Read _expire_at only once in has_expired()

[mbeijen]

Backport of encode/httpcore#1090.

Summary

In has_expired() the _expire_at attribute was read twice: once for the is not None guard and once for the comparison. On free-threaded builds another thread may reset it to None between the two reads, raising a TypeError. Read it once into a local instead.

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}

[codspeed-hq]

Merging this PR will not alter performance

✅ 15 untouched benchmarks
⏩ 7 skipped benchmarks¹

---

_{Comparing mbeijen:fix-expire-at-race (ff2953e) with main (82b9e2d)}

Footnotes

1. 7 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[Kludex]

And is locking a bad idea here?

[mbeijen]

And is locking a bad idea here?

I don't think locking helps here. has_expired() is sync but _state_lock is an AsyncLock, so locking would mean making it (and is_available/is_idle and their pool callers) async which is big change for little gain. These are meant to be cheap lock-free snapshots anyway.

The only real bug was reading _expire_at twice and racing with the = None reset on free-threaded builds, which crashes