dan walsh will be giving a talk on making a custom image, so syncing with him is good
dan walsh will be giving a talk on making a custom image, so syncing with him is good
Jand then spin up on zstd:chunked, though at this stage I think it's just waiting for bootc support
Nyeah, I plan on attending that.
Jactually if you see carl george go up to him and be like "jorge says you're the guy that's going to volunteer to enable centOS builds with a shiny epel gnome and kernel, is that true?"
JJoh man
Jyessssss
NLooks cool.
RSome interesting things from the bootc meeting today:
- There's early stage internal conversations about possibly moving rpm-ostree away from ostree-ext, and instead pulling images through bootc (https://github.com/coreos/rpm-ostree/issues/4994)
- Bootc can now pull other images (for use in quadlets) BEFORE boot - pulling the images at the same time as the update. Obviously the main use-case they mentioned is for LLMs so these don't need to be bundled in the same OS image.
- crun-vm has been updated to support running bootc images as a container runtime, which may be useful for testing (https://github.com/containers/crun-vm/commit/a9118ee63c76bf3b0f674eb164b35090f383f862)
- There's early stage internal conversations about possibly moving rpm-ostree away from ostree-ext, and instead pulling images through bootc (https://github.com/coreos/rpm-ostree/issues/4994)
- Bootc can now pull other images (for use in quadlets) BEFORE boot - pulling the images at the same time as the update. Obviously the main use-case they mentioned is for LLMs so these don't need to be bundled in the same OS image.
- crun-vm has been updated to support running bootc images as a container runtime, which may be useful for testing (https://github.com/containers/crun-vm/commit/a9118ee63c76bf3b0f674eb164b35090f383f862)
NSick!
AThere's early stage internal conversations about possibly moving rpm-ostree away from ostree-ext, and instead pulling images through bootc (https://github.com/coreos/rpm-ostree/issues/4994)
Asaw that, interesting
Apodman PR is still not merged
NThanks for attending those in my absence, been super busy as of late.
RThey said there's people working on this implementation "full-time". It touches many different repos
Ayeah saw updates, also opened this issue as related
RAnd composefs will only support zstd:chunked
at the current time, composefs requires zstd:chunked images
Asince yeah i had to convert an oci image into an OSTree commit
Ai ran into a bunch of problems head on
Ado we use composefs atm?
RNope
Ai think they mean shipping composefs
Apremade
Ai though ostree had a dial you could enable to turn on composefs locally
Aguess i can check my vm
Rhttps://github.com/containers/storage/commit/0b440b9b565548aae75a0d467be1181211ae4113
I think there is, yes. It does an "on-the-fly" conversion, but this obviously requires resources to do
I think there is, yes. It does an "on-the-fly" conversion, but this obviously requires resources to do
Aoh thats what you mean
Ai thought you meant ostree
Ayeah makes sense for podman
RNa. Sorry. I'm all over the place 
Aif you tried the on the fly backend you'd know
Aits hot garbage
Alike 4x slower download for gzip
Athey probably need to rewrite it
Aok seems like we are not using composefs yet
Jno, they're goingn to turn it on in 41, we have an open issue in main, it's not straightforward
Jfor sure don't mess with it on bare metal, use a vm
Ahad a couple of my users say cachyos is faster
Awould have been an easy first thing to check
Ahttps://github.com/ublue-os/main/issues/594 saw that, i made rechunk do this now, dunno if you kept up
Fexpected considering how customized it is regarding tweaks
Aa compiler aint gonna do that much of a difference
Athey said its noticeable
Fin my view, difference that doesn't matter
RFYI - Nvidia 560 was just merged to negativo's F39 & F40 branches. Hopefully we get this very soon
I think this update is fully automated now. We do not define Nvidia versions in any repos
I think this update is fully automated now. We do not define Nvidia versions in any repos
Jis it a matter of just firing off akmods then hwe?
RI think so. Or let the builds to their thing
Is there a way to check the package builds from negativo? There was a GitHub commit that enables 560 in the spec file, but I don't know if that means the package is ready yet
Is there a way to check the package builds from negativo? There was a GitHub commit that enables 560 in the spec file, but I don't know if that means the package is ready yet
JI can just fire off akmods and we can see what comes out
Awell matters or not something makes it feel substantially faster
Aboot speed on bazzite also seems slow, we should profile it
RStill 555.. Need to wait a while I guess
