asserting that the tstate pointer is non-NULL using assert

[Wulian233]

Bug report

Bug description:

https://github.com/python/cpython/blob/main/Python/pystate.c#L3016

To prevent dangling pointer issues, adding a check for NULL ensures that invalid pointers are not dereferenced

Linked PRs

• gh-122928: asserting that the tstate pointer is non-NULL using assert #122929

[ZeroIntensity]

As mentioned in the PR, a NULL check doesn't fix dangling pointers.

With that being said, calling _PyThreadState_MustExit with an invalid thread state sounds like a bug, is there a specific case of this happening?

[neonene]

My understanding is that a problematic dangling pointer could be caused by the legacy non-isolated subinterpreter created with Py_NewInterpreter(), allowing daemon threads. Now that Python uses and recommends the new isolated subinterpreter, I do not think the concern is worth much consideration.

See also #109794 (comment), #104341 (comment).

[picnixz]

I'm wondering whether the issue should still mention dangling pointer since a dangling pointer is a valid pointer (hence non-null) pointing to invalid memory. It's more about asserting that the pointer is non-NULL using assert (namely, enforcing pre-conditions).

[neonene]

Please change the title, or make the PR belong to gh-109793. The current title is not good for closing here as "completed".

[vstinner]

The change is not needed: #122929 (review)