In the discussion of the recent bad mesa update it was noted that running vkcube --validate showed the problem. However, it's not clear whether running it inside a non-accelerated VM is a useful test or not.
vkcube --validate
We should test (using the known-broken mesa) whether the check is useful in openQA context, and if it turns out it would've caught the bug, we should add a test for it.
@airlied let us know if you have any notes here - is running this check on llvmpipe in a qemu VM likely to be useful? Are there any other handy 'smoke test' commands we could run as well / instead?
Metadata Update from @adamwill: - Custom field story_points adjusted to 3
OK, I did the testing and it looks like this check would indeed have caught the bug even in a VM. vkcube --validate --c 10 with the bad mesa build causes a crash, with a good mesa build it runs and exits normally. So I'll go ahead and add a test.
vkcube --validate --c 10
https://pagure.io/fedora-qa/os-autoinst-distri-fedora/pull-request/465 adds this, but I found it actually fails on Rawhide, so we need to figure that out before we can land this.
This is merged now, though we may need to disable it on f42 as it seems to be failing there with only a warning message. I'll look into it some more.
Metadata Update from @adamwill: - Issue close_status updated to: Fixed - Issue status updated to: Closed (was: Open)
This issue has been migrated to Fedora Forge: https://forge.fedoraproject.org/quality/os-autoinst-distri-fedora/issues/464
Please continue any further discussion there.