Hooked up RtcEventLog. It lives in Voice Engine and pointers are propagated to ACM and Call.

webrtc/voice_engine/include/voe_codec.h

Index: webrtc/voice_engine/include/voe_codec.h
diff --git a/webrtc/voice_engine/include/voe_codec.h b/webrtc/voice_engine/include/voe_codec.h
index 5d94ac27365a91fa3df47901f28ad66422ab7b4f..a0703618e654a61c5dc378ccdf26af607ecf2a67 100644
--- a/webrtc/voice_engine/include/voe_codec.h
+++ b/webrtc/voice_engine/include/voe_codec.h
@@ -35,6 +35,7 @@
 
 namespace webrtc {
 
+class RtcEventLog;
 class VoiceEngine;
 
 class WEBRTC_DLLEXPORT VoECodec {
@@ -131,6 +132,10 @@ class WEBRTC_DLLEXPORT VoECodec {
   // success, and -1 if failed.
   virtual int SetOpusDtx(int channel, bool enable_dtx) = 0;
 
+  // Get a pointer to the event logging object associated with this Voice

[Henrik Grunell WebRTC, 2015/08/05 09:44:35]

Is the event logging object related to codec(s) only? Should this function
perhaps be in another "sub-API"?

[ivoc, 2015/08/06 10:57:51]

I was not entirely sure which of the sub-APIs is most appropriate for this
functionality. Initially I chose VoEBase, but Minyue@ said VoECodec would be
more appropriate because the actual logging is done in the ACM, and most of the
functions that relate to ACM are in VoECodec. If you have a better idea, feel
free to share.

[Henrik Grunell WebRTC, 2015/08/06 15:30:45]

Right, it's given to |audio_coding_| in the channel. This seems to be the right
place.

+  // Engine.
+  virtual RtcEventLog* GetEventLog() = 0;

[Henrik Grunell WebRTC, 2015/08/05 09:44:35]

Instead of getting a pointer and operate directly on that object, add functions
here that do what we need. E.g. Enable, Disable. The internal object should be
hidden.

[ivoc, 2015/08/06 10:57:52]

We need more functionality than just enable/disable. The pointer that is
obtained from this function is used in Call.cc to log RTP packets, RTCP packets
and video send/receive configs. We would have to basically replicate the whole
interface from the rtc_event_log.h header here, is that really desirable? 

[Henrik Grunell WebRTC, 2015/08/06 15:30:45]

OK. In that case I'd prefer the VoE user to have ownership of that object and
pass it to VoE. Let'd discuss offline tomorrow.

[Henrik Grunell WebRTC, 2015/08/07 11:30:03]

We talked offline and since (1) the path in libjingle between VoE and Call is
quite long with many layers, (2) the event log will eventually be moved
completely to Call, and (3) VoE outlives Call, I think it's OK to do it this
way.

+
  protected:
   VoECodec() {}
   virtual ~VoECodec() {}