RHEL8 release MK.2

[evgenyz]

This is a replacement for #121. On top of the changes in that PR we have here:

• Fully functional Packit builds for EPEL8/9
• Test matrix for all Fedoras, with a smoke test.
• FMF/TMT integration for ./tests/smoke.sh test (former test_of_the_basic_function.sh)
• Packit propose_downstream integration for fedora-all, epel-9, epel-8 (yet to be tested).

[evgenyz]

@matusmarhefka Would this fmf smoke test be enough as a starter for RHEL8/RHEL9? Do we need to adjust the config in some way? Could we be able to re-use the upstream test downstream?

[matusmarhefka]

@matusmarhefka Would this fmf smoke test be enough as a starter for RHEL8/RHEL9? Do we need to adjust the config in some way? Could we be able to re-use the upstream test downstream?

Sure and we can definitely re-use it in downstream. I see we are not testing the EPEL builds, what about testing them on CentOS Stream 8/9? It can be done by updating the job for tests similarly to https://github.com/ComplianceAsCode/content/blob/master/.packit.yaml#L27, so something like this should work:

- job: tests
  trigger: pull_request
  identifier: epel-tests
  skip_build: true
  targets:
    - epel-8
        distros: [centos-stream-8]
    - epel-9
        distros: [centos-stream-9]

[evgenyz]

@matusmarhefka Do you have any idea why epel-8 test is failing and epel-9 is not?

[Honny1]

@evgenyz I think the epel-* tests (including Fedora) will install the released package. The problem is probably caused by skip_build: true. The test uses the available released package. So it does not use the RPMs created in copr, so the change in PR will not be propagated in the used RPM. And epel-8 fails because openscap-report is not available in the EPEL8 repository.

This row in epel-9 log is proof.

10:09:09             out:  openscap-report       noarch  0.2.1-1.el9   epel                         569 k

[evgenyz]

Okay, the way Packit see it, if skip_build: true is provided then package installation is supposed to be part of tmt definition. The skip_build: true config option is a workaround for a completely different use-case, when the package is coming from a compose (or somewhere else rather than packit build process).

I propose to kick out the epel-8 config for Packit until they get rid of implicit builds for tests job. I suppose that package_dependency_test is enough for upstream testing.

[Honny1]

Why are the tests using RPM packages that have already been released? The RPM package used does not contain the latest changes that were introduced in the PR where the tests are running. Or is it a workaround for implicit builds for test jobs? If yes, kick out the epel-8 from config.