Don't call lookdict_index in delitemif_lock_held

[colesbury]

The delitemif_lock_held function calls lookdict_index and passes the returned hashpos to delitem_common. We should just pass the actual hash to delitem_common.

The existing code is confusing and unnecessary, but it's not exactly a bug. The behavior is fine because it's safe in this case to pretend the hashtable index (hashpos) is a hash value. The resulting size_t i = (size_t)hash & mask to compute the index from the hash is effectively a no-op.

cpython/Objects/dictobject.c

Lines 2635 to 2645 in 069de14

	hashpos = lookdict_index(mp->ma_keys, hash, ix);
	assert(hashpos >= 0);
	if (res > 0) {
	PyInterpreterState *interp = _PyInterpreterState_GET();
	uint64_t new_version = _PyDict_NotifyEvent(
	interp, PyDict_EVENT_DELETED, mp, key, NULL);
	return delitem_common(mp, hashpos, ix, old_value, new_version);
	} else {
	return 0;
	}

cpython/Objects/dictobject.c

Lines 2502 to 2504 in 069de14

	static int
	delitem_common(PyDictObject *mp, Py_hash_t hash, Py_ssize_t ix,
	PyObject *old_value, uint64_t new_version)

cpython/Objects/dictobject.c

Lines 995 to 1000 in 069de14

	static Py_ssize_t
	lookdict_index(PyDictKeysObject *k, Py_hash_t hash, Py_ssize_t index)
	{
	size_t mask = DK_MASK(k);
	size_t perturb = (size_t)hash;
	size_t i = (size_t)hash & mask;

Linked PRs

• gh-117826: Remove lookdict_index from delitemif_lock_held #117869

[corona10]

@colesbury Now we can close the issue?

[colesbury]

Yes, thanks for fixing this @corona10