* barrier.c: Remove static initialisation - irrelevent

	for this object.
	* pthread.h (PTHREAD_BARRIER_INITIALIZER): Removed.
	* rwlock.c (pthread_rwlock_wrlock): This routine is
	not a cancelation point - disable deferred
	cancelation around call to pthread_cond_wait().

tests/ChangeLog:
	* spin1.c: New; testing spinlocks.
	* spin2.c: New; testing spinlocks.
	* spin3.c: New; testing spinlocks.
	* spin4.c: New; testing spinlocks.
	* barrier1.c: New; testing barriers.
	* barrier2.c: New; testing barriers.
	* barrier3.c: New; testing barriers.
	* barrier4.c: New; testing barriers.
	* GNUmakefile: Add new tests.
	* Makefile: Add new tests.

1 files changed, 65 insertions, 0 deletions

diff --git a/tests/spin4.c b/tests/spin4.c
new file mode 100644
index 0000000..a435d04
--- /dev/null
+++ b/tests/spin4.c
@@ -0,0 +1,65 @@
+/* 
+ * spin4.c
+ *
+ * Declare a spinlock object, lock it, spin on it, 
+ * and then unlock it again.
+ *
+ * For this to work on a single processor machine we have
+ * to static initialise the spinlock. This bypasses the
+ * check of the number of processors done by pthread_spin_init.
+ * This is a non-portable side-effect of this implementation.
+ */
+
+#include "test.h"
+#include <sys/timeb.h>
+ 
+pthread_spinlock_t lock = PTHREADS_SPINLOCK_INITIALIZER;
+struct _timeb currSysTimeStart;
+struct _timeb currSysTimeStop;
+
+#define GetDurationMilliSecs(_TStart, _TStop) ((_TStop.time*1000+_TStop.millitm) \
+                                               - (_TStart.time*1000+_TStart.millitm))
+
+static int washere = 0;
+
+void * func(void * arg)
+{
+  _ftime(&currSysTimeStart);
+  assert(pthread_spin_lock(&lock) == 0);
+  assert(pthread_spin_unlock(&lock) == 0);
+  _ftime(&currSysTimeStop);
+  washere = 1;
+
+  return (void *) GetDurationMilliSecs(currSysTimeStart, currSysTimeStop);
+}
+ 
+int
+main()
+{
+  long result = 0;
+  pthread_t t;
+
+  assert(pthread_spin_lock(&lock) == 0);
+
+  assert(pthread_create(&t, NULL, func, NULL) == 0);
+
+  /*
+   * This should relinqish the CPU to the func thread enough times
+   * to waste approximately 2000 millisecs only if the lock really
+   * is spinning in the func thread (assuming 10 millisec CPU quantum).
+  for (i = 0; i < 200; i++)
+    {
+      sched_yield();
+    }
+
+  assert(pthread_spin_unlock(&lock) == 0);
+
+  assert(pthread_join(t, (void *) &result) == 0);
+  assert(result > 1000);
+
+  assert(pthread_spin_destroy(&lock) == 0);
+
+  assert(washere == 1);
+
+  return 0;
+}