gh-132413: Check if datetime current state is not NULL

[sergey-miryanov]

If it is a NULL then interpreter shutting down and we can do nothing

• Issue: SIGSEV in datetime.timedelta (possibly from datetime's C delta_new) #132413

[sergey-miryanov]

@neonene I cherry-picked and squashed your commits with test. If you don't like this - I will revert/remove it.

[StanFromIreland]

I presume it was closed because of this PR they opened, #136152, which seems like a more concrete approach.

[sergey-miryanov]

I presume it was closed because of this PR they opened, #136152, which seems like a more concrete approach.

Ough! My oversight! Thanks!

[StanFromIreland]

Looking deeper that one is also most likely superseded by #136189

[sergey-miryanov]

Looking deeper that one is also most likely superseded by #136189

IIUC #136189 doesn't solve #132413 by now.

[sergey-miryanov]

Closing in favor of #136152

[neonene]

This solution raises a ImportError rather than a crash, which is a breaking change. And the test needs to check stderr (subprocess exits with OK). I think this PR could be used as a last resort when the issue get stuck.

[sergey-miryanov]

This solution raises a ImportError rather than a crash, which is a breaking change.

Yeah, but I believe that exception (swallowed in this case) better than a crash. But I hope your PR will be approved and merged soon.