cefsrc: Apply running time on audio buffers

Also add an audio meta to buffers, for good measure, and set discont flag when
needed. This should help with A/V sync issues and audio cracks reported in centricular#59.

gstcefdemux.cc

@@ -6,6 +6,7 @@
 #define CEF_VIDEO_CAPS "video/x-raw, format=BGRA, width=[1, 2147483647], height=[1, 2147483647], framerate=[1/1, 60/1], pixel-aspect-ratio=1/1"
 #define CEF_AUDIO_CAPS "audio/x-raw, format=F32LE, rate=[1, 2147483647], channels=[1, 2147483647], layout=interleaved"
 
+#define gst_cef_demux_parent_class parent_class
 G_DEFINE_TYPE (GstCefDemux, gst_cef_demux, GST_TYPE_ELEMENT);
 
 static GstStaticPadTemplate gst_cef_demux_sink_template =
@@ -58,6 +59,7 @@ gst_cef_demux_push_events (GstCefDemux *demux)
         "channels", G_TYPE_INT, 2,
         "layout", G_TYPE_STRING, "interleaved",
         NULL);
+    gst_audio_info_from_caps (&demux->audio_info, audio_caps);
     gst_pad_push_event (demux->asrcpad, gst_event_new_caps (audio_caps));
     gst_caps_unref (audio_caps);
 
@@ -86,10 +88,21 @@ typedef struct
 static gboolean
 gst_cef_demux_push_audio_buffer (GstBuffer **buffer, guint idx, AudioPushData *push_data)
 {
-  GST_BUFFER_PTS (*buffer) += push_data->demux->ts_offset;
+  push_data->demux->last_audio_time = gst_element_get_current_running_time (GST_ELEMENT_CAST (push_data->demux));
+  GST_BUFFER_DTS (*buffer) = push_data->demux->last_audio_time;
+  GST_BUFFER_PTS (*buffer) = push_data->demux->last_audio_time;
+
+  gst_buffer_add_audio_meta (*buffer, &push_data->demux->audio_info, gst_buffer_get_size (*buffer), NULL);
+
+  GST_BUFFER_FLAG_UNSET (*buffer, GST_BUFFER_FLAG_DISCONT);
+  if (push_data->demux->need_discont) {
+    GST_BUFFER_FLAG_SET (*buffer, GST_BUFFER_FLAG_DISCONT);
+    push_data->demux->need_discont = FALSE;
+  }
+
   push_data->combined = gst_flow_combiner_update_pad_flow (push_data->flow_combiner, push_data->demux->asrcpad,
       gst_pad_push (push_data->demux->asrcpad, *buffer));
-  push_data->demux->last_audio_time = GST_BUFFER_PTS (*buffer) + GST_BUFFER_DURATION (*buffer);
+
   *buffer = NULL;
   return TRUE;
 }
@@ -127,11 +140,6 @@ gst_cef_demux_chain (GstPad * pad, GstObject * parent, GstBuffer * buffer)
 
   gst_cef_demux_push_events (demux);
 
-
-  if (!GST_CLOCK_TIME_IS_VALID (demux->ts_offset)) {
-    demux->ts_offset = GST_BUFFER_PTS (buffer);
-  }
-
   for (tmp = demux->cef_audio_stream_start_events; tmp; tmp = tmp->next) {
     const GstStructure *s = gst_event_get_structure ((GstEvent *) tmp->data);
 
@@ -228,6 +236,31 @@ gst_cef_demux_sink_query (GstPad *pad, GstObject *parent, GstQuery *query)
   return ret;
 }
 
+
+static GstStateChangeReturn
+gst_cef_demux_change_state (GstElement * element, GstStateChange transition)
+{
+  GstStateChangeReturn result;
+  GstCefDemux *demux = (GstCefDemux *) element;
+
+  GST_DEBUG_OBJECT (demux, "%s", gst_state_change_get_name (transition));
+  result = GST_CALL_PARENT_WITH_DEFAULT (GST_ELEMENT_CLASS , change_state, (element, transition), GST_STATE_CHANGE_FAILURE);
+
+  switch (transition) {
+  case GST_STATE_CHANGE_PAUSED_TO_READY:
+    gst_flow_combiner_reset (demux->flow_combiner);
+    break;
+  case GST_STATE_CHANGE_PLAYING_TO_PAUSED:
+    demux->need_discont = TRUE;
+    break;
+  default:
+    break;
+  }
+
+  return result;
+}
+
+
 static void
 gst_cef_demux_init (GstCefDemux * demux)
 {
@@ -248,11 +281,13 @@ gst_cef_demux_init (GstCefDemux * demux)
   gst_element_add_pad (GST_ELEMENT (demux), demux->asrcpad);
   gst_flow_combiner_add_pad (demux->flow_combiner, demux->asrcpad);
 
+  gst_audio_info_init (&demux->audio_info);
+
   demux->need_stream_start = TRUE;
   demux->need_caps = TRUE;
   demux->need_segment = TRUE;
-  demux->last_audio_time = 0;
-  demux->ts_offset = GST_CLOCK_TIME_NONE;
+  demux->need_discont = TRUE;
+  demux->last_audio_time = GST_CLOCK_TIME_NONE;
 }
 
 static void
@@ -273,6 +308,8 @@ gst_cef_demux_class_init (GstCefDemuxClass * klass)
 
   gobject_class->finalize = gst_cef_demux_finalize;
 
+  gstelement_class->change_state = GST_DEBUG_FUNCPTR(gst_cef_demux_change_state);
+
   gst_element_class_set_static_metadata (gstelement_class,
       "Chromium Embedded Framework demuxer", "Demuxer/Audio/Video",
       "Demuxes audio and video from cefsrc", "Mathieu Duponchelle <[email protected]>");

gstcefdemux.h

@@ -3,6 +3,7 @@
 
 #include <gst/gst.h>
 #include <gst/base/gstflowcombiner.h>
+#include <gst/audio/audio.h>
 
 G_BEGIN_DECLS
 
@@ -26,13 +27,14 @@ struct _GstCefDemux {
   gboolean need_stream_start;
   gboolean need_caps;
   gboolean need_segment;
+  gboolean need_discont;
   GstPad *vsrcpad;
   GstPad *asrcpad;
   GList *cef_audio_stream_start_events;
   GstEvent *vcaps_event;
   GstFlowCombiner *flow_combiner;
   GstClockTime last_audio_time;
-  GstClockTime ts_offset;
+  GstAudioInfo audio_info;
 };
 
 struct _GstCefDemuxClass {

gstcefsrc.cc

@@ -182,7 +182,6 @@ class AudioHandler : public CefAudioHandler
 
     mRate = params.sample_rate;
     mChannels = channels;
-    mCurrentTime = GST_CLOCK_TIME_NONE;
 
     GST_OBJECT_LOCK (mElement);
     mElement->audio_events = g_list_append (mElement->audio_events, event);
@@ -214,14 +213,7 @@ class AudioHandler : public CefAudioHandler
 
     GST_OBJECT_LOCK (mElement);
 
-    if (!GST_CLOCK_TIME_IS_VALID (mCurrentTime)) {
-      mCurrentTime = gst_util_uint64_scale (mElement->n_frames,
-          mElement->vinfo.fps_d * GST_SECOND, mElement->vinfo.fps_n);
-    }
-
-    GST_BUFFER_PTS (buf) = mCurrentTime;
     GST_BUFFER_DURATION (buf) = gst_util_uint64_scale (frames, GST_SECOND, mRate);
-    mCurrentTime += GST_BUFFER_DURATION (buf);
 
     if (!mElement->audio_buffers) {
       mElement->audio_buffers = gst_buffer_list_new();
@@ -245,7 +237,6 @@ class AudioHandler : public CefAudioHandler
   private:
 
     GstCefSrc *mElement;
-    GstClockTime mCurrentTime;
     gint mRate;
     gint mChannels;
     IMPLEMENT_REFCOUNTING(AudioHandler);