still fighting the same bug. now i must have killed it :)

(unfortunately it takes hours of testing to reproduce...)

CVS patchset: 5341
CVS date: 2003/09/04 13:56:23

1 files changed, 13 insertions, 2 deletions

diff --git a/src/xine-engine/audio_out.c b/src/xine-engine/audio_out.c
index be37f71ac..0505b440f 100644
--- a/src/xine-engine/audio_out.c
+++ b/src/xine-engine/audio_out.c
@@ -17,7 +17,7 @@
  * along with self program; if not, write to the Free Software
  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA
  *
- * $Id: audio_out.c,v 1.143 2003/09/04 00:21:09 miguelfreitas Exp $
+ * $Id: audio_out.c,v 1.144 2003/09/04 13:56:23 miguelfreitas Exp $
  *
  * 22-8-2001 James imported some useful AC3 sections from the previous alsa driver.
  *   (c) 2001 Andy Lo A Foe <andy@alsaplayer.org>
@@ -1681,6 +1681,9 @@ static void ao_flush (xine_audio_port_t *this_gen) {
     
     /* do not try this in paused mode */
     while( this->flush_audio_driver ) {
+      struct timeval  tv;
+      struct timespec ts;
+
       /* release mutex to get a buffer, otherwise a deadlock may happen */
       pthread_mutex_unlock(&this->flush_audio_driver_lock);
       buf = fifo_remove (this->free_fifo);
@@ -1689,7 +1692,15 @@ static void ao_flush (xine_audio_port_t *this_gen) {
       buf->num_frames = 0;
       buf->stream = NULL;
       fifo_append (this->out_fifo, buf);
-      pthread_cond_wait(&this->flush_audio_driver_reached, &this->flush_audio_driver_lock);
+
+      /* cond_timedwait was not supposed be needed here, but somehow it may still
+       * get stuck when using normal cond_wait. probably the signal is missed when
+       * we release the mutex above.
+       */
+      gettimeofday(&tv, NULL);
+      ts.tv_sec  = tv.tv_sec + 1;
+      ts.tv_nsec = tv.tv_usec * 1000;
+      pthread_cond_timedwait(&this->flush_audio_driver_reached, &this->flush_audio_driver_lock, &ts);
     }
     this->discard_buffers--;