output: signal the output thread when CANCEL is finished

After CANCEL, the output thread waits for another signal before it
continues playback, to synchronize with the caller.  There were some
situations where this signal wasn't sent properly.  This patch adds an
explicit g_cond_signal() at two code positions.

src/output_all.c

@@ -202,6 +202,29 @@ static void audio_output_wait_all(void)
 		notify_wait(&audio_output_client_notify);
 }
 
+/**
+ * Signals the audio output if it is open.  This function locks the
+ * mutex.
+ */
+static void
+audio_output_lock_signal(struct audio_output *ao)
+{
+	g_mutex_lock(ao->mutex);
+	if (audio_output_is_open(ao))
+		g_cond_signal(ao->cond);
+	g_mutex_unlock(ao->mutex);
+}
+
+/**
+ * Signals all audio outputs which are open.
+ */
+static void
+audio_output_signal_all(void)
+{
+	for (unsigned i = 0; i < num_audio_outputs; ++i)
+		audio_output_lock_signal(&audio_outputs[i]);
+}
+
 static void
 audio_output_reset_reopen(struct audio_output *ao)
 {
@@ -488,6 +511,13 @@ audio_output_all_cancel(void)
 	if (g_mp != NULL)
 		music_pipe_clear(g_mp, g_music_buffer);
 
+	/* the audio outputs are now waiting for a signal, to
+	   synchronize the cleared music pipe */
+
+	audio_output_signal_all();
+
+	/* invalidate elapsed_time */
+
 	audio_output_all_elapsed_time = -1.0;
 }
 

src/output_control.c

@@ -128,6 +128,10 @@ audio_output_open(struct audio_output *ao,
 			/* we're not using audio_output_cancel() here,
 			   because that function is asynchronous */
 			ao_command(ao, AO_COMMAND_CANCEL);
+
+			/* the audio output is now waiting for a
+			   signal; wake it up immediately */
+			g_cond_signal(ao->cond);
 		}
 
 		return true;