[Cast] Remove AudioDecoder's dependency on WebRTC, and refactor/clean-up AudioReceiver.

media/cast/audio_receiver/audio_decoder.cc

Index: media/cast/audio_receiver/audio_decoder.cc
diff --git a/media/cast/audio_receiver/audio_decoder.cc b/media/cast/audio_receiver/audio_decoder.cc
index b1a8256f2e9d7dd3dbdfcfc957d012987440d788..4e75473a6bcb14967aa2f5498a12c9b909cc89fe 100644
--- a/media/cast/audio_receiver/audio_decoder.cc
+++ b/media/cast/audio_receiver/audio_decoder.cc
@@ -2,165 +2,257 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
-#include "base/logging.h"
 #include "media/cast/audio_receiver/audio_decoder.h"
 
-#include "third_party/webrtc/modules/audio_coding/main/interface/audio_coding_module.h"
-#include "third_party/webrtc/modules/interface/module_common_types.h"
+#include "base/bind.h"
+#include "base/bind_helpers.h"
+#include "base/location.h"
+#include "base/logging.h"
+#include "base/memory/ref_counted.h"
+#include "base/stl_util.h"
+#include "base/sys_byteorder.h"
+#include "media/cast/cast_defines.h"
+#include "third_party/opus/src/include/opus.h"
 
 namespace media {
 namespace cast {
 
-AudioDecoder::AudioDecoder(scoped_refptr<CastEnvironment> cast_environment,
-                           const AudioReceiverConfig& audio_config,
-                           RtpPayloadFeedback* incoming_payload_feedback)
-    : cast_environment_(cast_environment),
-      audio_decoder_(webrtc::AudioCodingModule::Create(0)),
-      cast_message_builder_(cast_environment->Clock(),
-                            incoming_payload_feedback,
-                            &frame_id_map_,
-                            audio_config.incoming_ssrc,
-                            true,
-                            0),
-      have_received_packets_(false),
-      last_played_out_timestamp_(0) {
-  audio_decoder_->InitializeReceiver();
-
-  webrtc::CodecInst receive_codec;
-  switch (audio_config.codec) {
-    case transport::kPcm16:
-      receive_codec.pltype = audio_config.rtp_payload_type;
-      strncpy(receive_codec.plname, "L16", 4);
-      receive_codec.plfreq = audio_config.frequency;
-      receive_codec.pacsize = -1;
-      receive_codec.channels = audio_config.channels;
-      receive_codec.rate = -1;
-      break;
-    case transport::kOpus:
-      receive_codec.pltype = audio_config.rtp_payload_type;
-      strncpy(receive_codec.plname, "opus", 5);
-      receive_codec.plfreq = audio_config.frequency;
-      receive_codec.pacsize = -1;
-      receive_codec.channels = audio_config.channels;
-      receive_codec.rate = -1;
-      break;
-    case transport::kExternalAudio:
-      NOTREACHED() << "Codec must be specified for audio decoder";
-      break;
-  }
-  if (audio_decoder_->RegisterReceiveCodec(receive_codec) != 0) {
-    NOTREACHED() << "Failed to register receive codec";
+// Base class that handles the common problem of detecting dropped frames, and
+// then invoking the Decode() method implemented by the subclasses to convert
+// the encoded payload data into usable audio data.
+class AudioDecoder::ImplBase
+    : public base::RefCountedThreadSafe<AudioDecoder::ImplBase> {
+ public:
+  ImplBase(const scoped_refptr<CastEnvironment>& cast_environment,
+           transport::AudioCodec codec,
+           int num_channels,
+           int sampling_rate)
+      : cast_environment_(cast_environment),
+        codec_(codec),
+        num_channels_(num_channels),
+        cast_initialization_status_(STATUS_AUDIO_UNINITIALIZED),
+        seen_first_frame_(false) {
+    if (num_channels_ <= 0 || sampling_rate <= 0 || sampling_rate % 100 != 0)
+      cast_initialization_status_ = STATUS_INVALID_AUDIO_CONFIGURATION;
   }
 
-  audio_decoder_->SetMaximumPlayoutDelay(audio_config.rtp_max_delay_ms);
-  audio_decoder_->SetPlayoutMode(webrtc::streaming);
-}
+  CastInitializationStatus InitializationResult() const {
+    return cast_initialization_status_;
+  }
 
-AudioDecoder::~AudioDecoder() {}
+  void DecodeFrame(scoped_ptr<transport::EncodedAudioFrame> encoded_frame,
+                   const DecodeFrameCallback& callback) {
+    DCHECK_EQ(cast_initialization_status_, STATUS_AUDIO_INITIALIZED);
 
-bool AudioDecoder::GetRawAudioFrame(int number_of_10ms_blocks,
-                                    int desired_frequency,
-                                    PcmAudioFrame* audio_frame,
-                                    uint32* rtp_timestamp) {
-  DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::AUDIO));
-  // We don't care about the race case where a packet arrives at the same time
-  // as this function in called. The data will be there the next time this
-  // function is called.
-  lock_.Acquire();
-  // Get a local copy under lock.
-  bool have_received_packets = have_received_packets_;
-  lock_.Release();
-
-  if (!have_received_packets)
-    return false;
-
-  audio_frame->samples.clear();
-
-  for (int i = 0; i < number_of_10ms_blocks; ++i) {
-    webrtc::AudioFrame webrtc_audio_frame;
-    if (0 != audio_decoder_->PlayoutData10Ms(desired_frequency,
-                                             &webrtc_audio_frame)) {
-      return false;
+    scoped_ptr<AudioBus> decoded_audio;
+    if (encoded_frame->codec != codec_) {
+      NOTREACHED();
+      cast_environment_->PostTask(CastEnvironment::MAIN,
+                                  FROM_HERE,
+                                  base::Bind(callback,
+                                             base::Passed(&decoded_audio),
+                                             false));
     }
-    if (webrtc_audio_frame.speech_type_ == webrtc::AudioFrame::kPLCCNG ||
-        webrtc_audio_frame.speech_type_ == webrtc::AudioFrame::kUndefined) {
-      // We are only interested in real decoded audio.
-      return false;
-    }
-    audio_frame->frequency = webrtc_audio_frame.sample_rate_hz_;
-    audio_frame->channels = webrtc_audio_frame.num_channels_;
 
-    if (i == 0) {
-      // Use the timestamp from the first 10ms block.
-      if (0 != audio_decoder_->PlayoutTimestamp(rtp_timestamp)) {
-        return false;
+    COMPILE_ASSERT(sizeof(encoded_frame->frame_id) == sizeof(last_frame_id_),
+                   size_of_frame_id_types_do_not_match);
+    bool is_continuous = true;
+    if (seen_first_frame_) {
+      const uint32 frames_ahead = encoded_frame->frame_id - last_frame_id_;
+      if (frames_ahead > 1) {
+        RecoverBecauseFramesWereDropped();
+        is_continuous = false;
       }
-      lock_.Acquire();
-      last_played_out_timestamp_ = *rtp_timestamp;
-      lock_.Release();
+    } else {
+      seen_first_frame_ = true;
     }
-    int samples_per_10ms = webrtc_audio_frame.samples_per_channel_;
+    last_frame_id_ = encoded_frame->frame_id;
 
-    audio_frame->samples.insert(
-        audio_frame->samples.end(),
-        &webrtc_audio_frame.data_[0],
-        &webrtc_audio_frame.data_[samples_per_10ms * audio_frame->channels]);
+    decoded_audio = Decode(
+        reinterpret_cast<uint8*>(string_as_array(&encoded_frame->data)),
+        static_cast<int>(encoded_frame->data.size()));
+    cast_environment_->PostTask(CastEnvironment::MAIN,
+                                FROM_HERE,
+                                base::Bind(callback,
+                                           base::Passed(&decoded_audio),
+                                           is_continuous));
   }
-  return true;
-}
 
-void AudioDecoder::IncomingParsedRtpPacket(const uint8* payload_data,
-                                           size_t payload_size,
-                                           const RtpCastHeader& rtp_header) {
-  DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::MAIN));
-  DCHECK_LE(payload_size, kMaxIpPacketSize);
-  audio_decoder_->IncomingPacket(
-      payload_data, static_cast<int32>(payload_size), rtp_header.webrtc);
-  lock_.Acquire();
-  have_received_packets_ = true;
-  uint32 last_played_out_timestamp = last_played_out_timestamp_;
-  lock_.Release();
-
-  PacketType packet_type = frame_id_map_.InsertPacket(rtp_header);
-  if (packet_type != kNewPacketCompletingFrame)
-    return;
+ protected:
+  friend class base::RefCountedThreadSafe<ImplBase>;
+  virtual ~ImplBase() {}
 
-  cast_message_builder_.CompleteFrameReceived(rtp_header.frame_id,
-                                              rtp_header.is_key_frame);
+  virtual void RecoverBecauseFramesWereDropped() {}
 
-  frame_id_rtp_timestamp_map_[rtp_header.frame_id] =
-      rtp_header.webrtc.header.timestamp;
+  // Note: Implementation of Decode() is allowed to mutate |data|.
+  virtual scoped_ptr<AudioBus> Decode(uint8* data, int len) = 0;
 
-  if (last_played_out_timestamp == 0)
-    return;  // Nothing is played out yet.
+  const scoped_refptr<CastEnvironment> cast_environment_;
+  const transport::AudioCodec codec_;
+  const int num_channels_;
 
-  uint32 latest_frame_id_to_remove = 0;
-  bool frame_to_remove = false;
+  // Subclass' ctor is expected to set this to STATUS_AUDIO_INITIALIZED.
+  CastInitializationStatus cast_initialization_status_;
 
-  FrameIdRtpTimestampMap::iterator it = frame_id_rtp_timestamp_map_.begin();
-  while (it != frame_id_rtp_timestamp_map_.end()) {
-    if (IsNewerRtpTimestamp(it->second, last_played_out_timestamp)) {
-      break;
+ private:
+  bool seen_first_frame_;
+  uint32 last_frame_id_;
+
+  DISALLOW_COPY_AND_ASSIGN(ImplBase);
+};
+
+class AudioDecoder::OpusImpl : public AudioDecoder::ImplBase {
+ public:
+  OpusImpl(const scoped_refptr<CastEnvironment>& cast_environment,
+           int num_channels,
+           int sampling_rate)
+      : ImplBase(cast_environment,
+                 transport::kOpus,
+                 num_channels,
+                 sampling_rate),
+        decoder_memory_(new uint8[opus_decoder_get_size(num_channels)]),
+        opus_decoder_(reinterpret_cast<OpusDecoder*>(decoder_memory_.get())),
+        max_samples_per_frame_(
+            kOpusMaxFrameDurationMillis * sampling_rate / 1000),
+        buffer_(new float[max_samples_per_frame_ * num_channels]) {
+    if (ImplBase::cast_initialization_status_ != STATUS_AUDIO_UNINITIALIZED)
+      return;
+    if (opus_decoder_init(opus_decoder_, sampling_rate, num_channels) !=
+            OPUS_OK) {
+      ImplBase::cast_initialization_status_ =
+          STATUS_INVALID_AUDIO_CONFIGURATION;
+      return;
     }
-    frame_to_remove = true;
-    latest_frame_id_to_remove = it->first;
-    frame_id_rtp_timestamp_map_.erase(it);
-    it = frame_id_rtp_timestamp_map_.begin();
+    ImplBase::cast_initialization_status_ = STATUS_AUDIO_INITIALIZED;
   }
-  if (!frame_to_remove)
-    return;
 
-  frame_id_map_.RemoveOldFrames(latest_frame_id_to_remove);
+ private:
+  virtual ~OpusImpl() {}
+
+  virtual void RecoverBecauseFramesWereDropped() OVERRIDE {
+    // Passing NULL for the input data notifies the decoder of frame loss.
+    const opus_int32 result =
+        opus_decode_float(
+            opus_decoder_, NULL, 0, buffer_.get(), max_samples_per_frame_, 0);
+    DCHECK_GE(result, 0);
+  }
+
+  virtual scoped_ptr<AudioBus> Decode(uint8* data, int len) OVERRIDE {
+    scoped_ptr<AudioBus> audio_bus;
+    const opus_int32 num_samples_decoded = opus_decode_float(
+        opus_decoder_, data, len, buffer_.get(), max_samples_per_frame_, 0);
+    if (num_samples_decoded <= 0)
+      return audio_bus.Pass();  // Decode error.
+
+    // Copy interleaved samples from |buffer_| into a new AudioBus (where
+    // samples are stored in planar format, for each channel).
+    audio_bus = AudioBus::Create(num_channels_, num_samples_decoded).Pass();
+    // TODO(miu): This should be moved into AudioBus::FromInterleaved().
+    for (int ch = 0; ch < num_channels_; ++ch) {
+      const float* src = buffer_.get() + ch;
+      const float* const src_end = src + num_samples_decoded * num_channels_;
+      float* dest = audio_bus->channel(ch);
+      for (; src < src_end; src += num_channels_, ++dest)
+        *dest = *src;
+    }
+    return audio_bus.Pass();
+  }
+
+  const scoped_ptr<uint8[]> decoder_memory_;
+  OpusDecoder* const opus_decoder_;
+  const int max_samples_per_frame_;
+  const scoped_ptr<float[]> buffer_;
+
+  // According to documentation in third_party/opus/src/include/opus.h, we must
+  // provide enough space in |buffer_| to contain 120ms of samples.  At 48 kHz,
+  // then, that means 5760 samples times the number of channels.
+  static const int kOpusMaxFrameDurationMillis = 120;
+
+  DISALLOW_COPY_AND_ASSIGN(OpusImpl);
+};
+
+class AudioDecoder::Pcm16Impl : public AudioDecoder::ImplBase {
+ public:
+  Pcm16Impl(const scoped_refptr<CastEnvironment>& cast_environment,
+            int num_channels,
+            int sampling_rate)
+      : ImplBase(cast_environment,
+                 transport::kPcm16,
+                 num_channels,
+                 sampling_rate) {
+    if (ImplBase::cast_initialization_status_ != STATUS_AUDIO_UNINITIALIZED)
+      return;
+    ImplBase::cast_initialization_status_ = STATUS_AUDIO_INITIALIZED;
+  }
+
+ private:
+  virtual ~Pcm16Impl() {}
+
+  virtual scoped_ptr<AudioBus> Decode(uint8* data, int len) OVERRIDE {
+    scoped_ptr<AudioBus> audio_bus;
+    const int num_samples = len / sizeof(int16) / num_channels_;
+    if (num_samples <= 0)
+      return audio_bus.Pass();
+
+    int16* const pcm_data = reinterpret_cast<int16*>(data);
+#if defined(ARCH_CPU_LITTLE_ENDIAN)
+    // Convert endianness.
+    const int num_elements = num_samples * num_channels_;
+    for (int i = 0; i < num_elements; ++i)
+      pcm_data[i] = static_cast<int16>(base::NetToHost16(pcm_data[i]));
+#endif
+    audio_bus = AudioBus::Create(num_channels_, num_samples).Pass();
+    audio_bus->FromInterleaved(pcm_data, num_samples, sizeof(int16));
+    return audio_bus.Pass();
+  }
+
+  DISALLOW_COPY_AND_ASSIGN(Pcm16Impl);
+};
+
+AudioDecoder::AudioDecoder(
+    const scoped_refptr<CastEnvironment>& cast_environment,
+    const AudioReceiverConfig& audio_config)
+    : cast_environment_(cast_environment) {
+  switch (audio_config.codec) {
+    case transport::kOpus:
+      impl_ = new OpusImpl(cast_environment,
+                           audio_config.channels,
+                           audio_config.frequency);
+      break;
+    case transport::kPcm16:
+      impl_ = new Pcm16Impl(cast_environment,
+                            audio_config.channels,
+                            audio_config.frequency);
+      break;
+    default:
+      NOTREACHED() << "Unknown or unspecified codec.";
+      break;
+  }
 }
 
-bool AudioDecoder::TimeToSendNextCastMessage(base::TimeTicks* time_to_send) {
-  DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::MAIN));
-  return cast_message_builder_.TimeToSendNextCastMessage(time_to_send);
+AudioDecoder::~AudioDecoder() {}
+
+CastInitializationStatus AudioDecoder::InitializationResult() const {
+  if (impl_)
+    return impl_->InitializationResult();
+  return STATUS_UNSUPPORTED_AUDIO_CODEC;
 }
 
-void AudioDecoder::SendCastMessage() {
-  DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::MAIN));
-  cast_message_builder_.UpdateCastMessage();
+void AudioDecoder::DecodeFrame(
+    scoped_ptr<transport::EncodedAudioFrame> encoded_frame,
+    const DecodeFrameCallback& callback) {
+  DCHECK(encoded_frame.get());
+  DCHECK(!callback.is_null());
+  if (!impl_ || impl_->InitializationResult() != STATUS_AUDIO_INITIALIZED) {
+    callback.Run(make_scoped_ptr<AudioBus>(NULL), false);
+    return;
+  }
+  cast_environment_->PostTask(CastEnvironment::AUDIO,
+                              FROM_HERE,
+                              base::Bind(&AudioDecoder::ImplBase::DecodeFrame,
+                                         impl_,
+                                         base::Passed(&encoded_frame),
+                                         callback));
 }
 
 }  // namespace cast