gh-94673: Ensure subtypes are readied only once in math.trunc()

[neonene]

Fixes a typo in d2e2e53.
cc: @ericsnowcurrently

• Issue: [subinterpreters] Static types incorrectly share some objects between interpreters #94673

[ericsnowcurrently]

LGTM

(good catch!)

[miss-islington]

Thanks @neonene for the PR, and @ericsnowcurrently for merging it 🌮🎉.. I'm working now to backport this PR to: 3.12.
🐍🍒⛏🤖

[bedevere-bot]

GH-105471 is a backport of this pull request to the 3.12 branch.

[mdickinson]

Out of curiosity, why is the code for math_trunc different from that for math_floor and math_ceil? Why does math_trunc need the PyType_Ready call, but not math_floor or math_ceil?

[ericsnowcurrently]

Hmm, I have a guess. That code was added in 2007 (commit 15d3d04) as a builtin (bltinsmodule.c). In 2008 the function was moved to the math module. So that's probably the difference.

[mdickinson]

@ericsnowcurrently Nice find; that would make sense. (Well, maybe; I still can't quite figure out why the PyType_Ready would have been necessary, but it's a little more plausible that built-in functions could be being executed before some types are fully initialised.)

Do you see any reason not to just get rid of this code in mathmodule.c?

[ericsnowcurrently]

I suspect that code is an artifact of a time when folks were inconsistent about calling PyType_Ready(). However, I'm not sure what the actual status quo is at this point. This might be a good thing to ask about on python-dev.