you want some feedback on this recommendation? - Universal Blue

© 2026 Hedgehog Software, LLC

More

Communities Docs About Terms Privacy

bsherman

Bbsherman this probably should have it's "fork" threads cut https://github.com/ublue-os/s...

Arcitec

A

Arcitec•5/16/23, 6:30 PM

Definitely. We just got a ticket in ublue-os/base which got labeled inside ublue-os/startingpoint as "related".

The ties need to be cut for sure.

xyny

Xxyny Well it changed!

Arcitec

A

Arcitec•5/16/23, 6:30 PM

The rename even propagates to forks of startingpoint. This is cool!

bsherman

Bbsherman you want some feedback on this recommendation?

Arcitec

A

Arcitec•5/16/23, 6:32 PM

Yes would love to. But I was gonna mention that I figured out another big issue with how bluefin is forcibly writing to the dconf database of the system:

Rebasing away from bluefin breaks the system because all the bluefin settings (since they're settings rather than mere defaults) will stay behind.

On the other hand, if we soften it to use overridden per-image system defaults instead, then rebasing to bluefin would also break in reverse if a person rebases to bluefin but has modified some defaults on their system. Then those don't get synced.

So yeah it's a mess. Dconf was written long before the idea of rebasing between distros ever existed as a thought!

Arcitec

AArcitec Yes would love to. But I was gonna mention that I figured out another big issue ...

bsherman

B

bshermanOP•5/16/23, 6:33 PM

look at this: https://discord.com/channels/1072614816579063828/1087140903031943178/1107849123610234930

Arcitec

AArcitec Yes would love to. But I was gonna mention that I figured out another big issue ...

akdev

A

akdev•5/16/23, 6:43 PM

I think trying to fix home directory configuration hysteresis is out of scope for ostree-based systems so if you keep going this way you can probably find this problem with anything that writes to the home

akdev

A

akdev•5/16/23, 6:44 PM

Clean rebasing would need something to deal with all the homedir stuff

j0rge

J

j0rge•5/16/23, 6:44 PM

yeah

j0rge

J

j0rge•5/16/23, 6:45 PM

I am totally fine being like "hey so dconf is old so like, here's the gotchas when rebasing."

j0rge

J

j0rge•5/16/23, 6:45 PM

like I'm not filing issues on kde/gnome's home settings and paths, I ain't got time for that

Arcitec

A

Arcitec•5/16/23, 6:50 PM

gonna move this message to bluefin

Arcitec

A

Arcitec•5/16/23, 10:21 PM

Sherman and I discovered that we don't seem to be calling "cleanup" anywhere, which means that we're shipping the RPM caches of all installed packages. That's a lot of useless data.

Arcitec

A

Arcitec•5/16/23, 10:22 PM

RUN rpm-ostree install NetworkManager-wifi NetworkManager-wwan wpa_supplicant wireless-regdb && \
    rpm-ostree cleanup -m && \
    ostree container commit

RUN rpm-ostree install NetworkManager-wifi NetworkManager-wwan wpa_supplicant wireless-regdb && \
    rpm-ostree cleanup -m && \
    ostree container commit

This example of how to clean caches is given by cgwalters: https://github.com/coreos/rpm-ostree/issues/4178

Arcitec

A

Arcitec•5/16/23, 10:23 PM

I've done a ripgrep across all repo files and can't see cleanup anywhere: rg --hidden 'ostree.*cleanup'rg --hidden 'ostree.*cleanup'

Arcitec

A

Arcitec•5/16/23, 10:24 PM

Wait, maybe this is a false alarm. I was looking at fixing startingpoint and noticed that it does manual cleanup. I guess all repos do it this way.

RUN rm -rf \
        /tmp/* \
        /var/* && \
    ostree container commi

RUN rm -rf \
        /tmp/* \
        /var/* && \
    ostree container commi

Arcitec

A

Arcitec•5/16/23, 10:25 PM

It's only being done in startingpoint and main.

Arcitec

A

Arcitec•5/16/23, 10:25 PM

rg -g 'Containerfile' --hidden 'var'
os-startingpoint/Containerfile
35:        /var/* && \

os-main/Containerfile
19:RUN rm -rf /tmp/* /var/*
21:RUN mkdir -p /var/tmp && chmod -R 1777 /var/tmp

rg -g 'Containerfile' --hidden 'var'
os-startingpoint/Containerfile
35:        /var/* && \

os-main/Containerfile
19:RUN rm -rf /tmp/* /var/*
21:RUN mkdir -p /var/tmp && chmod -R 1777 /var/tmp

Arcitec

AArcitec Wait, maybe this is a false alarm. I was looking at fixing startingpoint and not...

bsherman

B

bshermanOP•5/16/23, 10:26 PM

yeah, but this is a good example where its a different step from another RUN where we do installs

and each RUN is a layer, so cleanup happens too late

Arcitec

A

Arcitec•5/16/23, 10:26 PM

Yep, exactly. And nvidia image doesn't clean up.

Arcitec

A

Arcitec•5/16/23, 10:26 PM

Out of the repos I have cloned (main, nvidia, startingpoint), only main and startingpoint wipe out temp files.

bsherman

B

bshermanOP•5/16/23, 10:27 PM

As far as I'm concerned, this can be part of https://github.com/ublue-os/main/issues/94

akdev

A

akdev•5/16/23, 10:27 PM

We should change to the rpm-ostree cleanup call if that works

akdev

A

akdev•5/16/23, 10:27 PM

Does it clean /tmp?

akdev

Aakdev Does it clean /tmp?

bsherman

B

bshermanOP•5/16/23, 10:28 PM

no, it just cleans the rpmdb metedata stuff

akdev

Aakdev We should change to the rpm-ostree cleanup call if that works

Arcitec

A

Arcitec•5/16/23, 10:28 PM

I was thinking the same but I think it only cleans /var/cache/var/cache folders related to rpms (downloaded packages).

j0rge

J

j0rge•5/16/23, 10:28 PM

we should check

j0rge

J

j0rge•5/16/23, 10:28 PM

when we started we had to do all this cleanup

j0rge

J

j0rge•5/16/23, 10:29 PM

and at some point colin told us that the ostree container commit was good enough

Arcitec

A

Arcitec•5/16/23, 10:29 PM

rg -g '*Containerfile*' --hidden 'var'
os-nvidia/install.Containerfile
21:RUN rm -rf /tmp/* /var/*
23:RUN mkdir -p /var/tmp && chmod -R 1777 /tmp /var/tmp

os-main/Containerfile
19:RUN rm -rf /tmp/* /var/*
21:RUN mkdir -p /var/tmp && chmod -R 1777 /var/tmp

os-nvidia/build.Containerfile
36:COPY --from=builder /var/cache /var/cache

os/Containerfile
35:        /var/* && \

rg -g '*Containerfile*' --hidden 'var'
os-nvidia/install.Containerfile
21:RUN rm -rf /tmp/* /var/*
23:RUN mkdir -p /var/tmp && chmod -R 1777 /tmp /var/tmp

os-main/Containerfile
19:RUN rm -rf /tmp/* /var/*
21:RUN mkdir -p /var/tmp && chmod -R 1777 /var/tmp

os-nvidia/build.Containerfile
36:COPY --from=builder /var/cache /var/cache

os/Containerfile
35:        /var/* && \

j0rge

J

j0rge•5/16/23, 10:29 PM

like I am reasonably certain we don't need to rm anything anymore or do the cleanup -m

j0rge

Jj0rge and at some point colin told us that the ostree container commit was good enough

Arcitec

A

Arcitec•5/16/23, 10:29 PM

Yeah definitely not good enough, since cgwalters said to use cleanup

https://github.com/coreos/rpm-ostree/issues/4178#issuecomment-1331441533

j0rge

J

j0rge•5/16/23, 10:29 PM

but I could be wrong

j0rge

Jj0rge and at some point colin told us that the ostree container commit was good enough

bsherman

B

bshermanOP•5/16/23, 10:29 PM

i think its good enough IF run in the same layer (RUN line)

j0rge

J

j0rge•5/16/23, 10:29 PM

is thaty the newest thing we can find?

Arcitec

A

Arcitec•5/16/23, 10:30 PM

Dunno but nov 2022 is new.

j0rge

J

j0rge•5/16/23, 10:30 PM

because something in rpm-ostree from november might as well be written on a clay tablet

j0rge

J

j0rge•5/16/23, 10:30 PM

so it could go either way, heh

bsherman

B

bshermanOP•5/16/23, 10:30 PM

colin walters is a machine

bsherman

B

bshermanOP•5/16/23, 10:30 PM

cranking out the goodness

j0rge

J

j0rge•5/16/23, 10:31 PM

https://github.com/coreos/layering-examples/blob/5644ba14808e3c439bd22204d07eef76bea6e089/wifi/Containerfile#L4

GitHub

layering-examples/Containerfile at 5644ba14808e3c439bd22204d07eef76...

Contribute to coreos/layering-examples development by creating an account on GitHub.

layering-examples/Containerfile at 5644ba14808e3c439bd22204d07eef76...

j0rge

J

j0rge•5/16/23, 10:31 PM

this is what we want right?

Arcitec

AArcitec ```sh rg -g '*Containerfile*' --hidden 'var' os-nvidia/install.Containerfile 21:...

Arcitec

A

Arcitec•5/16/23, 10:31 PM

^ So it turned out that the containerfiles have different names in the nvidia image. Turns out they do cleanup of varvar and tmptmp too.

bsherman

B

bshermanOP•5/16/23, 10:31 PM

There's no need to debate this. Just test it and demonstrate if it's helpful or not.

bsherman

Bbsherman no, it just cleans the rpmdb metedata stuff

akdev

A

akdev•5/16/23, 10:31 PM

Then we probably need a script to clean /tmp too

j0rge

Jj0rge https://github.com/coreos/layering-examples/blob/5644ba14808e3c439bd22204d07eef7...

Arcitec

A

Arcitec•5/16/23, 10:31 PM

Yeah we may want that in addition to the rm -rfrm -rf

akdev

A

akdev•5/16/23, 10:31 PM

Oh I see it is not needed anymore

j0rge

J

j0rge•5/16/23, 10:32 PM

yeah, I think what happened is he told us what to do and we did that, but when we merged stuff from other repos we might have added the rm -rf

j0rge

Jj0rge like I am reasonably certain we don't need to rm anything anymore or do the clea...

akdev

A

akdev•5/16/23, 10:32 PM

I remember reading Collin's notes he wanted the commit to do the cleanup

akdev

Aakdev I remember reading Collin's notes he wanted the commit to do the cleanup

Arcitec

A

Arcitec•5/16/23, 10:32 PM

Yes this would make sense, but does it do it automatically or is it still a wish?

j0rge

J

j0rge•5/16/23, 10:32 PM

I am reasonably certain it's automatic

Arcitec

AArcitec Yes this would make sense, but does it do it automatically or is it still a wish...

akdev

A

akdev•5/16/23, 10:33 PM

I'll have to check rpm-ostree to really know

but I trust Jorge

rg -g '*Containerfile*' --hidden 'var'
os-nvidia/install.Containerfile
21:RUN rm -rf /tmp/* /var/*
23:RUN mkdir -p /var/tmp && chmod -R 1777 /tmp /var/tmp

os-main/Containerfile
19:RUN rm -rf /tmp/* /var/*
21:RUN mkdir -p /var/tmp && chmod -R 1777 /var/tmp

os-nvidia/build.Containerfile
36:COPY --from=builder /var/cache /var/cache

os/Containerfile
35:        /var/* && \

rg -g '*Containerfile*' --hidden 'var'
os-nvidia/install.Containerfile
21:RUN rm -rf /tmp/* /var/*
23:RUN mkdir -p /var/tmp && chmod -R 1777 /tmp /var/tmp

os-main/Containerfile
19:RUN rm -rf /tmp/* /var/*
21:RUN mkdir -p /var/tmp && chmod -R 1777 /var/tmp

os-nvidia/build.Containerfile
36:COPY --from=builder /var/cache /var/cache

os/Containerfile
35:        /var/* && \