Use ProcessReverseStream in VoiceEngines OutputMixer

webrtc/voice_engine/output_mixer.cc

 /*
  *  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 498 matching lines @@
         {
             // Pure stereo mode (we are receiving a stereo signal).
         }
 
         assert(_audioFrame.num_channels_ == 2);
         AudioFrameOperations::Scale(_panLeft, _panRight, _audioFrame);
     }
 
     // --- Far-end Voice Quality Enhancement (AudioProcessing Module)
     if (feed_data_to_apm) {
-      // Convert from mixing to AudioProcessing sample rate, similarly to how it
-      // is done on the send side. Downmix to mono.
-      AudioFrame frame;
-      frame.num_channels_ = 1;
-      frame.sample_rate_hz_ = _audioProcessingModulePtr->input_sample_rate_hz();
-      RemixAndResample(_audioFrame, &audioproc_resampler_, &frame);
-
-      if (_audioProcessingModulePtr->AnalyzeReverseStream(&frame) != 0) {
+      if (_audioProcessingModulePtr->ProcessReverseStream(&_audioFrame) != 0) {

[peah-webrtc, 2016/03/09 22:07:54]

Do we know the implications of this? This changes the APM input from being mono
to being stereo. How is this handled inside APM? Do we downmix to mono before
doing AEC, AGC and AECM processing? If not, afaics this change will actually
double the complexity for these subcomponents as the APM applies separate
instances of these for each channel. On top of that, this may trigger a bug
related to how the processing is applied on the capture channels.

Lets discuss the implications of this change.

[peah-webrtc, 2016/03/10 10:49:30]

After offline discussions: I think this change is fine as there is a
functionality in APM that also internally downmixes the signal to 1 channels. 
This changes must, however, be properly tested as the internal downmixing has
not been active (due to the code being removed in this CL) before this CL.

         WEBRTC_TRACE(kTraceWarning, kTraceVoice, VoEId(_instanceId, -1),
                      "AudioProcessingModule::AnalyzeReverseStream() => error");
         RTC_DCHECK(false);
       }
     }
 
     // --- External media processing
     {
         rtc::CritScope cs(&_callbackCritSect);
         if (_externalMedia)
@@ skipping 61 matching lines @@
             _audioFrame.data_[2 * i + 1] = 0;
         }
     }
     assert(_audioFrame.samples_per_channel_ == toneSamples);
 
     return 0;
 }
 
 }  // namespace voe
 }  // namespace webrtc