-
Notifications
You must be signed in to change notification settings - Fork 18
Update CI VM images for newer packages #284
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Re: rawhide build error @lsm5 do think it's the |
rawhide now has |
@cevich isn't |
Actually that's expected unfortunately. The image-build happens via Packer, which runs in a CentOS VM. This is because Packer needs to interface directly with the GCE APIs. Best bet is to use |
Just had a thought. I wonder if this is simply a "We don't test upgrades to rawhide" problem. I could just nuke the (presumed) |
1ff4e58
to
b2d6e22
Compare
Argh, didn't work. @lsm5 do you know if it's even possible to update/upgrade from force-push: Wild-guess: Try |
My current workstation is rawhide. I don't remember any issues while upgrading from dnf to dnf5 when I did f38 -> rawhide. Let me try it in ci_vm |
@cevich how should I be running get_ci_vm here? I see there's a |
Okay, I think that's helpful. So either the problem here is unique to here, or maybe some new upgrade problem crept in. You're probably right, some manual tinkering is probably required to figure this out.
This repo. produces the git_ci_vm container that the script uses. So you can't use it here directly. There are a few debugging make targets, but I don't think that's what you want. They're all geared toward debugging the build process, not the output VMs. I'd suggest using get_ci_vm from another repo, like podman or buildah if you need a quick VM to play on. There's also container images (F38 and F37), but fair-warning: It's a HUGE image. Anyway, (unless you're curios) don't kill yourself over this. I can go hands-on just as easily, and see if I can figure it out. I'll ping you if I get stuck again. |
SGTM :) |
Force-push: Experiments showed that upgrading |
hmm, didn't notice it during the local upgrade sometime last week, but won't be surprised if things changed in that short span. Glad that it works though :D |
Ugg, well, it worked in a container. It still fails here (same error). I'm going to see about using some duct-tape to convince |
@cevich maybe you could start directly from a rawhide image? Fedora makes those available too |
Thanks for the suggestion. I'll keep that as "Plan B" since it probably involves a lot more work. I think I figured out a workaround: Remove For future ref. / in case it helps. I used |
Yay progress! Now the rawhide image build is failing with:
Eed! Poking around, I see there was a build a month ago for this version. I clicked around on stuff (I don't know what I'm doing) and found https://koji.fedoraproject.org/koji/taskinfo?taskID=102699029 (for the @flouthoc do you remember if/what we need wasm for in our CI images? I grepped down the podman |
err yes, that's a problem i hit in containers/crun#1000 as well. I need to check with the wasmedge maintainers. The wasmedge package might need a rebuild. |
Fedora automation is pretty neat: https://bugzilla.redhat.com/show_bug.cgi?id=2219457 |
@hydai @dm4 could you please take a look at this bugzilla? wasmedge is currently not installable on rawhide. I think it would need a rebuild. |
It seems like a dependency problem. I found the |
1bd41fc
to
24c5074
Compare
Force-push: Added |
@cevich It was needed for this issue containers/podman#16501, but I don't think its a hard requirement. If its easy then its good, but we should not block on it afaics.
Not sure if |
Hi @lsm5 |
Thanks for the quick response @hydai, I'll will let you know if it breaks.
Dunno, it's setup as a dependency though so I'm keen to trust the package maintainer.
Thanks for the feedback @flouthoc, PRs linked to that issue seem pretty low-level (i.e. not just some tests). I'll leave it as-is for now, since there's a packaging fix. |
33b69fb
to
938917a
Compare
@lsm5 ping - the joy continues 😠 This is a new error as of this morning:
How did you find that automatic rawhide BZ before, (I can't seem to find anything on the new error)? |
do you know what package actually requires this ? Maybe we can just get rid of it? EDIT: I see it's pylint. Hmm, probably not something we can avoid. Can you try with
I simply searched for wasmedge in bugzilla and the BZ was there. I can't find anything yet for astroid, maybe the bz account only runs periodically. |
we can also |
Yeah, it's needed for a few repos CI. I'll try |
Ref: containers/podman#18612 (comment) Also: * Switch to the distro. version of `passt` since development has cooled down. It's also now available in both F37 and 38. Note, for the F38 images, it will still grab it from updates-testing. * Implement a few fixes to cope with the `dnf` to `dnf5` update when switching from F38 to rawhide. * De-duplicate & force use of `DEBIAN_FRONTEND=noninteractive` by including it into the `$SUDO` variable. * Add debugging to `base_images/debian_base-setup.sh` to help verify `DEBIAN_FRONTEND=noninteractive` is set. * Move python packages out of Fedora rawhide due to a broken dependnecy: `nothing provides (python3.12dist(astroid) <= 2.17~~dev0 with python3.12dist(astroid) >= 2.15.2) needed by python3-pylint-2.17.2-2.fc39.noarch` * Fix the name of the debian kernel headers package to not mention the currently booted kernel version. This fixes an issue where an older kernel is in use and doesn't match currently available headers package. Signed-off-by: Chris Evich <[email protected]>
no love @lsm5 same error. I think maybe we don't need many/most python things in rawhide. Trying with them only in Fedora/prior-Fedora... 🤞 |
Cirrus CI build successful. Found built image names and IDs:
|
@cevich i heard in fedora-ci channel that rawhide testing is broken due to dnf5. So, maybe you wanna put this on hold for a bit? But, I see all checks have passed, so your call :) |
I'm even more torn since I'm headed out on PTO for two weeks. At the same time, these updates in podman are VERY long overdue 😢 Perhaps I won't tag the images, and instead just try them out only in podman... |
@cevich SGTM. |
...podman results are improved. I'm going to to leave this PR open, as a reminder to me to followup when I return from PTO. Then we can decide if |
Looks like I need to build updated Debian images, merging this and starting a new PR. |
Ref:
containers/podman#18612 (comment)
Also:
Switch to the distro. version of
passt
since development hascooled down. It's also now available in both F37 and 38. Note, for
the F38 images, it will still grab it from updates-testing.
Implement a few fixes to cope with the
dnf
todnf5
updatewhen switching from F38 to rawhide.
De-duplicate & force use of
DEBIAN_FRONTEND=noninteractive
byincluding it into the
$SUDO
variable.Add debugging to
base_images/debian_base-setup.sh
to help verifyDEBIAN_FRONTEND=noninteractive
is set.Move python packages out of Fedora rawhide due to a broken
dependnecy:
nothing provides (python3.12dist(astroid) <= 2.17~~dev0 with python3.12dist(astroid) >= 2.15.2) needed by python3-pylint-2.17.2-2.fc39.noarch
Fix the name of the debian kernel headers package to not mention the
currently booted kernel version. This fixes an issue where an older
kernel is in use and doesn't match currently available headers package.