gh-151728: Clear the typing caches at interpreter shutdown

[wjakob]

typing caches every subscripted type in module-level lru_cachees. Those caches are only released when the typing module itself is torn down, which does not happen when some extension module leaks a reference to it. Since typing is nearly universal, a single such leak (e.g. via torch) keeps the cached types alive past interpreter shutdown, including types owned by unrelated, correctly written extensions. Those extensions then look like they are leaking: valgrind reports them, and nanobind prints warnings about types, functions, and instances that were never freed.

typing._cleanups already holds callables to clean caches but does nothing with them. This PR registers an atexit callback that triggers them during Python finalization so that the process is leak-free upon termination. This complements gh-98253, which broke a cycle inside _tp_cache but left the retention itself in place.

The PR adds no tests because reproducing the problem requires an extension module that deliberately leaks a reference (i.e. one with a missing tp_traverse). A minimal one is at https://github.com/wjakob/typing-leak.

• Issue: typing caches keep extension types alive past shutdown #151728

[read-the-docs-community]

Documentation build overview

📚 cpython-previews | 🛠️ Build #33810774 | 📁 Comparing a191941 against main (9ccd5bb)

  🔍 Preview build  

295 files changed · + 1 added · ± 284 modified · - 10 deleted

+ Added

• library/asyncio-policy.html

± Modified

• bugs.html
• glossary.html
• index.html
• license.html
• py-modindex.html
• c-api/allocation.html
• c-api/buffer.html
• c-api/bytes.html
• c-api/code.html
• c-api/complex.html
• and 274 more...

- Deleted

• c-api/sentinel.html
• c-api/slots.html
• deprecations/pending-removal-in-3.21.html
• deprecations/soft-deprecations.html
• howto/abi3t-migration.html
• library/asyncio-threading.html
• library/asyncio-tools.html
• library/tkinter.fontchooser.html
• library/tkinter.systray.html
• whatsnew/3.16.html

[encukou]

Looks reasonable.
Now that there's a function around the loop, could you call it in Lib/test/libregrtest/utils.py (in clear_caches, near the end)?

[wjakob]

Done.

[ZeroIntensity]

Quick thought: Should we perhaps add this to sys._clear_internal_caches, and then call that at shutdown instead?

If not, this looks good to me anyway.

[encukou]

Should we perhaps add this to sys._clear_internal_cache

sys is for interpreter internals, I don't think it should be responsible for cleaning up typing.
Maybe we should have a generic module cleanup mechanism, but then again, atexit should be enough.