Bug report

Bug description:

When cibuildwheel runs the musllinux_1_2_i686 configuration, I often see exceptions logged when the end-of-process GC runs:

Exception ignored in: <function _DeleteDummyThreadOnDel.__del__ at 0xf2be5578>
  Traceback (most recent call last):
    File "/opt/_internal/cpython-3.13.0/lib/python3.13/threading.py", line 1383, in __del__
  TypeError: 'NoneType' object does not support the context manager protocol
  .Traceback (most recent call last):
    File "<string>", line 1, in <module>
      from tornado.ioloop import IOLoop; classname = lambda x: x.__class__.__name__; from tornado.platform.asyncio import AsyncIOMainLoop; AsyncIOMainLoop().install(); print(classname(IOLoop.current()))

It looks like something about musl libc is causing _active_limbo_lock to get GC'd before _thread_local_info._track_dummy_thread_ref. This happens on musllinux (1.2) on i686 builds. 64-bit builds (of x86 and arm) and manylinux builds are not affected. Older versions of python are not affected; I have not tested with 3.14 alphas but I think they are likely affected since the relevant code doesn't appear to have changed.

A complete log with this failure can be found at https://github.com/bdarnell/tornado/actions/runs/13459578244/job/37733273132

This issue is not a priority for me. I'm just going to skip this configuration in my CI and I'm filing an issue so I have something to point to.

CPython versions tested on:

3.13

Operating systems tested on:

Linux

Linked PRs

• gh-130522: Fix threading errors during garbage collection #131537
• [3.14] gh-130522: Fix unraisable TypeError in threading at interpreter shutdown (GH-131537) #137105
• [3.13] gh-130522: Fix unraisable TypeError in threading at interpreter shutdown (GH-131537) #137106