bpo-33612: Remove PyThreadState_Clear() assertion

[vstinner]

bpo-25612, bpo-33612: Remove an assertion from PyThreadState_Clear()
which failed at Python shutdown or on fork if a thread was running a
generator.

https://bugs.python.org/issue33612

[markshannon]

LGTM.
The tests tstate->exc_info->previous_item == NULL and tstate->exc_info == &tstate->exc_state are equivalent, so the assert is redundant if correct and needlessly causes a crash when wrong.

[serhiy-storchaka]

I agree, it is safe to clear just tstate->exc_state and remove the assertion.

[vstinner]

I agree, it is safe to clear just tstate->exc_state and remove the assertion.

Wait, do you mean that exc_state and/or exc_info must be explicitly cleared? These fields are pointers to data hold by a generator. The best that we can do is to set these pointers to NULL, but it is very likely the the whole tstate memory is going to be freed anyway, so I'm not sure that it's useful.

I'm not sure that I understood you correctly. Do you think that the current PR is correct? Or do we need extra changes?

[serhiy-storchaka]

I think that the current PR is correct.

exc_state is cleared few lines above, exc_info is a link to the state owned by the generator, and doesn't need any action. But setting it to &exc_state wouldn't make the code worse.

[bedevere-bot]

@vstinner: Please replace # with GH- in the commit message next time. Thanks!

[miss-islington]

Thanks @vstinner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.7.
🐍🍒⛏🤖

[bedevere-bot]

GH-7074 is a backport of this pull request to the 3.7 branch.

[vstinner]

exc_state is cleared few lines above

exc_state fields are cleared, not exc_state itself.

exc_info is a link to the state owned by the generator, and doesn't need any action. But setting it to &exc_state wouldn't make the code worse.

I don't feel confortable to make this change. Please write a separated PR if you want to do that.