Bug 1543622 - Make number of channels out param of GetAudioFrame; r=pehrsons

The number of channels is available in mAudioFrame in GetAudioFrame so
there is no reason to calculate it after the fact in MediaPipeline.

Differential Revision: https://phabricator.services.mozilla.com/D50934

--HG--
extra : moz-landing-system : lando

media/webrtc/signaling/gtest/mediaconduit_unittests.cpp

@@ -197,7 +197,8 @@ void AudioSendAndReceive::GenerateAndReadSamples() {
   auto audioInput = mozilla::MakeUnique<int16_t[]>(PLAYOUT_SAMPLE_LENGTH);
   auto audioOutput = mozilla::MakeUnique<int16_t[]>(PLAYOUT_SAMPLE_LENGTH);
   short* inbuf;
-  int sampleLengthDecoded = 0;
+  size_t numChannels = 0;
+  size_t sampleLengthDecoded = 0;
   unsigned int SAMPLES = (PLAYOUT_SAMPLE_FREQUENCY / 100);  // 10 milliseconds
   int CHANNELS = 1;                                         // mono audio
   int sampleLengthInBytes = sizeof(int16_t) * PLAYOUT_SAMPLE_LENGTH;
@@ -246,7 +247,7 @@ void AudioSendAndReceive::GenerateAndReadSamples() {
 
     PR_Sleep(PR_MillisecondsToInterval(10));
     mOtherSession->GetAudioFrame(audioOutput.get(), PLAYOUT_SAMPLE_FREQUENCY,
-                                 10, sampleLengthDecoded);
+                                 10, numChannels, sampleLengthDecoded);
     if (sampleLengthDecoded == 0) {
       cerr << " Zero length Sample " << endl;
     }

media/webrtc/signaling/src/media-conduit/AudioConduit.cpp

@@ -583,7 +583,8 @@ MediaConduitErrorCode WebrtcAudioConduit::SendAudioFrame(
 MediaConduitErrorCode WebrtcAudioConduit::GetAudioFrame(int16_t speechData[],
                                                         int32_t samplingFreqHz,
                                                         int32_t capture_delay,
-                                                        int& lengthSamples) {
+                                                        size_t& numChannels,
+                                                        size_t& lengthSamples) {
   CSFLogDebug(LOGTAG, "%s ", __FUNCTION__);
 
   // validate params
@@ -614,16 +615,23 @@ MediaConduitErrorCode WebrtcAudioConduit::GetAudioFrame(int16_t speechData[],
     return kMediaConduitSessionNotInited;
   }
 
-  int lengthSamplesAllowed = lengthSamples;
+  size_t lengthSamplesAllowed = lengthSamples;
   lengthSamples = 0;  // output paramter
 
   mRecvChannelProxy->GetAudioFrameWithInfo(samplingFreqHz, &mAudioFrame);
+  numChannels = mAudioFrame.num_channels_;
+
+  if (numChannels == 0) {
+    CSFLogError(LOGTAG, "%s Audio frame has zero channels", __FUNCTION__);
+    return kMediaConduitPlayoutError;
+  }
+
   // XXX Annoying, have to copy to our buffers -- refactor?
   lengthSamples = mAudioFrame.samples_per_channel_ * mAudioFrame.num_channels_;
   MOZ_RELEASE_ASSERT(lengthSamples <= lengthSamplesAllowed);
   PodCopy(speechData, mAudioFrame.data(), lengthSamples);
 
-  CSFLogDebug(LOGTAG, "%s GetAudioFrame:Got samples: length %d ", __FUNCTION__,
+  CSFLogDebug(LOGTAG, "%s GetAudioFrame:Got samples: length %zu ", __FUNCTION__,
               lengthSamples);
   return kMediaConduitNoError;
 }

media/webrtc/signaling/src/media-conduit/AudioConduit.h

@@ -137,6 +137,8 @@ class WebrtcAudioConduit : public AudioSessionConduit,
    * @param capture_delay [in]: Estimated Time between reading of the samples
    *                            to rendering/playback
    * @param lengthSamples [in]: Contain maximum length of speechData array.
+   * @param numChannels [out]: Number of channels in the audio frame,
+   *                           guaranteed to be non-zero.
    * @param lengthSamples [out]: Will contain length of the audio frame in
    *                             samples at return.
    *                             Ex: A value of 160 implies 160 samples each of
@@ -151,7 +153,8 @@ class WebrtcAudioConduit : public AudioSessionConduit,
   MediaConduitErrorCode GetAudioFrame(int16_t speechData[],
                                       int32_t samplingFreqHz,
                                       int32_t capture_delay,
-                                      int& lengthSamples) override;
+                                      size_t& numChannels,
+                                      size_t& lengthSamples) override;
 
   /**
    * Webrtc transport implementation to send and receive RTP packet.

media/webrtc/signaling/src/media-conduit/MediaConduitInterface.h

@@ -536,6 +536,8 @@ class AudioSessionConduit : public MediaSessionConduit {
    *                             Hertz (16000, 32000,..)
    * @param capture_delay [in]: Estimated Time between reading of the samples
    *                            to rendering/playback
+   * @param numChannels [out]: Number of channels in the audio frame,
+   *                           guaranteed to be non-zero.
    * @param lengthSamples [out]: Will contain length of the audio frame in
    *                             samples at return.
    *                             Ex: A value of 160 implies 160 samples each of
@@ -549,7 +551,8 @@ class AudioSessionConduit : public MediaSessionConduit {
   virtual MediaConduitErrorCode GetAudioFrame(int16_t speechData[],
                                               int32_t samplingFreqHz,
                                               int32_t capture_delay,
-                                              int& lengthSamples) = 0;
+                                              size_t& numChannels,
+                                              size_t& lengthSamples) = 0;
 
   /**
    * Checks if given sampling frequency is supported

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

@@ -1320,11 +1320,12 @@ class MediaPipelineReceiveAudio::PipelineListener
     // in the graph rate.
 
     while (mPlayedTicks < aDesiredTime) {
-      const int scratchBufferLength =
+      constexpr size_t scratchBufferLength =
           AUDIO_SAMPLE_BUFFER_MAX_BYTES / sizeof(int16_t);
       int16_t scratchBuffer[scratchBufferLength];
 
-      int samplesLength = scratchBufferLength;
+      size_t channelCount = 0;
+      size_t samplesLength = scratchBufferLength;
 
       // This fetches 10ms of data, either mono or stereo
       MediaConduitErrorCode err =
@@ -1332,7 +1333,7 @@ class MediaPipelineReceiveAudio::PipelineListener
               ->GetAudioFrame(scratchBuffer, mRate,
                               0,  // TODO(ekr@rtfm.com): better estimate of
                                   // "capture" (really playout) delay
-                              samplesLength);
+                              channelCount, samplesLength);
 
       if (err != kMediaConduitNoError) {
         // Insert silence on conduit/GIPS failure (extremely unlikely)
@@ -1341,6 +1342,7 @@ class MediaPipelineReceiveAudio::PipelineListener
                  " (desired %" PRId64 " -> %f)",
                  err, mPlayedTicks, aDesiredTime,
                  mSource->TrackTimeToSeconds(aDesiredTime)));
+        channelCount = 1;
         // if this is not enough we'll loop and provide more
         samplesLength = samplesPer10ms;
         PodArrayZero(scratchBuffer);
@@ -1349,16 +1351,12 @@ class MediaPipelineReceiveAudio::PipelineListener
       MOZ_RELEASE_ASSERT(samplesLength <= scratchBufferLength);
 
       MOZ_LOG(gMediaPipelineLog, LogLevel::Debug,
-              ("Audio conduit returned buffer of length %u", samplesLength));
+              ("Audio conduit returned buffer of length %zu", samplesLength));
 
       RefPtr<SharedBuffer> samples =
           SharedBuffer::Create(samplesLength * sizeof(uint16_t));
       int16_t* samplesData = static_cast<int16_t*>(samples->Data());
       AudioSegment segment;
-      // We derive the number of channels of the stream from the number of
-      // samples the AudioConduit gives us, considering it gives us packets of
-      // 10ms and we know the rate.
-      uint32_t channelCount = samplesLength / samplesPer10ms;
       AutoTArray<int16_t*, 2> channels;
       AutoTArray<const int16_t*, 2> outputChannels;
       size_t frames = samplesLength / channelCount;