Debugging corrupt package failures

[kenodegard]

Description

Noticing an increase in corrupt package test failures:

• https://github.com/conda/conda-build/actions/runs/14114350528/job/39540950302
• https://github.com/conda/conda-build/actions/runs/14131729595/job/39593520975?pr=5654
• https://github.com/conda/conda-build/actions/runs/14114164270/job/39540208622?pr=5652

Checklist - did you ...

• Add a file to the news directory (using the template) for the next release's release notes?
• Add / update necessary tests?
• Add / update outdated documentation?

[kenodegard]

great now the failure has vanished from this PR...

[beckermr]

So the test jobs marked as "serial" are being run with two pytest-xdist workers. Is that maybe the issue?

[kenodegard]

interestingly all of the jobs are now cache hits

[kenodegard]

So the test jobs marked as "serial" are being run with two purest-xdist workers. Is that maybe the issue?

yeah that's been the case for a while now: #5078

I'd be surprised if that's the reason, plus we're seeing failures on both the "parallel" (tests/cli/test_main_build.py::test_slash_in_recipe_arg_keeps_build_id) and the "serial" (tests/test_api_build.py::test_recipe_builds[relative_url_download]) jobs

[beckermr]

Maybe add some retry test markers to have the test suite try twice as a stopgap?

[kenodegard]

We could add the retry but I don't understand how effective that would be since the corrupt package is in the cache (/home/runner/conda_pkgs_dir/) so a rerun would just reuse the corrupt package?

[beckermr]

Hrmmm. Can we try things actually in serial as a test?