Created attachment 155099 [details]
Code snippet copied from stackoverflow and fixed to compile

SUMMARY

Destroying a std::condition_variable after calling notify_all() results in a helgrind warning "pthread_cond_destroy: destruction of condition variable being waited upon", even though this behaviour is apparently allowed by the C++ spec (http://en.cppreference.com/w/cpp/thread/condition_variable/~condition_variable).

See this stackoverflow question (https://stackoverflow.com/questions/48093715/destruction-of-condition-variable-randomly-loses-notification) for a full discussion, although only the posts by the OP and the user mevets are worth reading. Ultimately it seems the program is correct, although the discussion is contentious which is why I've included "potential" in the bug report title.

STEPS TO REPRODUCE

Run the snippet in the stackoverflow question (also attached) under helgrind.

OBSERVED RESULT

The error message

==71731== Thread #1: pthread_cond_destroy: destruction of condition variable being waited upon
==71731==    at 0x484AA83: ??? (in /usr/libexec/valgrind/vgpreload_helgrind-amd64-linux.so)
==71731==    by 0x4959E3D: __gthread_cond_destroy (gthr-default.h:885)
==71731==    by 0x4959E3D: ~__condvar (std_mutex.h:141)
==71731==    by 0x4959E3D: std::condition_variable::~condition_variable() (condition_variable.cc:36)
==71731==    by 0x1098DB: notify_on_delete::~notify_on_delete() (in a.out)
==71731==    by 0x109381: main (in a.out)

EXPECTED RESULT

No error messages.

SOFTWARE/OS VERSIONS

valgrind-3.20.0.GIT

Target: x86_64-pc-linux-gnu
gcc version 12.2.0 (GCC)