Another tiny test case

Check that PTHREAD_MUTEX_NORMAL mutexes will self-deadlock and that
pthread_mutex_timedlock() will block and then timeout on self-deadlock

Round-robin behavior isn't guaranteed by SCHED_OTHER and the kernel
doesn't have SCHED_FIFO or SCHED_RR support, so use sleep(1) to 'ensure'
the second thread gets the lock when expected

Add a test for recursive pthread_mutex_trylock()

ok guenther@

after mutex_unlock, there's no guarantee that a waiting thread will
start running. calling pthread_yield makes test pass with rthreads

-Wall cleanup.

various cleanups; david says results are same

Fix pthread_mutex regression test... undefined behavior does not
mean != EPERM in one test, minor changes to recursive mutex test.

remove printf in signal test signal handler, run test long enough
t

Fix pthread_mutex regression test... undefined behavior does not
mean != EPERM in one test, minor changes to recursive mutex test.

remove printf in signal test signal handler, run test long enough
to ensure signal hander is called at least twice, reduce output.

Note: running these tests over a remote console may indicate bogus
falures due to console output latency. Any test that takes longer
than 5 seconds to complete may be flagged as a failure.

show more ...

per man page and gcc 3.0.1, exit requires prototype from stdlib.h

Regression tests for libc_r (pthreads) library.
Thanks to pval@ for resolving the license stuff.