From: Niels Provos <provos@gmail.com>
Date: Thu, 19 Nov 2009 22:02:33 +0000 (+0000)
Subject: nick found a race condition in the pthreads test case
X-Git-Tag: release-2.0.3-alpha~1
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=986500de4db591134832923f5dd8d78477a62f2e;p=libevent

nick found a race condition in the pthreads test case

svn:r1554
---

diff --git a/ChangeLog b/ChangeLog
index be01b27b..7b2f34ea 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -52,6 +52,7 @@ Changes in 2.0.3-alpha:
  o Finally expose the IOCP-based bufferevent backend.  It passes its unit tests, but probably still has some bugs remaining.  Code by Nick Mathewson and Christopher Davis.
  o Numerous other bugfixes.
  o On FreeBSD and other OSes, connect can return ECONREFUSED immediately; instead of failing the function call, pretend with faileld in the callback.
+ o Fix a race condition in the pthreads test case; found by Nick Mathewson
 
 Changes in 2.0.2-alpha:
  o Add a new flag to bufferevents to make all callbacks automatically deferred.
diff --git a/test/regress_pthread.c b/test/regress_pthread.c
index acfd4327..bc611d0f 100644
--- a/test/regress_pthread.c
+++ b/test/regress_pthread.c
@@ -73,9 +73,11 @@ basic_thread(void *arg)
 	for (i = 0; i < 100; i++) {
 		struct timeval tv;
 		evutil_timerclear(&tv);
-		assert(evtimer_add(&ev, &tv) == 0);
 
 		assert(pthread_mutex_lock(&cw.lock) == 0);
+		/* we need to make sure that even does not happen before
+		 * we get to wait on the conditional variable */
+		assert(evtimer_add(&ev, &tv) == 0);
 		assert(pthread_cond_wait(&cw.cond, &cw.lock) == 0);
 		assert(pthread_mutex_unlock(&cw.lock) == 0);