build: patch osbuild; set img ref properly #201

dustymabe · 2025-12-04T16:52:58Z

We currently set the imgref to something we cannot use later with bootc
or rpm-ostree. This should fix the reference and allow bootc upgrade
to automatically work.

In service of the above goal, backport a few patches from osbuild upstream.

None

dustymabe · 2025-12-04T16:53:23Z

Opening this PR to see if I can get some feedback on if it works or not.

Rather than write into container-storage and then export to an oci-archive let's just go straight there. Signed-off-by: Dusty Mabe <[email protected]>

We need to patch OSBuild with patches from upstream PR [1] to allow us to set the imgref to the value we need and not have the image building fail with an error. [1] osbuild/osbuild#2270 Signed-off-by: Dusty Mabe <[email protected]>

We currently set the imgref to something we cannot use later with bootc or rpm-ostree. This should fix the reference and allow `bootc upgrade` to automatically work. Signed-off-by: Brent Baude <[email protected]>

github-actions · 2025-12-04T20:43:32Z

Image Downloads for cirrus build 5317168310517760:

x86_64	aarch64
podman-machine.x86_64.applehv.raw.zst	podman-machine.aarch64.applehv.raw.zst
podman-machine.x86_64.hyperv.vhdx.zst	podman-machine.aarch64.hyperv.vhdx.zst
podman-machine.x86_64.qemu.qcow2.zst	podman-machine.aarch64.qemu.qcow2.zst
podman-machine.x86_64.wsl.tar.zst	podman-machine.aarch64.wsl.tar.zst
podman-machine.x86_64.tar	podman-machine.aarch64.tar

Everything zip

dustymabe · 2025-12-04T21:04:57Z

I'll mark this as ready for review once osbuild/osbuild#2270 merges

baude · 2025-12-06T15:30:27Z

This dupes and extends #200 so i will close the previous PR as such.

dustymabe · 2025-12-09T15:34:04Z

I'll mark this as ready for review once osbuild/osbuild#2270 merges

merged upstream!

baude

LGTM ... at what point do you think we can stop patching? any chance there is backports going on or is it the next version.

openshift-ci · 2025-12-09T15:40:34Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: baude, dustymabe

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

~~OWNERS~~ [baude]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

Luap99 · 2025-12-09T15:41:48Z

build.sh

+echo "Patching OSBuild"
+dnf install -y --quiet patch
+patch_osbuild


how urgent is this here?

I rather not see this merged until the fixes land properly in the main repos. Any runtime install like this is prone to introduce plenty of flakes.

how urgent is this here?

That's a question for your team to answer.

@Luap99 this is holding up my work on updates

If you need this now sure, as long as we agree that we don't make the final 6.0 release with this I am good merging.

I absolutely agree with you. But I will leave the merge up to you.

dustymabe · 2025-12-09T15:48:12Z

@baude
LGTM ... at what point do you think we can stop patching? any chance there is backports going on or is it the next version.

Probably mid-january. If you wanted to open a PR to backport the patch to current OSBuild RPM then maybe faster.

When the new version of OSBuild makes it to the repos this patch would start failing. That's your cue to drop the patching (can probably just revert the commit that introduced it).

Luap99 · 2025-12-09T18:04:14Z

/lgtm

Luap99 · 2025-12-09T19:05:31Z

This broke CI I believe, the publish task can no longer push the images to quay

https://cirrus-ci.com/task/4738727450247168

I guess the removal of podman save removes the actual image name from the archive so the load was unable to restart the name, does rpm-ostree compose build-chunked-oci add the name to the archive, maybe the name is a bit different now?

dustymabe · 2025-12-09T19:55:28Z

This broke CI I believe, the publish task can no longer push the images to quay

https://cirrus-ci.com/task/4738727450247168

I guess the removal of podman save removes the actual image name from the archive so the load was unable to restart the name, does rpm-ostree compose build-chunked-oci add the name to the archive, maybe the name is a bit different now?

It has nothing to do with the archive I don't guess. The log you linked is trying to build a manifest from containers-storage. I didn't realize there was a later step that ran that leveraged the images from containers-storage. We can just revert 16ff701 then. Sorry about that.

Luap99 · 2025-12-09T20:10:12Z

This broke CI I believe, the publish task can no longer push the images to quay
https://cirrus-ci.com/task/4738727450247168
I guess the removal of podman save removes the actual image name from the archive so the load was unable to restart the name, does rpm-ostree compose build-chunked-oci add the name to the archive, maybe the name is a bit different now?

It has nothing to do with the archive I don't guess. The log you linked is trying to build a manifest from containers-storage. I didn't realize there was a later step that ran that leveraged the images from containers-storage. We can just revert 16ff701 then. Sorry about that.

No it is not about the local storage, the build tasks uploads the oci archive and the the publish task downloads the archive and all the other artifacts of the build and loads the archive as image so we can add it to the manifest list and then push that to quay.io. My guess (untested) is that the archive creation does not add the image name metadata adn as such the load doesn't restore the name.

If the direct creation via rpm-ostree is faster I think we should keep that. Restoring the name is cheap and simple enough I assume, just an extra step after load.

openshift-ci bot added do-not-merge/work-in-progress do-not-merge/release-note-label-needed labels Dec 4, 2025

build: write directly to oci-archive when rechunking

16ff701

Rather than write into container-storage and then export to an oci-archive let's just go straight there. Signed-off-by: Dusty Mabe <[email protected]>

dustymabe force-pushed the dusty-set-imgref branch from 06bc4d5 to 2154b33 Compare December 4, 2025 16:54

dustymabe mentioned this pull request Dec 4, 2025

Set img ref properly #200

Closed

dustymabe force-pushed the dusty-set-imgref branch 2 times, most recently from 6a2b2f1 to 18ee4ee Compare December 4, 2025 17:15

build: patch OSBuild to fix setting imgref

7e0793a

We need to patch OSBuild with patches from upstream PR [1] to allow us to set the imgref to the value we need and not have the image building fail with an error. [1] osbuild/osbuild#2270 Signed-off-by: Dusty Mabe <[email protected]>

dustymabe force-pushed the dusty-set-imgref branch from 18ee4ee to 0fd9ff4 Compare December 4, 2025 17:31

build: set img ref properly

8383444

We currently set the imgref to something we cannot use later with bootc or rpm-ostree. This should fix the reference and allow `bootc upgrade` to automatically work. Signed-off-by: Brent Baude <[email protected]>

dustymabe force-pushed the dusty-set-imgref branch from 0fd9ff4 to 8383444 Compare December 4, 2025 20:00

dustymabe mentioned this pull request Dec 8, 2025

stages/ostree.{aleph,deploy.container}: support more imgref formats; make aleph stage more reliable osbuild/osbuild#2270

Merged

dustymabe marked this pull request as ready for review December 9, 2025 15:33

openshift-ci bot removed the do-not-merge/work-in-progress label Dec 9, 2025

baude approved these changes Dec 9, 2025

View reviewed changes

openshift-ci bot added the approved label Dec 9, 2025

Luap99 reviewed Dec 9, 2025

View reviewed changes

openshift-ci bot assigned Luap99 Dec 9, 2025

openshift-ci bot added lgtm release-note-none and removed do-not-merge/release-note-label-needed labels Dec 9, 2025

openshift-merge-bot bot merged commit 02518ff into containers:main Dec 9, 2025
12 checks passed

build: patch osbuild; set img ref properly #201

build: patch osbuild; set img ref properly #201

Conversation

dustymabe commented Dec 4, 2025 • edited by baude Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dustymabe commented Dec 4, 2025

Uh oh!

github-actions bot commented Dec 4, 2025

Uh oh!

dustymabe commented Dec 4, 2025

Uh oh!

baude commented Dec 6, 2025

Uh oh!

dustymabe commented Dec 9, 2025

Uh oh!

baude left a comment

Choose a reason for hiding this comment

Uh oh!

openshift-ci bot commented Dec 9, 2025

Uh oh!

Luap99 Dec 9, 2025

Choose a reason for hiding this comment

Uh oh!

dustymabe Dec 9, 2025

Choose a reason for hiding this comment

Uh oh!

baude Dec 9, 2025

Choose a reason for hiding this comment

Uh oh!

Luap99 Dec 9, 2025

Choose a reason for hiding this comment

Uh oh!

baude Dec 9, 2025

Choose a reason for hiding this comment

Uh oh!

dustymabe commented Dec 9, 2025

Uh oh!

Luap99 commented Dec 9, 2025

Uh oh!

Uh oh!

Luap99 commented Dec 9, 2025

Uh oh!

dustymabe commented Dec 9, 2025

Uh oh!

Luap99 commented Dec 9, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

dustymabe commented Dec 4, 2025 •

edited by baude

Loading