[PATCH] nptl: Use a PI-aware lock for internal pthread_cond-related locking

[Florian Weimer]

* Adhemerval Zanella Netto:

>> Does POSIX define what happens if a non-robust mutex is accessed again
>> if a thread terminates without unlocking it first?  If yes, then this is
>> indeed a problem.
>
> Afaik this is UB for non-robust mutexes, although the resolution for Austin
> Issue 755  [1] is not clear IMHO.
>
> [1] https://www.austingroupbugs.net/view.php?id=755#c1875

Ah, so it's UB in practically all implementations (for at least some
mutex types), but not UB in POSIX.

Thanks,
Florian