Changeset 8600

Timestamp:

10/27/10 17:20:09 (3 years ago)

Author:

bligon

Message:

In function dbpf_sync_context_destroy(), helgrind was complaining that we shouldn't
try to destory a mutex after it's been locked. So, I added an unlock before the
destroy.

Files:

• branches/Orange-Branch/src/io/trove/trove-dbpf/dbpf-sync.c (modified) (1 diff)

branches/Orange-Branch/src/io/trove/trove-dbpf/dbpf-sync.c

@@ -91 +91 @@
     for(c=0; c < COALESCE_CONTEXT_LAST; c++)
     {
+        /* grab lock...should be the last one, since we are shutting down */
         gen_mutex_lock(&sync_array[c][context_index].mutex);
+      
+        /* we have to unlock the mutex before we can destroy it */
+        gen_mutex_unlock(&sync_array[c][context_index].mutex);
+
+        /* destroy the mutex */
         gen_mutex_destroy(&sync_array[c][context_index].mutex);
+
+        /* cleanup the op queue */
         dbpf_op_queue_cleanup(sync_array[c][context_index].sync_queue);
     }