Remove frame time scheduing in IncomingVideoStream

talk/media/webrtc/webrtcvideoengine2.cc

Index: talk/media/webrtc/webrtcvideoengine2.cc
diff --git a/talk/media/webrtc/webrtcvideoengine2.cc b/talk/media/webrtc/webrtcvideoengine2.cc
index 101ed15bddd996d4e51d462d537f4d50ee613dd4..b11d4f6479702ca222a9251e4b361ae09f67de3b 100644
--- a/talk/media/webrtc/webrtcvideoengine2.cc
+++ b/talk/media/webrtc/webrtcvideoengine2.cc
@@ -1226,6 +1226,8 @@ bool WebRtcVideoChannel2::AddRecvStream(const StreamParams& sp,
 
   config.rtp.remb = send_codec_ ? HasRemb(send_codec_->codec) : false;
 
+  config.renderer_has_time_control = options_.renderer_has_time_control;

[pbos, 2015/11/16 18:38:46]

I want this to be an option on the renderer for VideoRenderer:

pbos@deimos:~/webrtc/src (master)$ git diff
diff --git a/webrtc/video_renderer.h b/webrtc/video_renderer.h
index fedd28b..058a37a 100644
--- a/webrtc/video_renderer.h
+++ b/webrtc/video_renderer.h
@@ -25,6 +25,8 @@ class VideoRenderer {
 
   virtual bool IsTextureSupported() const = 0;
 
+  virtual bool HasTimeControl() const { return false; }
+
  protected:
   virtual ~VideoRenderer() {}
 };


Then you override that one in the renderer implementation (which IIRC is
WebRtcVideoReceiveStream).

[pbos, 2015/11/16 18:40:55]

Maybe ::SchedulesFrames(), I dunno, wdyt?

[pthatcher1, 2015/11/16 20:25:27]

That gets kind of complicated, because there can be multiple renderers.  So what
do you do if one does its own time control but one doesn't?  We ran into a
similar issue with frame rotation.   It gets pretty messy and quickly leads into
a discussion of refactoring all of our renderers everywhere.

[qiangchen, 2015/11/16 20:44:14]

For the current data structure, one VideoRenderer (essentially
WebRtcVideoReceiveStream) corresponds to one IncomingVideoStream. In another
word one IncomingVideoStream will serve one VideoRenderer only, so there is no
such problem as "some has time control, some not".

[qiangchen, 2015/11/16 20:44:14]

The reason I chose Config to carry this piece of info is that it is passed to
downstream class on the current code path, so this way we do not need to add
parameter to any functions.

Just noticed that Config has a back pointer to the VideoRenderer, so it is
feasible to move this property into VideoRenderer.

I'll do it.

+
   receive_streams_[ssrc] = new WebRtcVideoReceiveStream(
       call_, sp, config, external_decoder_factory_, default_stream,
       recv_codecs_);