Enable passing receive stream ID to the decoder factory.

webrtc/media/engine/webrtcvideoengine2.cc

Index: webrtc/media/engine/webrtcvideoengine2.cc
diff --git a/webrtc/media/engine/webrtcvideoengine2.cc b/webrtc/media/engine/webrtcvideoengine2.cc
index 82d32b0e1865dff74586b16030a8731a6634a5ec..e1e50486d53ed65b75545a82455a68f9f99250b4 100644
--- a/webrtc/media/engine/webrtcvideoengine2.cc
+++ b/webrtc/media/engine/webrtcvideoengine2.cc
@@ -919,6 +919,7 @@ webrtc::RtpParameters WebRtcVideoChannel2::GetRtpReceiveParameters(
   for (const VideoCodec& codec : recv_params_.codecs) {
     rtp_params.codecs.push_back(codec.ToCodecParameters());
   }
+  rtp_params.ssrcs = it->second->GetSsrcs();

[pthatcher1, 2016/06/15 19:40:00]

This should be:

if (!it->second-GetSsrcs().empty()) {
  rtp_params.encodings[0] = it->second->GetSsrcs()[0];
}

[perkj_webrtc, 2016/06/15 20:25:52]

Why is this change needed at all ? 

[pthatcher1, 2016/06/15 21:08:17]

Now that you mention it, this part of the change isn't needed at all if sp.id is
used as the received_stream_id, because the app can just the ID passed into the
SessionDescription to correlate the decoder to the receive stream, even without
having the SSRC exposed in the RtpParameters.

So, as much as I like a part of the standard implemented, you are correct that
it's not needed for the purpose of this CL.


As for the purpose: my understanding is that certain applications want to have a
hardward-based VideoSinkInterface<EncodedFrame> for certain payload types but
not others.  We don't have support.  So they are hacking around it by making a
Decoder act like a VideoSinkInterface<EncodedFrame>.  But then they need to
correlate a Decoder to an RtpReceiver/ReceiveStream.  And do that, they
basically need the Decoder to know some ID of the RtpReceiver/ReceiveStream.  

   return rtp_params;
 }
 
@@ -2206,6 +2207,7 @@ WebRtcVideoChannel2::WebRtcVideoReceiveStream::WebRtcVideoReceiveStream(
     const std::vector<VideoCodecSettings>& recv_codecs,
     bool red_disabled_by_remote_side)
     : call_(call),
+      receive_stream_id_(sp.id),

[pthatcher1, 2016/06/15 19:40:00]

Interesting.   That's now what I had in mind originally, but that will probably
be easier to use and implement for now, since that comes right from the
SessionDescription passed into SetLocalDescription.

       ssrcs_(sp.ssrcs),
       ssrc_groups_(sp.ssrc_groups),
       stream_(NULL),
@@ -2267,7 +2269,8 @@ WebRtcVideoChannel2::WebRtcVideoReceiveStream::CreateOrReuseVideoDecoder(
 
   if (external_decoder_factory_ != NULL) {
     webrtc::VideoDecoder* decoder =
-        external_decoder_factory_->CreateVideoDecoder(type);
+        external_decoder_factory_->CreateVideoDecoder(type,
+            { receive_stream_id_ });
     if (decoder != NULL) {
       return AllocatedDecoder(decoder, type, true);
     }