[Yasser-Ameur]

NOTE Erasing or replacing the contents of this template will result in your pull
request being summarily closed without consideration!

Changes

Fixes #.

TemporaryDirectory.cleanup() runs its cleanup through to_thread.run_sync, which
checkpoints on entry. When the host task is already cancelled, and it usually is,
because cleanup() gets awaited from a finally: block during unwind, the checkpoint
aborts before the worker thread runs and the directory is left on disk.

#1304 fixed exactly this for __aexit__ with a shielded CancelScope. cleanup(), a
few lines below, kept the unshielded call. This gives it the same shield. The other
to_thread.run_sync teardown in the module, AsyncFile.aclose(), was already shielded
by #1314, so this was the last one.

test_cleanup_method_with_cancellation mirrors the test #1304 added for __aexit__.
It fails on the unmodified tree on all four backend variants and passes with the fix.

Checklist

If this is a user-facing code change, like a bugfix or a new feature, please ensure that
you've fulfilled the following conditions (where applicable):

• You've added tests (in tests/) which would fail without your patch
• You've updated the documentation (in docs/), in case of behavior changes or new
  features
• You've added a new changelog entry (in docs/versionhistory.rst).

[Yasser-Ameur]

Good catch, you're right. I've added await checkpoint_if_cancelled() right after the shielded block, so a pending cancellation is still delivered to the caller once the directory is safely removed, matching the shield-then-checkpoint pattern already used elsewhere in this file. I extended the regression test to assert both that the directory is gone and that the cancellation is still raised, and confirmed the new assertion fails without the checkpoint. __aexit__ has the same shield without a checkpoint after it; I left it alone since the question was about cleanup(), but I can add it in this PR if you'd like.

[agronholm]

Thanks.