Add read support of RtpStreamId/RepairedRtpStreamId header extensions.

webrtc/common_types.h

 /*
  *  Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 699 matching lines @@
   bool voiceActivity;
   uint8_t audioLevel;
 
   // For Coordination of Video Orientation. See
   // http://www.etsi.org/deliver/etsi_ts/126100_126199/126114/12.07.00_60/
   // ts_126114v120700p.pdf
   bool hasVideoRotation;
   VideoRotation videoRotation;
 
   PlayoutDelay playout_delay = {-1, -1};
+
+  // For identification of a stream when ssrc is not signaled. See
+  // https://tools.ietf.org/html/draft-ietf-avtext-rid-09
+  // TODO(danilchap): Update url from draft to release version.
+  // Stream id is limited to 16 bytes because it is the maximum length
+  // that can be encoded with one-byte header extensions.
+  size_t stream_id_size = 0;
+  char stream_id[16];

[nisse-webrtc, 2017/04/10 07:15:27]

This is intended as an arbitrary octet string, which could include nul bytes?
And the octet strings "foo", "foo\0", "foo\0\0" must be treated as distinct?

Then it seems uint8_t would be more appropriate than char, and that would also
signal that str* functions aren't appropriate. (And std::basic_string<uint8_t>
instead of std::string where a string class is wanted).

[danilchap, 2017/04/10 08:39:41]

according to spec valid characters are [0-9A-Za-z], so "foo\0" is an invalid
value.
This check is not enforced during parsing for simplicity:
It makes sense to validate the configured rsid, then invalid rsid would never
match valid one.

[nisse-webrtc, 2017/04/10 09:18:33]

Then one alternative would be to always zero-pad the values to 16 bytes, and
then drop the *_size members?

That would make things look simpler here, and comparisons could be done with
memcmp(stream_id, ..., sizeof(stream_id)). Might make things complex elsewhere,
though, depending on how this is used. This option is similar to the arcane unix
strncpy semantics.

Or make it 17 bytes and always nul-terminated (and if needed to avoid false
matches, reject received ids including nul characters).

I don't have a strong opinion here, but I think it would be nice if the
representation could be simplified.

[danilchap, 2017/04/10 11:44:16]

I've created a dedicated class StreamId and moved all complexity there. This way
it become easy to remove size property and make rest of the code simpler.

+  size_t repaired_stream_id_size = 0;
+  char repaired_stream_id[16];
 };
 
 struct RTPHeader {
   RTPHeader();
 
   bool markerBit;
   uint8_t payloadType;
   uint16_t sequenceNumber;
   uint32_t timestamp;
   uint32_t ssrc;
@@ skipping 105 matching lines @@
 enum class RtcpMode { kOff, kCompound, kReducedSize };
 
 enum NetworkState {
   kNetworkUp,
   kNetworkDown,
 };
 
 }  // namespace webrtc
 
 #endif  // WEBRTC_COMMON_TYPES_H_