Fix handling of buffer underruns in ALSA. Probably closes #461

We passed the delay verbatim to ...frames_to_bytes function which doesn't like negative numbers.

modules/audio_output/alsa.c

@@ -818,6 +818,8 @@ static void ALSAFill( aout_instance_t * p_aout )
             /* Here the device should be either in the RUNNING state.
              * p_status is valid. */
 
+#if 0
+    /* This apparently does not work correctly in Alsa 1.0.11 */
             snd_pcm_status_get_tstamp( p_status, &ts_next );
             next_date = (mtime_t)ts_next.tv_sec * 1000000 + ts_next.tv_usec;
             if( next_date )
@@ -826,19 +828,16 @@ static void ALSAFill( aout_instance_t * p_aout )
                         * 1000000 / p_aout->output.output.i_rate;
             }
             else
+#endif
             {
                 /* With screwed ALSA drivers the timestamp is always zero;
                  * use another method then */
-                snd_pcm_sframes_t delay;
+                snd_pcm_sframes_t delay = 0;
                 ssize_t i_bytes = 0;
 
-                if( !snd_pcm_delay( p_sys->p_snd_pcm, &delay ) )
-                {
-                    i_bytes = snd_pcm_frames_to_bytes(p_sys->p_snd_pcm, delay);
-                }
-                next_date = mdate() + (mtime_t)i_bytes * 1000000
-                        / p_aout->output.output.i_bytes_per_frame
-                        / p_aout->output.output.i_rate
+                snd_pcm_delay( p_sys->p_snd_pcm, &delay );
+                next_date = mdate() + (mtime_t)(delay) * 1000000 /
+                          p_aout->output.output.i_rate
                         * p_aout->output.output.i_frame_length;
             }
         }