[libcxx-commits] [libcxx] [libc++] fix `counting_semaphore` lost wakeups (PR #79265)
Jan Kokemüller via libcxx-commits
libcxx-commits at lists.llvm.org
Sat Feb 3 01:55:59 PST 2024
================
@@ -94,20 +95,25 @@ public:
auto __old = __a_.fetch_add(__update, memory_order_release);
_LIBCPP_ASSERT_ARGUMENT_WITHIN_DOMAIN(
__update <= _LIBCPP_SEMAPHORE_MAX - __old, "update is greater than the expected value");
-
- if (__old > 0) {
- // Nothing to do
- } else if (__update > 1)
+ (void)__old;
+ if (__update > 1) {
__a_.notify_all();
- else
+ } else {
__a_.notify_one();
+ }
}
_LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void acquire() {
auto const __test_fn = [this]() -> bool {
auto __old = __a_.load(memory_order_relaxed);
return (__old != 0) && __a_.compare_exchange_strong(__old, __old - 1, memory_order_acquire, memory_order_relaxed);
};
- __cxx_atomic_wait(&__a_.__a_, __test_fn);
+ auto const __test_with_old = [this](__cxx_contention_t& __monitor) -> bool {
----------------
jiixyj wrote:
> when this function is called, it is possible that the atomic variable has already been changed and __monitor is the new value. so we are waiting for a different value now. (and also there is no happens-before so __test_fn could be evaluated before taking the monitor). so basically we need to fix this function.
Agreed! But I don't think the `__test_fn` can be reordered before taking the monitor, as taking the monitor is a "acquire" operation?
https://github.com/llvm/llvm-project/pull/79265
More information about the libcxx-commits
mailing list