Allow parsing empty RTCP TargetBitrate messages, but stop sending them.

webrtc/common_types.h

Index: webrtc/common_types.h
diff --git a/webrtc/common_types.h b/webrtc/common_types.h
index 73a1c8340d62e93b65b3d0f4894583826c336705..f41c646000705eb95a2f13b36772980a19115531 100644
--- a/webrtc/common_types.h
+++ b/webrtc/common_types.h
@@ -631,12 +631,15 @@ class BitrateAllocation {
   inline bool operator!=(const BitrateAllocation& other) const {
     return !(*this == other);
   }
+  std::string ToString() const;
 
  private:
   uint32_t sum_;
   uint32_t bitrates_[kMaxSpatialLayers][kMaxTemporalStreams];
 };
 
+std::ostream& operator<<(std::ostream& os, const BitrateAllocation& bitrate);

[stefan-webrtc, 2017/06/22 11:10:38]

Where is this used? To me it seems like it should be in a test helper if we need
it only for testing? 

[sprang_webrtc, 2017/06/22 11:28:29]

It is not used, at the moment.
It is however something I use frequently when debugging, and I have grown tired
of constantly having to re-implement or patch this locally when that happens.
The operator<< is useful if you want to log it in a test, or just for getting an
immediate text representation inside a graphical debugger.
Where I exactly I've used it has varied. Sometime in the pacing module,
sometimes in the video coding module, sometimes in the rtp/rtcp module... so
having it in common types makes sense to me.

[stefan-webrtc, 2017/06/22 11:38:12]

Ok, I'm mostly worried about adding usage of stringstream, which I remember that
tommi has been trying to get rid of from webrtc since it's very heavy. That's
why I'd prefer to only build this into tests...

[sprang_webrtc, 2017/06/22 12:06:57]

Right.. I'll add a comment about about using this only in tests for now, perhaps
I can switch to something simpler after discussing with him?

+
 // Bandwidth over-use detector options.  These are used to drive
 // experimentation with bandwidth estimation parameters.
 // See modules/remote_bitrate_estimator/overuse_detector.h