Negotiate H264 profiles in SDP

webrtc/media/base/codec.h

Index: webrtc/media/base/codec.h
diff --git a/webrtc/media/base/codec.h b/webrtc/media/base/codec.h
index bfbff23c9416e336a303d90c2d11b60b943e3f05..adbdd660dc2b478e2c51dd74f72ee7822fd2029c 100644
--- a/webrtc/media/base/codec.h
+++ b/webrtc/media/base/codec.h
@@ -154,6 +154,10 @@ struct VideoCodec : public Codec {
   VideoCodec(VideoCodec&& c);
   virtual ~VideoCodec() = default;
 
+  // Indicates if this codec is compatible with the specified codec. For H264,
+  // the profiles must be the same. H264 levels however are not compared.

[hta-webrtc, 2016/11/08 13:59:53]

if Codec has a Matches function, shouldn't there be an "override" below?

(overriding on a reference argument may be a problem.)

[magjed_webrtc, 2016/11/08 16:18:26]

No, this is overloading, not overriding.

+  bool Matches(const VideoCodec& codec) const;

[hta-webrtc, 2016/11/08 13:59:53]

This is actually an issue with the concept of "codec" as encompassing both an
encoder and a decoder. If we have codecs A(encoder E, decoder D) and codec
a(encoder e, decoder d), it's perfectly reasonable to have encoder E produce a
stream that is decodable by d, but encoder e can produce a stream that is not
decodable by D.

Example: A=constrained baseline, a=high profile.

Matching needs to be on an encoder vs a decoder.

[magjed_webrtc, 2016/11/08 16:18:26]

The codec here does not really encompass both an encoder and a decoder. The
codec we offer/answer over SDP happens to represent both the encoder and the
decoder, but I would say that's an SDP artifact and not built into this class.

+
   std::string ToString() const;
 
   VideoCodec& operator=(const VideoCodec& c);