Bugfix that fixes the integration issue that cause WebRTC AEC mobile to fail

webrtc/voice_engine/output_mixer.cc

Index: webrtc/voice_engine/output_mixer.cc
diff --git a/webrtc/voice_engine/output_mixer.cc b/webrtc/voice_engine/output_mixer.cc
index d46ee6296d6a946c179641ad41e53f77fa1b8c08..4d280185c3ce26e7dccfbe525c7e4665505a5edc 100644
--- a/webrtc/voice_engine/output_mixer.cc
+++ b/webrtc/voice_engine/output_mixer.cc
@@ -550,9 +550,17 @@ OutputMixer::DoOperationsOnCombinedSignal(bool feed_data_to_apm)
 // ----------------------------------------------------------------------------
 
 void OutputMixer::APMProcessReverseStream() {

[aluebs-webrtc, 2015/12/15 19:16:08]

APMAnalyzeReverseStream?

[the sun, 2015/12/15 19:37:14]

If the name is a problem I'd suggest making this code inline in the one place
where it is being called.

[peah-webrtc, 2015/12/15 21:29:45]

I decided to keep it since I expected that this CL would be followed up with
another fix that allowed the ProcessReverseStream call to be reintroduced. I
will change according to the inline suggestion instead though. Please let me
know if you are ok with that!

[peah-webrtc, 2015/12/15 21:29:45]

Done.

[aluebs-webrtc, 2015/12/15 21:40:34]

Sounds good to me.

-  if (_audioProcessingModulePtr->ProcessReverseStream(&_audioFrame) != 0) {
-    WEBRTC_TRACE(kTraceError, kTraceVoice, VoEId(_instanceId, -1),
-                 "AudioProcessingModule::ProcessReverseStream() => error");
+  // 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);

[the sun, 2015/12/15 19:37:14]

What are the expected performance effects of this?

[peah-webrtc, 2015/12/15 21:29:45]

In what sense do you mean? There is a computational cost for the resampling and
how big that is I cannot tell. But note that before Aug 14, this resampling was
always done so it should not be anything that the platforms where it is used
cannot handle.

The resampling is needed to at all be able to run AECM but on a longer term
basis we should perhaps consider to do it in the band-splitting mechanism
instead.

+
+  if (_audioProcessingModulePtr->AnalyzeReverseStream(&frame) != 0) {
+    WEBRTC_TRACE(kTraceWarning, kTraceVoice, VoEId(_instanceId, -1),
+                 "AudioProcessingModule::AnalyzeReverseStream() => error");
+    assert(false);

[hlundin-webrtc, 2015/12/15 16:07:05]

DCHECK

[peah-webrtc, 2015/12/15 21:29:45]

Done.

   }
 }