Bug#1057562: Some severities to reconsider due to flaky tests

[Jeremy Bícha]

On Wed, Apr 9, 2025 at 8:42 AM Santiago Vila <sanvila at debian.org> wrote:
> What happens in the buildds remains in the buildds. Whoever wants to
> rebuild the package from source will not do so in the official buildds.
>
> You are still violating Policy 4.2 for no particular reason.
>
> If build-time dependencies are specified, it must be possible to build the package and produce > working binaries on a system with only essential and build-essential packages installed and also those required to satisfy the build-time relationships (including any implied relationships).
>
> We should not rely on a test which segfaults to determine if the package
> is good or not. If the Release Managers still think this is ok
> "because it works in the buildds", I can still call the TC for
> your unwillingness to follow Debian Policy 4.2.

I do not believe Debian Policy is as precise on this detail as you
believe. It does not mandate that everything in Debian main be
buildable on a single CPU in an AWS instance.

> Do you need a patch for that?

I did help a bit in that trixie's debhelper by default now provides
verbose output for dh_auto_test even when passing. It looks like
something goes wrong in your builder at step 8 (
/gck/object/find_objects ) of the gcr:gck / object test. I am willing
to accept a patch to disable that subtest.

Thank you,
Jeremy Bícha