Add test to run basic qm image build with a-i-b

[mkemel]

Introducing a new sanity test, that builds a basic qm image with automotive-image-builder.

Resolves: #675

[dougsland]

looks sane to me @Yarboa PTAL

[Yarboa]

@mkemel
AFAICT, I see that no gate run this test.
Few questions here:

• In general the test itself just build qm qcow image. is it?
• Does it check that vm also running?
• Do you have dependency in /dev/kvm?
• Dont you think it should be part of automotive-sig?

One more thing here, What the tests should really test?
Is it a new QM image in autosd image with latest qm image?

[Yarboa]

please refer this
#697 (comment)

[mkemel]

Hi @Yarboa

1. Yes
2. No
3. Not that I'm aware of, don't think that osbuild has such a dependency.
4. The test should verify that a PR does not break QM compatibility with a-i-b and the build runs successfully. So it should be in QM.

One question from my side is - the requirement is for CI only. So maybe it would be better to create a separate plan for that?

[mkemel]

@alexlarsson care to take a look?

[Yarboa]

Hi @Yarboa

1. Yes
2. No
3. Not that I'm aware of, don't think that osbuild has such a dependency.

As per this https://github.com/osbuild/osbuild?tab=readme-ov-file#requirements
only qemu-image, ack

4. The test should verify that a PR does not break QM compatibility with a-i-b and the build runs successfully. So it should be in QM.
   So i thing there is an issue here, I will review PR again

And please add better description

One question from my side is - the requirement is for CI only. So maybe it would be better to create a separate plan for that?

The issue here is not the plan it is packit,
I would creatye tests with special hardware in pacvklit.yaml

[mkemel]

The issue here is not the plan it is packit,
I would creatye tests with special hardware in pacvklit.yaml

Currently the test is run by packit as part of this job:

  - job: tests
    trigger: pull_request
    identifier: qm-tier-0
    tmt_plan: /plans/e2e/tier-0
    targets:
      - epel-9-x86_64
    tf_extra_params:
      environments:
        - artifacts:
          - *bluechi_copr_repo
          hardware:
            disk:
             - size: ">= 20 GB"

And the only hardware requirement is >= 20GB, and the test does not require more than that. But do we want this test have its own job?

[Yarboa]

tmt_plan: /plans/e2e/tier-0

I do not think it is related,
All tests running are running on QM environment, and its prepare step, set QM env
Current test, have no relation with qm deployments, just verify qm image is build success with qm rpm PR generated.

Isnt it?

[mkemel]

I do not think it is related,
All tests running are running on QM environment, and its prepare step, set QM env
Current test, have no relation with qm deployments, just verify qm image is build success with qm rpm PR generated.

Yes. I wasn't sure if I should just add this one to sanity tests (this is a sanity test after all), and run it in an environment with QM (which does not affect this test), or create a separate plan and run it as a separate job. I chose the former in avoid creating separate plan and job for one test. But if you think that it would be better to have a separate job for this one - I can add one, it's not a problem.

[Yarboa]

I do not think it is related,
All tests running are running on QM environment, and its prepare step, set QM env
Current test, have no relation with qm deployments, just verify qm image is build success with qm rpm PR generated.

Yes. I wasn't sure if I should just add this one to sanity tests (this is a sanity test after all), and run it in an environment with QM (which does not affect this test), or create a separate plan and run it as a separate job. I chose the former in avoid creating separate plan and job for one test. But if you think that it would be better to have a separate job for this one - I can add one, it's not a problem.

Ack, ask for different machine with specific plan+test running, there is no prepare step here,
it needs only c9s machine with minimum of memory and storage, in case of no nested kvm, no need to ask it from testing farm

[Yarboa]

The issue here is not the plan it is packit,
I would creatye tests with special hardware in pacvklit.yaml

Currently the test is run by packit as part of this job:

  - job: tests
    trigger: pull_request
    identifier: qm-tier-0
    tmt_plan: /plans/e2e/tier-0
    targets:
      - epel-9-x86_64
    tf_extra_params:
      environments:
        - artifacts:
          - *bluechi_copr_repo
          hardware:
            disk:
             - size: ">= 20 GB"

And the only hardware requirement is >= 20GB, and the test does not require more than that. But do we want this test have its own job?

better to maintain it separately in case it is broken, gate run also on bare c9s/fedora, it should not block merge of non related PRs

[dougsland]

Worth add for reference this comment and discussion: #674 (comment) I guess this is the root cause for this PR.

[dougsland]

If @michalskrivanek @Yarboa are satisfied we can proceed. I trust in their review.

[Yarboa]

If @michalskrivanek @Yarboa are satisfied we can proceed. I trust in their review.

I thougt there is agreement on that
#697 (comment), waiting for another update

BTW, it could be moved to qm-next also

[michalskrivanek]

well, I'm NOT happy with copying the logic from auto-image-builder.sh. It even changed in the meantime...

[mkemel]

1. Created separate plan and Packit job to run it
2. Now use auto-image-builder.sh

[mkemel]

/packit retest-failed

[Yarboa]

LGTM

[Yarboa]

@mkemel please check precommit errors
https://github.com/containers/qm/actions/runs/12790524518/job/35695989011?pr=697

[mkemel]

@Yarboa It took me some time to make this work, the whole quotes thing in this script is quite complicated since we want to pass a mix of single and double quotes to a shell script which passes it to podman which passes it to container. Suppressed the warnings, shellcheck does not really recognize what I'm doing there.

[Yarboa]

@Yarboa It took me some time to make this work, the whole quotes thing in this script is quite complicated since we want to pass a mix of single and double quotes to a shell script which passes it to podman which passes it to container. Suppressed the warnings, shellcheck does not really recognize what I'm doing there.

Did you try this one?
EXTRA_REPOS="extra_repos=[{\"id\":\"qm_build\",\"baseurl\":\"$COPR_URL\"}]"

And that one?
/bin/bash auto-image-builder.sh build --export qcow2 --define "$EXTRA_REPOS" qm.aib.yml qm.qcow2

this one was passing for me
find -name '*.sh' -exec shellcheck --severity=error -f gcc {} \;

[mkemel]

Did you try this one?
EXTRA_REPOS="extra_repos=[{"id":"qm_build","baseurl":"$COPR_URL"}]"

This would work, but I like it less tbh, I think using single quote is much better than escaping every double quote, it makes reading much more difficult to my taste

And that one?
/bin/bash auto-image-builder.sh build --export qcow2 --define "$EXTRA_REPOS" qm.aib.yml qm.qcow2

This wouldn't work because we need single quotes there