solution right now.
I forgor how to make a release without release please lol
GitHub
Release v0.1 · ublue-os/titanoboa
What's Changed
feat: use system's qemu if installed by @ledif in #2
fix: use /app instead of /tmp by @ledif in #3
feat: create livecd user and autologin for SDDM/LightDM by @ledif in #6
f...
ok now what?
We copy the webui.spec and add to staging?
Then update the bluefin hook to use the updated webui to poke at it?
Oh I thought we were just respinning ISOs not packaging webui
why can't we use it as is?
evolution and thunderbird are equivalent for the majority of basic users
thunderbird does have far more features overall
and is better supported
Oh just respin
in UM GNOME we take the lcd path and ship geary
Removed the ram entry
clicking
Geary has to use online accounts..... Which is a fail for quite a few oauth2 locked items
EWS/o365 works out of the box on evolution.
It doesn't have the whole Firefox instance meaning that you will open PDFs/links using other programs.
I also like the "single" workflow for all of the ISOs
i've used evolution in thunderbird i swear
may be wrong
You can use them in thunderbird using an add-on. OWL works. But out of the box no support for ews/o365.
Evolution's pain point is gtk3 and older UI. I do not recall if it supports gmail tags
The version name in the webui is just the pretty name in /etc/os-release. We can change it to anything.
For my image, I've set it to be "<Project Name> (<Image Ref>)". e.g. "Eternal Linux (ghcr.io/rsturla/eternal-linux/lumina:42)"
For my image, I've set it to be "<Project Name> (<Image Ref>)". e.g. "Eternal Linux (ghcr.io/rsturla/eternal-linux/lumina:42)"
yeah I know we're on it I just wanted to preempt whichever random nerd is going to yell
Are we able to change the default EFI (or /boot?) size in anaconda to allow people to pin more Nvidia images?
that should be possible in the conf file I think
something like this maybe:
https://github.com/rhinstaller/anaconda-webui/releases/tag/37.1 looks like there is a new tag for webui, seems to fix the issues with keyboard on non-gnome DEs
GitHub
Release 37.1 · rhinstaller/anaconda-webui
show bootloader devices on review screen
components: storage: fix disabled radio buttons
storage: avoid rendering wizard pages while Cockpit Storage is active
components: keyboard: run Gnome keyboa...
That doesn't seem to work unfortunately
We only need to increase /boot right now
damn
just keeping this alive it's like the only way i can keep track of this busy place lol
GitHub
Support Flatpak preinstallation as part of a DNF install by owt...
Add new functionality for running Flatpak preinstallation after installing the base system. It&#39;s initially implemented only for the DNF payload.
Flatpak installation is done as an extra &am...
where can I test the newest readymade changes so I can test my locale?
probably the worst time for asking this lol
from git or with readymade-nightly from terra
helium is getting some design upgrades (most notably for readymade, better page titles
should be landing relatively soon
that music player app is actually real if you want to use it
https://github.com/tau-OS/victrola
GitHub
GitHub - tau-OS/victrola: Play your music, in an elegant way.
Play your music, in an elegant way. Contribute to tau-OS/victrola development by creating an account on GitHub.
@madomado I see important pushes! I assume I'm building a pushing a new ISO?
nope, I'm tired so I haven't done testings
I'll ping you when I confirm everything works
:3
haha ok
@jorge hey, a bit late, but tested both with and without encryption, seems to be working fine
also bumped to v0.12.3
still trying to make the packages build in terra
cc @j0rge
merged, in a call
LMAOO
aaaaaaaaaaaaaa
Hi, I want to report my failure on installing Bluefin via readymade.
So, previously my laptop has:
- Windows 10 in drive A
- Windows 11 (Tiny11 ver) in drive B partition 1 (+ partition 2 for data)
Before I use the readymade live CD, I booted into the old Win10 instance on drive A first to do one final backup.
Though, I remember that I failed to boot on the Win10 for a bit. After a restart I got the Windows boot option with Windows 11 and Windows 10 and go through with it... (So, a windows boot rebuild?)
I guess why my Bluefin installation is failed once I formatted the drive A in the readymade process.
Seems like I had to repair my Windows boot manager first? Luckily the files on both the Tiny11 and the Data partition are still untouched... I just have to find a spare flashdisk to flash the Tiny11 ISO and try repair these, before using my current flashdisk to install Bluefin again.
Here's the full readymade errors
also yeah that's already the 4th trial (after a reboot)
here's the first process/error receipt
here's the second and third trial (where the process went fast, straight up failed, I ended up rebooting)
Thanks for the report
We recommend removing any drive except the one you're trying to install to. Not just when using Readymade, but when using Anaconda and installing Windows/other distros
You're welcome, though I don't think I can do that since I'm on a laptop (the best I can do is just unmounting)
Oh. Could you explain your setup? You're not trying to dual boot on a single drive, are you?
Nope, I would just boot different drive from the BIOS boot menu
And did you previously have another Linux distro installed before installing your Windows OS?
We've seen cases where Windows installs it's boot contents to a different drive than the actual install data when it detects another OS
Worked fine back when I had the Win10 / ublue cosmic-atomic-main spin (But I encountered failure after a daily update so I ended up wiping it and installing Tiny11. Seems like it's a better choice for a 50GB OS partition + a data partition compared to a ublue image)
hence the Win10 in drive A (120GB) and Tiny11 in drive B setup
Thats probably it then.
Your tiny11 install used the drive A boot files (EFI), so when Readymade tried to wipe drive A to prepare for the Bluefin install, it removed tiny11's boot files
well for some reason Impression always resort to the live Bluefin flashdisk instead of the Verbatim spare drive I want to use haha
oh it finally worked
are you using bios?
we currently only support uefi
Aah alright
So... back to the anaconda ISO then?
I think ublue generally doesn't support bios
correct me if I'm wrong
@Robert (p5) meow, throwing the ball back to you
i'm in a thinkpad x230
and installing bluefin went properly back then, a month before 42 iirc
anyway the tiny11 flashing is done, gotta try repairing it first before flashing the anaconda version
welp, blank, so i gotta flash it on rufus kek
Yeah. Ideally nobody should be using BIOS, and it's not something we're testing against.
Any semi-modern hardware from the past 10 years should support UEFI
Anaconda supports bios. However I'm really confused on your setup
Sounds like either a dual drive laptop (though I've not seen many of those) or installing to an external disk
my x230 is dual drive!
Made this issue since it's the second time we've ran into this
https://github.com/FyraLabs/readymade/issues/110
GitHub
Warn when wiping disk containing an existing EFI partition · Issue...
There's been two instances so far in Bluefin land where Windows decided to install it's EFI partition on another drive, and Readymade cleans that drive when installing the OS. If there'...
ah I see…
im pretty sure its blowing up because we have the retry logic :(
Hi, it's morning for me and I'm back again. Tried the command here https://answers.microsoft.com/en-us/windows/forum/all/windows-boot-option-disappearing-from-boot-manager/d50956d9-94e6-41ae-9212-373591c34a0c and got "Element not found"....
Gotta search again and exhaust the options I have with this flashdisk B (Windows). If I can't do anything again I'll just use flashdisk A (Readymade) to flash an Anaconda Bluefin to flashdisk B and install bluefin from there
EaseUS
How to Fix Bootrec /fixboot Element Not Found Issue on Windows 11/10/7
Are you receiving bootrec /fixboot element not found error on Windows PC? Then follow this guide to fix the problem using advanced recovery, system restore, and other tools.
finally!!!!
ah yaaaay finally
Hi hi, I'm back on Bluefin. I was able to install it using the new Anaconda installer.... until the last part, Finalization, where it got stuck (perhaps because of the memory leak? I know my Linux will gets freezed once it reaches 3.3 GB or more, but sometimes I will wait until it responded.... though after 30 minutes there's no sign of anything so I just gave up and force turning it off.
Booted into Bluefin. Nice. I create account and such. Sadly all Flatpak-based apps can't be opened at all, so I guess that's where it's broken?
I was able to do
ujust update, I updated from stable-20250525 (the ISO) to stable-20250608, but yes I got these messages in the end:
At least the main browser I used is installable via bash script so now I can type and describe this.
TLDR: What should I do if I failed the Finalization stage on Anaconda installation and flatpak stuff got broken because of it? Should I try reinstalling?
Currently trying to rebase it to stable-daily branch
Hmmm
Okay, I got some "progress" thaks to
sudo ujust install-system-flatpaks (? i just choose it)
Only 15/62 apps got installed, still almost no progress, but at least there's one. I'll try it again laterSounds like the flatpak install failed. It uses rsync and I'm guessing it failed to copy over correctly.
Likely the permission errors is due to SELinux labels.
Try
sudo restorecon -Rfv /var/lib/flatpak
Then run the ujust for install flatpaks.
So to understand, your computer froze during the finalize step?
I'm wondering if it's possible to limit rsync to only use 50% of total memory or something. But it's effectively doing a clone of the /var/lib/flatpak dir on the ISO to the new installYup. Though I use wifi back then, opened another window of firefox (but I closed it as soon as the deployment stuff is done)
When I reinstall it without connecting to internet the installation is done properly, problems solved. Thank you sir!
@imbev
I see that you are using BIB. Anything you've learned about it to make the build/install process smoother?
Find a tag that works and stick with it
yeah... for whatever reason the image today will not generate a working iso for me
Does it work with an older tag?
Disk Images work is the weird part only ISOs are choking.
They also don't seem to have tags?
I only see digests on quay
But latest seems quite broken for me building a centOS image
That's what I meant, sorry.
Also try an older tag/digest of CentOS, it might be the image instead of bib
Trying with Cayo right now, will see if the behavior also exists with an modified centOS
GitHub
Install of
readymade-bluefin-lts-aarch64.iso Fails On PinebookPro...I downloaded the ISO, wrote it to an external USB, and booted up just fine. The installer ran most of (?) the way through and failed running the bootupctrl to setup the EFI. See the log below: 0: R...
i have this hw
it's mainline but tow boot hates fedora's EFI
It doesn't require uboot?
tow boot is a uboot distro that provides efi
@M2 would it be viable to run anaconda-webui in a container for GTS and LTS? since the the bootc image we are deploying doesn't need to match the "host OS" for Anaconda?
We sorta were doing that before. Except what we were changing was the bundled OCI archive compared to the host.
LTS has a weirder consideration of having a different EFI directory name. GTS should be fine.
I also think this will bloat ISO sizes because of package differences
gts solves itself in a few months.
LTS may solve itself too, seems like work continues on the webui, idk if there's any timeline set. Maybe they will wait until rhel 11 but that seems like a long time to wait
yeah
Hey hey, I'm trying to add centos support to titanoboa, looks like the builder is only fedora, would that mean it can only build fedora liveos right now?
nope, it should be fine with other distros as well i think
it was supposed to be at least if no one changed anything
it's using only fedora/fedora as builder idk if that makes things not work