[rtp_rtcp] tmmbr helper simplified by moving set_to_send member out

webrtc/modules/rtp_rtcp/source/rtcp_sender.h

Index: webrtc/modules/rtp_rtcp/source/rtcp_sender.h
diff --git a/webrtc/modules/rtp_rtcp/source/rtcp_sender.h b/webrtc/modules/rtp_rtcp/source/rtcp_sender.h
index 977c1ff43a04579abd227c6c914513999f663dd8..409db3fab28c8db9415b1e2ff4116b5413dae5bd 100644
--- a/webrtc/modules/rtp_rtcp/source/rtcp_sender.h
+++ b/webrtc/modules/rtp_rtcp/source/rtcp_sender.h
@@ -132,7 +132,7 @@ class RTCPSender {
 
   void SetTMMBRStatus(bool enable);
 
-  int32_t SetTMMBN(const TMMBRSet* boundingSet);
+  void SetTMMBN(const TMMBRSet* boundingSet);

[philipel, 2016/03/04 10:49:27]

While you are at it, change to const TMMBRSet& instead.

[danilchap, 2016/03/04 14:06:13]

Unfortunately right now caller can't guarantee non-null parameter.
So leaving this change until later CL where this guarantee would be obvious.

 
   int32_t SetApplicationSpecificData(uint8_t subType,
                                      uint32_t name,
@@ -238,6 +238,8 @@ class RTCPSender {
   std::vector<uint32_t> remb_ssrcs_ GUARDED_BY(critical_section_rtcp_sender_);
 
   TMMBRHelp tmmbr_help_ GUARDED_BY(critical_section_rtcp_sender_);
+  std::vector<rtcp::TmmbItem> tmmbn_to_send_

[philipel, 2016/03/04 10:49:27]

Don't you think it would be clearer to use a TMMBRSet instead of a vector here?
Even if you don't need any of the convenience functions I still think it is
better.

[danilchap, 2016/03/04 14:06:13]

Yes, it is inconsistent with SetTMMBN function, but
TMMBRSet type doesn't follow type name convention and doesn't add any real
features to vector. Goal is to remove this type completely.
So, to be clearer, I've updated SetTMMBN function instead.

+      GUARDED_BY(critical_section_rtcp_sender_);
   uint32_t tmmbr_send_ GUARDED_BY(critical_section_rtcp_sender_);
   uint32_t packet_oh_send_ GUARDED_BY(critical_section_rtcp_sender_);