Fix unit tests

[schaefi]

This change is two fold

Fixed docopt docstring

When using doctopt argument information that contains the "|" symbol it gets wrongly evaluated as an option
in newer versions of docopt. This commit changes the argument information to use a "," instead

Fix unit tests

Due to us carrying a custom kiwi-revert-bls-default-for-suse.patch patch there is a different grub bootloader BLS behavior which causes some unit tests to fail when they run outside of the general source. This commit fixes it.

[Conan-Kudo]

Test builds running:

• F45: https://koji.fedoraproject.org/koji/taskinfo?taskID=143295668
• F44: https://koji.fedoraproject.org/koji/taskinfo?taskID=143295669
• F43: https://koji.fedoraproject.org/koji/taskinfo?taskID=143295670
• F42: https://koji.fedoraproject.org/koji/taskinfo?taskID=143295676

[Conan-Kudo]

The builds have failed.

[schaefi]

@Conan-Kudo There is little I can do here. The error from the build for x86_64 and aarch64 is

profile vmxFlavour not found for host arch i686

And that would actually be correct. This means in this mock environment the call for

import platform

print(platform.machine())

would return i686 and that is causing the test issues.

As such I believe something is wrong with the architecture emulation in the build environment
It's just an assumption though

[schaefi]

@Conan-Kudo wait maybe it's just me ... I took a closer look and there is some issue I overlooked.

Stay tuned, I'm on it

[schaefi]

@Conan-Kudo

urks this is a race condition in the parallel invocation of pytest. I see you are using 8 workers

PYTEST_XDIST_AUTO_NUM_WORKERS=8

I'm using 4 workers and in this scenario the problem does not happen. If I switch to 8 it fails for me in the same way. I believe this is because several tests mock the architecture to run an arch specific test. That arch mock is however global and if at the same time another test runs which does not work with this arch mock it fails.

I need to check where the race comes from :(

[Conan-Kudo]

Well, that explains why it doesn't fail on RHEL... We don't have the ability to run pytest in parallel there. 😅

[schaefi]

@Conan-Kudo ok, I think I found the race condition. Can you give it a new try ? Thanks

[Conan-Kudo]

Test builds running:

• F45: https://koji.fedoraproject.org/koji/taskinfo?taskID=143320527
• F44: https://koji.fedoraproject.org/koji/taskinfo?taskID=143320529
• F43: https://koji.fedoraproject.org/koji/taskinfo?taskID=143320532
• F42: https://koji.fedoraproject.org/koji/taskinfo?taskID=143320538

[schaefi]

Thanks, still some issues, checking

[schaefi]

@Conan-Kudo ok next try please and thanks for your patience. race conditions are hard to debug, but there was also a test that was really not well written from my side and I actually wondered why it did not fail more often.

[Conan-Kudo]

Test builds running:

• F45: https://koji.fedoraproject.org/koji/taskinfo?taskID=143322051
• F44: https://koji.fedoraproject.org/koji/taskinfo?taskID=143322056
• F43: https://koji.fedoraproject.org/koji/taskinfo?taskID=143322057
• F42: https://koji.fedoraproject.org/koji/taskinfo?taskID=143322058

[schaefi]

This looks promising :)

[Conan-Kudo]

Looks like everything passes now.

[schaefi]

@Conan-Kudo yes can you approve, then I'll zip a new release and you will be unblocked.