[podman-5.8] Backport working CI here#1010
Conversation
Make sure all go module related changes are committed properly. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 6f52244) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
|
Packit jobs failed. @containers/packit-build please check. |
Do not define GIT_CHECK_EXCLUDE in two places and just define it once and make the action use the Makefile. In addtion we don't want to lint the full set of commits, this seems to cause side effects with merge commits as it lints commits that may no actual be in the PR and thus it can fail. Using "github.event.pull_request.base.sha" is the right base commit we care about and makes the validation work. Thanks to Miloslav for finding that value. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit f250115) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
No longer built fuse-overlayfs and bats from source, we have latest versions in the CI images already. The CI broke as fuse-overlayfs was rewritten in rust we can no longer compile it like this. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit c4beb46) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com> Signed-off-by: Tim Zhou <timcoding1988@gmail.com> (cherry picked from commit e7a48b3) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com> Signed-off-by: Tim Zhou <timcoding1988@gmail.com> (cherry picked from commit 5bf07fd) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> (cherry picked from commit 755f4d7) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This makes for a much nicer web page, focusing on the actual tests. Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit 112166a) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit fa543cd) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
... and note that we should have it Renovate-managed. Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit b23fe32) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This should let us benefit from the latest diagnostics without having to manage a specific version. Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit b05c61d) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit 42797be) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This matches the previous situation + comment, but I don't know personally why this was done and I don't have a confirmation by the original author. Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit ee32565) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Should not change behavior. Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit 492b37b) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The original rationale no longer applies (our Debian images use a newer kernel) and it's not clear why we need this. Signed-off-by: Miloslav Trmač <mitr@redhat.com> (cherry picked from commit 0027eb9) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The images should have all we need already installed. Runtime installs are bad because they are flaky and can start failing all of the sudden. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 4bd0192) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Should include the sq binary for the image tests. Fixes: podman-container-tools#912 Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit b386240) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> (cherry picked from commit ef73911) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
With the modernize lint this is not needed anymore. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 319db77) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Do it like in podman and buildah and use a extra lima.yml to deduplicate the yml file. Also while at it touch up the task names, the name is set like the arguments to hack/ci/ci.yml like in podman so we can just copy paste this to reproduce any CI failure locally with that. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 54e86e4) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Mainly so we trigger it from the main ci.yml and work it in the task dependencies so we do not have to have extra tasks in the merge protection settings. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 0f2151f) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit ea28f61) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This should safe some test resources if basic validation fails already. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 5eadb77) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Because merge protection gates on Total Success we should have a job to ensure the yaml really has the right deps for it to work correctly. Also add a check to ensure all tasks are pinned with a version comment. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit ddbfa72) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The default github timeout is 6 hours which is way to long. In case there is ever a hang abort after 10 minutes as all these jobs should be quite fast. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit e53fa8c) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
All the jobs should be much faster than this so we do not needs such long timeouts. In case of hangs it is better to abort early. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 1b4b968) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This action has write access to PRs so we should guard against a compromised tag and only update explicitly via renovate PRs. Signed-off-by: Paul Holzinger <pholzing@redhat.com> (cherry picked from commit 726f5d0) Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Not a cherry-pick as it causes a lot more conflicts and this was a simple rm .github/workflows/opendownstream-pr.yml Upstream commit: 5b17ba4 Signed-off-by: Paul Holzinger <pholzing@redhat.com>
|
Ok I am a bit clueless for the image test, I have basically a one to one copy yet it works on main but fails here |
I have a local fix for the common problem but I do not understand what is going on with the skopeo tests here? |
|
@mtrmac would be nice if you could have a short look at the skopeo test failure if you know what the error is about. |
|
@Luap99 That looks like |
|
but that is the thing I do not get, I choose an skopeo image before that change went in 6f419aa here so I thought that should be compatible. I can certainly try a newer image and apply that skopeo patch |
|
I didn’t now inspect the exact contents of the image — note that the relevant point is not containers/automation_images#437 , but an earlier image update which updated the |
Yes sure, the image ID I picked is the one that is still being seen in the .cirrus.yml on the release-1.22 branch there which implies to me that image ID must have worked with that skopeo branch, at least until before cirrus was turned off. But maybe something is wrong with the copy logic here |
With lima we already run as rootless user so just run the tests as that user. Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Netavark v2 is not supported with the old podman 5.X branches so we should not test again it. Signed-off-by: Paul Holzinger <pholzing@redhat.com>
We have no supported it since 5.0 upstream, the latest ci image does not have all the deps for it so just remove it. Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This target does not exists on this branch and was only added later. Signed-off-by: Paul Holzinger <pholzing@redhat.com>
We need to test against the right skopeo branch, while at it update the skopeo git url to the new org. Signed-off-by: Paul Holzinger <pholzing@redhat.com>
|
@TomSweeneyRedHat @mtrmac PTAL Once this is merged #1007 can be rebased and have proper tests |
mtrmac
left a comment
There was a problem hiding this comment.
ACK, thanks!
“ci: do not use custom ssh user for image tests” could be forward-ported back to main I think.
yes that is on its way, I have to many branches at once right now |
cff8f4a
into
podman-container-tools:podman-5.8
|
I manually choose these three commits prior to the ci rework:
6f52244
f250115
c4beb46
And then the rest of the commits I got with
git log --format=%H e7a48b3642^..main -- .github/workflows/ hack/ci/ | tacfor cherry-picking to get all the stuff from main since the migration. With that I only had to solve some minor conflicts.Lastly more commits to make the new ci yaml check happy by removing .github/workflows/opendownstream-pr.yml, fixing some permission problem, fix the netavark tests and remove the cni tests, and then set the right skopeo branch for this container-libs version.