Cleanup of the AudioDeviceBuffer class

webrtc/modules/audio_device/audio_device_buffer.cc

Index: webrtc/modules/audio_device/audio_device_buffer.cc
diff --git a/webrtc/modules/audio_device/audio_device_buffer.cc b/webrtc/modules/audio_device/audio_device_buffer.cc
index ba8b6a5b3e1f773be7acc0a55a98ec7a77b7fe1a..d157c1cb1ae661483675e18411c0bd57d6a6d449 100644
--- a/webrtc/modules/audio_device/audio_device_buffer.cc
+++ b/webrtc/modules/audio_device/audio_device_buffer.cc
@@ -22,9 +22,6 @@
 
 namespace webrtc {
 
-static const int kHighDelayThresholdMs = 300;
-static const int kLogHighDelayIntervalFrames = 500;  // 5 seconds.
-
 static const char kTimerQueueName[] = "AudioDeviceBufferTimer";
 
 // Time between two sucessive calls to LogStats().
@@ -33,30 +30,26 @@ static const size_t kTimerIntervalInMilliseconds =
     kTimerIntervalInSeconds * rtc::kNumMillisecsPerSec;
 
 AudioDeviceBuffer::AudioDeviceBuffer()
-    : _ptrCbAudioTransport(nullptr),
+    : audio_transport_cb_(nullptr),
       task_queue_(kTimerQueueName),
       timer_has_started_(false),
-      _recSampleRate(0),
-      _playSampleRate(0),
-      _recChannels(0),
-      _playChannels(0),
-      _recChannel(AudioDeviceModule::kChannelBoth),
-      _recBytesPerSample(0),
-      _playBytesPerSample(0),
-      _recSamples(0),
-      _recSize(0),
-      _playSamples(0),
-      _playSize(0),
-      _recFile(*FileWrapper::Create()),
-      _playFile(*FileWrapper::Create()),
-      _currentMicLevel(0),
-      _newMicLevel(0),
-      _typingStatus(false),
-      _playDelayMS(0),
-      _recDelayMS(0),
-      _clockDrift(0),
-      // Set to the interval in order to log on the first occurrence.
-      high_delay_counter_(kLogHighDelayIntervalFrames),
+      rec_sample_rate_(0),
+      play_sample_rate_(0),
+      rec_channels_(0),
+      play_channels_(0),
+      rec_channel_(AudioDeviceModule::kChannelBoth),
+      rec_bytes_per_sample_(0),
+      play_bytes_per_sample_(0),
+      rec_samples_per_10ms_(0),
+      rec_bytes_per_10ms_(0),
+      play_samples_per_10ms_(0),
+      play_bytes_per_10ms_(0),
+      current_mic_level_(0),
+      new_mic_level_(0),
+      typing_status_(false),
+      play_delay_ms_(0),
+      rec_delay_ms_(0),
+      clock_drift_(0),
       num_stat_reports_(0),
       rec_callbacks_(0),
       last_rec_callbacks_(0),
@@ -68,8 +61,6 @@ AudioDeviceBuffer::AudioDeviceBuffer()
       last_play_samples_(0),
       last_log_stat_time_(0) {
   LOG(INFO) << "AudioDeviceBuffer::ctor";
-  memset(_recBuffer, 0, kMaxBufferSizeBytes);
-  memset(_playBuffer, 0, kMaxBufferSizeBytes);
 }
 
 AudioDeviceBuffer::~AudioDeviceBuffer() {
@@ -93,27 +84,19 @@ AudioDeviceBuffer::~AudioDeviceBuffer() {
     LOG(INFO) << "average: "
              << static_cast<float>(total_diff_time) / num_measurements;
   }
-
-  _recFile.Flush();
-  _recFile.CloseFile();
-  delete &_recFile;
-
-  _playFile.Flush();
-  _playFile.CloseFile();
-  delete &_playFile;
 }
 
 int32_t AudioDeviceBuffer::RegisterAudioCallback(
-    AudioTransport* audioCallback) {
+    AudioTransport* audio_callback) {
   LOG(INFO) << __FUNCTION__;
   rtc::CritScope lock(&_critSectCb);
-  _ptrCbAudioTransport = audioCallback;
+  audio_transport_cb_ = audio_callback;
   return 0;
 }
 
 int32_t AudioDeviceBuffer::InitPlayout() {
-  RTC_DCHECK(thread_checker_.CalledOnValidThread());
   LOG(INFO) << __FUNCTION__;
+  RTC_DCHECK(thread_checker_.CalledOnValidThread());
   last_playout_time_ = rtc::TimeMillis();
   if (!timer_has_started_) {
     StartTimer();
@@ -123,8 +106,8 @@ int32_t AudioDeviceBuffer::InitPlayout() {
 }
 
 int32_t AudioDeviceBuffer::InitRecording() {
-  RTC_DCHECK(thread_checker_.CalledOnValidThread());
   LOG(INFO) << __FUNCTION__;
+  RTC_DCHECK(thread_checker_.CalledOnValidThread());
   if (!timer_has_started_) {
     StartTimer();
     timer_has_started_ = true;
@@ -135,38 +118,40 @@ int32_t AudioDeviceBuffer::InitRecording() {
 int32_t AudioDeviceBuffer::SetRecordingSampleRate(uint32_t fsHz) {
   LOG(INFO) << "SetRecordingSampleRate(" << fsHz << ")";
   rtc::CritScope lock(&_critSect);
-  _recSampleRate = fsHz;
+  rec_sample_rate_ = fsHz;
   return 0;
 }
 
 int32_t AudioDeviceBuffer::SetPlayoutSampleRate(uint32_t fsHz) {
   LOG(INFO) << "SetPlayoutSampleRate(" << fsHz << ")";
   rtc::CritScope lock(&_critSect);
-  _playSampleRate = fsHz;
+  play_sample_rate_ = fsHz;
   return 0;
 }
 
 int32_t AudioDeviceBuffer::RecordingSampleRate() const {
-  return _recSampleRate;
+  return rec_sample_rate_;
 }
 
 int32_t AudioDeviceBuffer::PlayoutSampleRate() const {
-  return _playSampleRate;
+  return play_sample_rate_;
 }
 
 int32_t AudioDeviceBuffer::SetRecordingChannels(size_t channels) {
+  LOG(INFO) << "SetRecordingChannels(" << channels << ")";
   rtc::CritScope lock(&_critSect);
-  _recChannels = channels;
-  _recBytesPerSample =
+  rec_channels_ = channels;
+  rec_bytes_per_sample_ =
       2 * channels;  // 16 bits per sample in mono, 32 bits in stereo
   return 0;
 }
 
 int32_t AudioDeviceBuffer::SetPlayoutChannels(size_t channels) {
+  LOG(INFO) << "SetPlayoutChannels(" << channels << ")";
   rtc::CritScope lock(&_critSect);
-  _playChannels = channels;
+  play_channels_ = channels;
   // 16 bits per sample in mono, 32 bits in stereo
-  _playBytesPerSample = 2 * channels;
+  play_bytes_per_sample_ = 2 * channels;
   return 0;
 }
 
@@ -174,135 +159,101 @@ int32_t AudioDeviceBuffer::SetRecordingChannel(
     const AudioDeviceModule::ChannelType channel) {
   rtc::CritScope lock(&_critSect);
 
-  if (_recChannels == 1) {
+  if (rec_channels_ == 1) {
     return -1;
   }
 
   if (channel == AudioDeviceModule::kChannelBoth) {
     // two bytes per channel
-    _recBytesPerSample = 4;
+    rec_bytes_per_sample_ = 4;
   } else {
     // only utilize one out of two possible channels (left or right)
-    _recBytesPerSample = 2;
+    rec_bytes_per_sample_ = 2;
   }
-  _recChannel = channel;
+  rec_channel_ = channel;
 
   return 0;
 }
 
 int32_t AudioDeviceBuffer::RecordingChannel(
     AudioDeviceModule::ChannelType& channel) const {
-  channel = _recChannel;
+  channel = rec_channel_;
   return 0;
 }
 
 size_t AudioDeviceBuffer::RecordingChannels() const {
-  return _recChannels;
+  return rec_channels_;
 }
 
 size_t AudioDeviceBuffer::PlayoutChannels() const {
-  return _playChannels;
+  return play_channels_;
 }
 
 int32_t AudioDeviceBuffer::SetCurrentMicLevel(uint32_t level) {
-  _currentMicLevel = level;
+  current_mic_level_ = level;
   return 0;
 }
 
-int32_t AudioDeviceBuffer::SetTypingStatus(bool typingStatus) {
-  _typingStatus = typingStatus;
+int32_t AudioDeviceBuffer::SetTypingStatus(bool typing_status) {
+  typing_status_ = typing_status;
   return 0;
 }
 
 uint32_t AudioDeviceBuffer::NewMicLevel() const {
-  return _newMicLevel;
+  return new_mic_level_;
 }
 
-void AudioDeviceBuffer::SetVQEData(int playDelayMs,
-                                   int recDelayMs,
-                                   int clockDrift) {
-  if (high_delay_counter_ < kLogHighDelayIntervalFrames) {
-    ++high_delay_counter_;
-  } else {
-    if (playDelayMs + recDelayMs > kHighDelayThresholdMs) {
-      high_delay_counter_ = 0;
-      LOG(LS_WARNING) << "High audio device delay reported (render="
-                      << playDelayMs << " ms, capture=" << recDelayMs << " ms)";
-    }
-  }
-
-  _playDelayMS = playDelayMs;
-  _recDelayMS = recDelayMs;
-  _clockDrift = clockDrift;
+void AudioDeviceBuffer::SetVQEData(int play_delay_ms,
+                                   int rec_delay_ms,
+                                   int clock_drift) {
+  play_delay_ms_ = play_delay_ms;
+  rec_delay_ms_ = rec_delay_ms;
+  clock_drift_ = clock_drift;
 }
 
 int32_t AudioDeviceBuffer::StartInputFileRecording(
     const char fileName[kAdmMaxFileNameSize]) {
-  rtc::CritScope lock(&_critSect);
-
-  _recFile.Flush();
-  _recFile.CloseFile();
-
-  return _recFile.OpenFile(fileName, false) ? 0 : -1;
+  LOG(LS_WARNING) << "Not implemented";
+  return 0;
 }
 
 int32_t AudioDeviceBuffer::StopInputFileRecording() {
-  rtc::CritScope lock(&_critSect);
-
-  _recFile.Flush();
-  _recFile.CloseFile();
-
+  LOG(LS_WARNING) << "Not implemented";
   return 0;
 }
 
 int32_t AudioDeviceBuffer::StartOutputFileRecording(
     const char fileName[kAdmMaxFileNameSize]) {
-  rtc::CritScope lock(&_critSect);
-
-  _playFile.Flush();
-  _playFile.CloseFile();
-
-  return _playFile.OpenFile(fileName, false) ? 0 : -1;
+  LOG(LS_WARNING) << "Not implemented";
+  return 0;
 }
 
 int32_t AudioDeviceBuffer::StopOutputFileRecording() {
-  rtc::CritScope lock(&_critSect);
-
-  _playFile.Flush();
-  _playFile.CloseFile();
-
+  LOG(LS_WARNING) << "Not implemented";
   return 0;
 }
 
-int32_t AudioDeviceBuffer::SetRecordedBuffer(const void* audioBuffer,
-                                             size_t nSamples) {
-  rtc::CritScope lock(&_critSect);
-
-  if (_recBytesPerSample == 0) {
-    assert(false);
-    return -1;
-  }
+int32_t AudioDeviceBuffer::SetRecordedBuffer(const void* audio_buffer,
+                                             size_t num_samples) {
+  AllocateRecordingBufferIfNeeded();
+  RTC_CHECK(rec_buffer_);
+  // WebRTC can only receive audio in 10ms chunks, hence we fail if the native
+  // audio layer tries to deliver something else.
+  RTC_CHECK_EQ(num_samples, rec_samples_per_10ms_);
 
-  _recSamples = nSamples;
-  _recSize = _recBytesPerSample * nSamples;  // {2,4}*nSamples
-  if (_recSize > kMaxBufferSizeBytes) {
-    assert(false);
-    return -1;
-  }
+  rtc::CritScope lock(&_critSect);
 
-  if (_recChannel == AudioDeviceModule::kChannelBoth) {
-    // (default) copy the complete input buffer to the local buffer
-    memcpy(&_recBuffer[0], audioBuffer, _recSize);
+  if (rec_channel_ == AudioDeviceModule::kChannelBoth) {
+    // Copy the complete input buffer to the local buffer.
+    memcpy(&rec_buffer_[0], audio_buffer, rec_bytes_per_10ms_);
   } else {
-    int16_t* ptr16In = (int16_t*)audioBuffer;
-    int16_t* ptr16Out = (int16_t*)&_recBuffer[0];
-
-    if (AudioDeviceModule::kChannelRight == _recChannel) {
+    int16_t* ptr16In = (int16_t*)audio_buffer;
+    int16_t* ptr16Out = (int16_t*)&rec_buffer_[0];
+    if (AudioDeviceModule::kChannelRight == rec_channel_) {
       ptr16In++;
     }
-
-    // exctract left or right channel from input buffer to the local buffer
-    for (size_t i = 0; i < _recSamples; i++) {
+    // Exctract left or right channel from input buffer to the local buffer.
+    for (size_t i = 0; i < rec_samples_per_10ms_; i++) {
       *ptr16Out = *ptr16In;
       ptr16Out++;
       ptr16In++;
@@ -310,52 +261,40 @@ int32_t AudioDeviceBuffer::SetRecordedBuffer(const void* audioBuffer,
     }
   }
 
-  if (_recFile.is_open()) {
-    // write to binary file in mono or stereo (interleaved)
-    _recFile.Write(&_recBuffer[0], _recSize);
-  }
-
   // Update some stats but do it on the task queue to ensure that the members
   // are modified and read on the same thread.
   task_queue_.PostTask(
-      rtc::Bind(&AudioDeviceBuffer::UpdateRecStats, this, nSamples));
-
+      rtc::Bind(&AudioDeviceBuffer::UpdateRecStats, this, num_samples));
   return 0;
 }
 
 int32_t AudioDeviceBuffer::DeliverRecordedData() {
+  RTC_CHECK(rec_buffer_);
+  RTC_DCHECK(audio_transport_cb_);
   rtc::CritScope lock(&_critSectCb);
-  // Ensure that user has initialized all essential members
-  if ((_recSampleRate == 0) || (_recSamples == 0) ||
-      (_recBytesPerSample == 0) || (_recChannels == 0)) {
-    RTC_NOTREACHED();
-    return -1;
-  }
 
-  if (!_ptrCbAudioTransport) {
+  if (!audio_transport_cb_) {
     LOG(LS_WARNING) << "Invalid audio transport";
     return 0;
   }
 
   int32_t res(0);
   uint32_t newMicLevel(0);
-  uint32_t totalDelayMS = _playDelayMS + _recDelayMS;
-  res = _ptrCbAudioTransport->RecordedDataIsAvailable(
-      &_recBuffer[0], _recSamples, _recBytesPerSample, _recChannels,
-      _recSampleRate, totalDelayMS, _clockDrift, _currentMicLevel,
-      _typingStatus, newMicLevel);
+  uint32_t totalDelayMS = play_delay_ms_ + rec_delay_ms_;
+  res = audio_transport_cb_->RecordedDataIsAvailable(
+      &rec_buffer_[0], rec_samples_per_10ms_, rec_bytes_per_sample_,
+      rec_channels_, rec_sample_rate_, totalDelayMS, clock_drift_,
+      current_mic_level_, typing_status_, newMicLevel);
   if (res != -1) {
-    _newMicLevel = newMicLevel;
+    new_mic_level_ = newMicLevel;
+  } else {
+    LOG(LS_ERROR) << "RecordedDataIsAvailable() failed";
   }
 
   return 0;
 }
 
-int32_t AudioDeviceBuffer::RequestPlayoutData(size_t nSamples) {
-  uint32_t playSampleRate = 0;
-  size_t playBytesPerSample = 0;
-  size_t playChannels = 0;
-
+int32_t AudioDeviceBuffer::RequestPlayoutData(size_t num_samples) {
   // Measure time since last function call and update an array where the
   // position/index corresponds to time differences (in milliseconds) between
   // two successive playout callbacks, and the stored value is the number of
@@ -367,37 +306,17 @@ int32_t AudioDeviceBuffer::RequestPlayoutData(size_t nSamples) {
   last_playout_time_ = now_time;
   playout_diff_times_[diff_time]++;
 
-  // TOOD(henrika): improve bad locking model and make it more clear that only
-  // 10ms buffer sizes is supported in WebRTC.
-  {
-    rtc::CritScope lock(&_critSect);
-
-    // Store copies under lock and use copies hereafter to avoid race with
-    // setter methods.
-    playSampleRate = _playSampleRate;
-    playBytesPerSample = _playBytesPerSample;
-    playChannels = _playChannels;
-
-    // Ensure that user has initialized all essential members
-    if ((playBytesPerSample == 0) || (playChannels == 0) ||
-        (playSampleRate == 0)) {
-      RTC_NOTREACHED();
-      return -1;
-    }
-
-    _playSamples = nSamples;
-    _playSize = playBytesPerSample * nSamples;  // {2,4}*nSamples
-    RTC_CHECK_LE(_playSize, kMaxBufferSizeBytes);
-    RTC_CHECK_EQ(nSamples, _playSamples);
-  }
-
-  size_t nSamplesOut(0);
+  AllocatePlayoutBufferIfNeeded();
+  RTC_CHECK(play_buffer_);
+  // WebRTC can only provide audio in 10ms chunks, hence we fail if the native
+  // audio layer asks for something else.
+  RTC_CHECK_EQ(num_samples, play_samples_per_10ms_);
 
   rtc::CritScope lock(&_critSectCb);
 
   // It is currently supported to start playout without a valid audio
   // transport object. Leads to warning and silence.
-  if (!_ptrCbAudioTransport) {
+  if (!audio_transport_cb_) {
     LOG(LS_WARNING) << "Invalid audio transport";
     return 0;
   }
@@ -405,9 +324,11 @@ int32_t AudioDeviceBuffer::RequestPlayoutData(size_t nSamples) {
   uint32_t res(0);
   int64_t elapsed_time_ms = -1;
   int64_t ntp_time_ms = -1;
-  res = _ptrCbAudioTransport->NeedMorePlayData(
-      _playSamples, playBytesPerSample, playChannels, playSampleRate,
-      &_playBuffer[0], nSamplesOut, &elapsed_time_ms, &ntp_time_ms);
+  size_t num_samples_out(0);
+  res = audio_transport_cb_->NeedMorePlayData(
+      play_samples_per_10ms_, play_bytes_per_sample_, play_channels_,
+      play_sample_rate_, &play_buffer_[0], num_samples_out, &elapsed_time_ms,
+      &ntp_time_ms);
   if (res != 0) {
     LOG(LS_ERROR) << "NeedMorePlayData() failed";
   }
@@ -415,23 +336,46 @@ int32_t AudioDeviceBuffer::RequestPlayoutData(size_t nSamples) {
   // Update some stats but do it on the task queue to ensure that access of
   // members is serialized hence avoiding usage of locks.
   task_queue_.PostTask(
-      rtc::Bind(&AudioDeviceBuffer::UpdatePlayStats, this, nSamplesOut));
-
-  return static_cast<int32_t>(nSamplesOut);
+      rtc::Bind(&AudioDeviceBuffer::UpdatePlayStats, this, num_samples_out));
+  return static_cast<int32_t>(num_samples_out);
 }
 
-int32_t AudioDeviceBuffer::GetPlayoutData(void* audioBuffer) {
+int32_t AudioDeviceBuffer::GetPlayoutData(void* audio_buffer) {
   rtc::CritScope lock(&_critSect);
-  RTC_CHECK_LE(_playSize, kMaxBufferSizeBytes);
-
-  memcpy(audioBuffer, &_playBuffer[0], _playSize);
+  memcpy(audio_buffer, &play_buffer_[0], play_bytes_per_10ms_);
+  return static_cast<int32_t>(play_samples_per_10ms_);
+}
 
-  if (_playFile.is_open()) {
-    // write to binary file in mono or stereo (interleaved)
-    _playFile.Write(&_playBuffer[0], _playSize);
-  }
+void AudioDeviceBuffer::AllocatePlayoutBufferIfNeeded() {
+  RTC_CHECK(play_bytes_per_sample_);
+  if (play_buffer_)
+    return;
+  LOG(INFO) << __FUNCTION__;
+  rtc::CritScope lock(&_critSect);
+  // Derive the required buffer size given sample rate and number of channels.
+  play_samples_per_10ms_ = static_cast<size_t>(play_sample_rate_ * 10 / 1000);
+  play_bytes_per_10ms_ = play_bytes_per_sample_ * play_samples_per_10ms_;
+  LOG(INFO) << "playout samples per 10ms: " << play_samples_per_10ms_;
+  LOG(INFO) << "playout bytes per 10ms: " << play_bytes_per_10ms_;
+  // Allocate memory for the playout audio buffer. It will always contain audio
+  // samples corresponding to 10ms of audio to be played out.
+  play_buffer_.reset(new int8_t[play_bytes_per_10ms_]);
+}
 
-  return static_cast<int32_t>(_playSamples);
+void AudioDeviceBuffer::AllocateRecordingBufferIfNeeded() {
+  RTC_CHECK(rec_bytes_per_sample_);
+  if (rec_buffer_)
+    return;
+  LOG(INFO) << __FUNCTION__;
+  rtc::CritScope lock(&_critSect);
+  // Derive the required buffer size given sample rate and number of channels.
+  rec_samples_per_10ms_ = static_cast<size_t>(rec_sample_rate_ * 10 / 1000);
+  rec_bytes_per_10ms_ = rec_bytes_per_sample_ * rec_samples_per_10ms_;
+  LOG(INFO) << "recorded samples per 10ms: " << rec_samples_per_10ms_;
+  LOG(INFO) << "recorded bytes per 10ms: " << rec_bytes_per_10ms_;
+  // Allocate memory for the recording audio buffer. It will always contain
+  // audio samples corresponding to 10ms of audio.
+  rec_buffer_.reset(new int8_t[rec_bytes_per_10ms_]);
 }
 
 void AudioDeviceBuffer::StartTimer() {
@@ -455,7 +399,7 @@ void AudioDeviceBuffer::LogStats() {
     uint32_t diff_samples = rec_samples_ - last_rec_samples_;
     uint32_t rate = diff_samples / kTimerIntervalInSeconds;
     LOG(INFO) << "[REC : " << time_since_last << "msec, "
-              << _recSampleRate / 1000
+              << rec_sample_rate_ / 1000
               << "kHz] callbacks: " << rec_callbacks_ - last_rec_callbacks_
               << ", "
               << "samples: " << diff_samples << ", "
@@ -464,7 +408,7 @@ void AudioDeviceBuffer::LogStats() {
     diff_samples = play_samples_ - last_play_samples_;
     rate = diff_samples / kTimerIntervalInSeconds;
     LOG(INFO) << "[PLAY: " << time_since_last << "msec, "
-              << _playSampleRate / 1000
+              << play_sample_rate_ / 1000
               << "kHz] callbacks: " << play_callbacks_ - last_play_callbacks_
               << ", "
               << "samples: " << diff_samples << ", "