Bug 996853: handle AUDIO_FORMAT_SILENCE in MediaPipeline and AudioSegment::WriteTo r=roc

content/media/AudioSegment.cpp

@@ -175,7 +175,7 @@ AudioSegment::WriteTo(uint64_t aID, AudioStream* aOutput, AudioMixer* aMixer)
     // To avoid overbuffering in the AudioStream, we simply drop the silence,
     // here. The stream will underrun and output silence anyways.
     if (c.mBuffer || aOutput->GetWritten()) {
-      if (c.mBuffer) {
+      if (c.mBuffer && c.mBufferFormat != AUDIO_FORMAT_SILENCE) {
         channelData.SetLength(c.mChannelData.Length());
         for (uint32_t i = 0; i < channelData.Length(); ++i) {
           channelData[i] = c.mChannelData[i];

media/webrtc/signaling/src/mediapipeline/MediaPipeline.cpp

@@ -1010,6 +1010,9 @@ void MediaPipelineTransmit::PipelineListener::ProcessAudioChunk(
           ConvertAudioSamplesWithScale(buf, samples, chunk.mDuration, chunk.mVolume);
         }
         break;
+      case AUDIO_FORMAT_SILENCE:
+        memset(samples, 0, chunk.mDuration * sizeof(samples[0]));
+        break;
       default:
         MOZ_ASSERT(PR_FALSE);
         return;
@@ -1017,9 +1020,7 @@ void MediaPipelineTransmit::PipelineListener::ProcessAudioChunk(
     }
   } else {
     // This means silence.
-    for (uint32_t i = 0; i < chunk.mDuration; ++i) {
-      samples[i] = 0;
-    }
+    memset(samples, 0, chunk.mDuration * sizeof(samples[0]));
   }
 
   MOZ_ASSERT(!(rate%100)); // rate should be a multiple of 100