Re: Patch: POSIX update for glibc 2.1

[Matt Welsh <mdw@cs.berkeley.edu>]

Tom Tromey <tromey@cygnus.com> writes:
>  #elif defined (PTHREAD_MUTEX_HAVE___M_COUNT)
> -  int r = pmu->__m_count == 1;
> +  int r = (pthread_t) pmu->__m_owner == pthread_self ();

I'm not sure this is quite right, since __m_owner will still be
pthread_self() if the mutex was unlocked before entering this
routine.

The correct check is 
	((pmu->__m_owner == pthread_self()) && (pmu->__m_count == 0))

which means that PthreadCheckMonitor was the first to lock the mutex.

Still, this is ugly and possibly slow, depending on how pthread_self()
is implemented. In the long run it's probably best to have our own 
"fast check" for a mutex being locked or not. 

Matt Welsh