gh-109959: attempt to fix hypothesis tests for test_glob

[picnixz]

Just trying to see if this approach would work or not. Tests are likely to be flaky but I want at least to isolate the failure if possible...

• Issue: test_glob: test_selflink() fails randomly on Linux #109959

[picnixz]

I've rerun lots of times the hypothesis workflow and didn't encounter issues but I really don't know if this is meaningful. @barneygale WDYT?

[barneygale]

I know I speculated that this area of code could be related, but on reflection I think it's very unlikely. Given the behaviour Alyssa analysed, if this code were related, then only a bug in itertools.chain() could explain it, and why would such a bug show up only here? I think the issue must elsewhere in glob.py.

[picnixz]

What's also really weird is that I'm locally never seeing this failure. I don't know if this may be something related to the machine the hypothesis tests are actually running.

[picnixz]

After Victor's investigation, it appears to be an OS issue itself. I'm going to close this one as I don't think I can come up with a good fix (though, is it possible to rerun it automatically if it failed, perhaps with a few exponential backoff strategy to avoid a busy host due to other tests running?)

[vstinner]

The root issue is in the Linux kernel: #109959 (comment)

I don't think that changing iglob() is going to work around the issue.

[bedevere-app]

When you're done making the requested changes, leave the comment: I have made the requested changes; please review again.

[picnixz]

Ok so nothing I can do except skipping the test as we discussed on Discord. Closing in favor of #128812 then.