Added unit test from MediaCodecPlayerAndroid

media/filters/android/media_codec_audio_decoder.cc

Index: media/filters/android/media_codec_audio_decoder.cc
diff --git a/media/filters/android/media_codec_audio_decoder.cc b/media/filters/android/media_codec_audio_decoder.cc
index f23452a4ca38edff8eb860c3ffe473e62348a8b4..3650ccb57c17f7590bc3910fd68bed87176ed9b2 100644
--- a/media/filters/android/media_codec_audio_decoder.cc
+++ b/media/filters/android/media_codec_audio_decoder.cc
@@ -11,7 +11,9 @@
 #include "base/thread_task_runner_handle.h"
 #include "media/base/android/sdk_media_codec_bridge.h"
 #include "media/base/audio_buffer.h"
+#include "media/base/audio_timestamp_helper.h"
 #include "media/base/bind_to_current_loop.h"
+#include "media/base/timestamp_constants.h"
 
 namespace media {
 
@@ -116,6 +118,8 @@ void MediaCodecAudioDecoder::Initialize(const AudioDecoderConfig& config,
   }
 
   config_ = config;
+  timestamp_helper_.reset(
+      new AudioTimestampHelper(config_.samples_per_second()));
   output_cb_ = BindToCurrentLoop(output_cb);
 
   SetState(STATE_READY);
@@ -126,8 +130,15 @@ void MediaCodecAudioDecoder::Decode(const scoped_refptr<DecoderBuffer>& buffer,
                                     const DecodeCB& decode_cb) {
   DecodeCB bound_decode_cb = BindToCurrentLoop(decode_cb);
 
+  if (!buffer->end_of_stream() && buffer->timestamp() == kNoTimestamp()) {
+    DVLOG(2) << __FUNCTION__ << " " << buffer->AsHumanReadableString()
+             << ": no timestamp, skipping this buffer";
+    bound_decode_cb.Run(kDecodeError);
+    return;
+  }
+
   if (state_ == STATE_ERROR) {
-    // We get here if an error happens in DecodeOutput() or Reset().
+    // We get here if an error happens in DequeueOutput() or Reset().
     DVLOG(2) << __FUNCTION__ << " " << buffer->AsHumanReadableString()
              << ": Error state, returning decode error for all buffers";
     ClearInputQueue(kDecodeError);
@@ -171,6 +182,9 @@ void MediaCodecAudioDecoder::Reset(const base::Closure& closure) {
     success = !!media_codec_;
   }
 
+  // Reset AudioTimestampHelper.
+  timestamp_helper_->SetBaseTimestamp(kNoTimestamp());
+
   SetState(success ? STATE_READY : STATE_ERROR);
 
   task_runner_->PostTask(FROM_HERE, closure);
@@ -204,6 +218,16 @@ void MediaCodecAudioDecoder::DoIOTask() {
 bool MediaCodecAudioDecoder::QueueInput() {
   DVLOG(2) << __FUNCTION__;
 
+  bool did_work = false;
+  while (QueueOneInputBuffer())
+    did_work = true;
+
+  return did_work;
+}
+
+bool MediaCodecAudioDecoder::QueueOneInputBuffer() {
+  DVLOG(2) << __FUNCTION__;
+
   if (input_queue_.empty())
     return false;
 
@@ -277,6 +301,7 @@ MediaCodecAudioDecoder::DequeueInputBuffer() {
       media_codec_->DequeueInputBuffer(NoWaitTimeout(), &input_buf_index);
   switch (status) {
     case media::MEDIA_CODEC_DEQUEUE_INPUT_AGAIN_LATER:
+      DVLOG(2) << __FUNCTION__ << ": MEDIA_CODEC_DEQUEUE_INPUT_AGAIN_LATER";
       break;
 
     case media::MEDIA_CODEC_ERROR:
@@ -501,8 +526,17 @@ void MediaCodecAudioDecoder::OnDecodedFrame(const OutputBufferInfo& out) {
       kSampleFormatS16, config_.channel_layout(), channel_count,
       config_.samples_per_second(), frame_count);
 
-  // Set timestamp.
-  audio_buffer->set_timestamp(out.pts);
+  // Calculate and set buffer timestamp.
+
+  const bool first_buffer =
+      timestamp_helper_->base_timestamp() == kNoTimestamp();
+  if (first_buffer) {
+    // Clamp the base timestamp to zero.
+    timestamp_helper_->SetBaseTimestamp(std::max(base::TimeDelta(), out.pts));
+  }
+
+  audio_buffer->set_timestamp(timestamp_helper_->GetTimestamp());

[DaleCurtis, 2016/02/29 21:51:30]

I believe instead of a timestamp helper you actually want a discard helper here
so that MSE's partial append window trimming will function correctly. See how
OpusAudioDecoder and FFmpegAudioDecoder use their AudioDiscardHelper classes.

[Tima Vaisburd, 2016/03/01 01:08:07]

I have a question.
As far as I can get these decoders modify the output buffer using all input
buffers that contributed to that output buffer. In MediaCodec, however, the
association between input and output buffers is lost. We can build it ourselves
based on timestamps, but this is not completely reliable. Do you see an easier
way to accomplish this and,  do we have to have this discard for this decoder?

[DaleCurtis, 2016/03/01 01:16:44]

Right now it's fine to just use the timestamp helper, but without it gapless
playback using MSE will not work (used by Play Music and likely transitively by
Cast Android receiver).

Likely to make this work we'll need to extend AudioDiscardHelper's ability to
deal with discards that span multiple buffers and (as you say) attach the input
buffers based on frame count (or timing).

+  timestamp_helper_->AddFrames(frame_count);
 
   // Copy data into AudioBuffer.
   CHECK_LE(out.size, audio_buffer->data_size());