I think we probably need the scratch-build test to ensure that the packages being verified are ALSO included in the buildroot for that scratch-build. This would have prevented the issue we hit today, since the fedora-release build would then have failed in gating and never been pushed to production, breaking the buildroot for everyone else.
(If this should be filed elsewhere, please let me know)
When do you need this? (YYYY/MM/DD) No strict timeline
When is this no longer needed or useful? (YYYY/MM/DD) The heat-death of the universe or the termination of the Fedora Project package gating system, whichever comes first.
If we cannot complete your request, what is the impact? We will continue to encounter avoidable buildroot breakages like the one above.
See e.g. https://src.fedoraproject.org/rpms/redhat-rpm-config/blob/rawhide/f/tests/build-self/runtest.sh for how we have done it in redhat-rpm-config
Maybe I'm missing something, but I don't see how that test does what I'm asking for.
Metadata Update from @phsmoura: - Issue tagged with: low-gain, low-trouble, ops
So, I am not sure what releng can do here. ;)
Perhaps this should be against centos-ci? Or perhaps @adamwill has some ideas for openqa?
Or we just need to adjust what ci we do on fedora-release?
Metadata Update from @kevin: - Issue untagged with: low-gain, low-trouble, ops