gh-117657: Quiet erroneous TSAN reports of data races in _PySeqLock

[mpage]

TSAN reports a couple of data races between the compare/exchange in _PySeqLock_LockWrite:

cpython/Python/lock.c

Line 483 in a23fa33

else if (_Py_atomic_compare_exchange_uint32(&seqlock->sequence, &prev, prev + 1)) {

and the non-atomic loads in _PySeqLock_AbandonWrite (example race):

cpython/Python/lock.c

Lines 493 to 495 in a23fa33

	void _PySeqLock_AbandonWrite(_PySeqLock *seqlock)
	{
	uint32_t new_seq = seqlock->sequence - 1;

and _PySeqLock_UnlockWrite (example race):

cpython/Python/lock.c

Lines 500 to 502 in a23fa33

	void _PySeqLock_UnlockWrite(_PySeqLock *seqlock)
	{
	uint32_t new_seq = seqlock->sequence + 1;

This is another instance of TSAN incorrectly modeling failed compare/exchange as a write instead of a load.

• Issue: Make TSAN tests pass with the GIL disabled in free-threaded builds #117657

[DinoV]

LGTM!

[cpython-cla-bot]

All commit authors signed the Contributor License Agreement.