Enable pcm_s32le audio decoding.

To accomodate this change, a new kSampleFormatS24 has been introduced.
It is only used with S24LE PCM input. FFmpeg treats pcm_s24le as an
encoded format that can be decoded into s32le samples; kSampleFormatS24
only exists on the Chrome side to distinguish FFmpeg's pcm_s24le from
pcm_s32le, both of which use Chrome's catch-all kCodecPCM codec and
the same real sample format.

BUG=557170,internal b/22378279
TEST=media_unittests --gtest_filter=AudioBufferTest.*
media_unittests --gtest_filter=PipelineIntegrationTest.S32PlaybackHashed

Review URL: https://codereview.chromium.org/1422113002

Cr-Commit-Position: refs/heads/master@{#361545}

15 files changed, 67 insertions, 33 deletions

diff --git a/DEPS b/DEPS
index ad41411..1a4785a 100644
--- a/DEPS
+++ b/DEPS
@@ -187,7 +187,7 @@ deps = {
    Var('chromium_git') + '/webm/libvpx.git' + '@' +  '204cde580a5f6dd5e7511c932c47c068046d9671',
 
   'src/third_party/ffmpeg':
-   Var('chromium_git') + '/chromium/third_party/ffmpeg.git' + '@' + 'a3f3bd78577a16bb5ec49c9d4f4ebd039efc9123',
+   Var('chromium_git') + '/chromium/third_party/ffmpeg.git' + '@' + 'f962bcb6a8aed397bb2d418527a4e4a23f5ab955',
 
   'src/third_party/libjingle/source/talk':
     Var('chromium_git') + '/external/webrtc/trunk/talk.git' + '@' + '5a1f5b5548e30780dcb5211bcdf8ce4f4a937838', # commit position 10776
diff --git a/chromecast/media/cma/base/decoder_config_adapter.cc b/chromecast/media/cma/base/decoder_config_adapter.cc
index 0bcf49c..fa8ac20 100644
--- a/chromecast/media/cma/base/decoder_config_adapter.cc
+++ b/chromecast/media/cma/base/decoder_config_adapter.cc
@@ -44,6 +44,8 @@ SampleFormat ToSampleFormat(const ::media::SampleFormat sample_format) {
       return kSampleFormatU8;
     case ::media::kSampleFormatS16:
       return kSampleFormatS16;
+    case ::media::kSampleFormatS24:
+      return kSampleFormatS24;
     case ::media::kSampleFormatS32:
       return kSampleFormatS32;
     case ::media::kSampleFormatF32:
@@ -131,6 +133,8 @@ VideoProfile ToVideoProfile(const ::media::VideoCodecProfile codec_profile) {
       return ::media::kSampleFormatU8;
     case kSampleFormatS16:
       return ::media::kSampleFormatS16;
+    case kSampleFormatS24:
+      return ::media::kSampleFormatS24;
     case kSampleFormatS32:
       return ::media::kSampleFormatS32;
     case kSampleFormatF32:
diff --git a/chromecast/public/media/decoder_config.h b/chromecast/public/media/decoder_config.h
index cf84654..e2cf78d 100644
--- a/chromecast/public/media/decoder_config.h
+++ b/chromecast/public/media/decoder_config.h
@@ -45,9 +45,10 @@ enum SampleFormat {
   kSampleFormatPlanarS16,  // Signed 16-bit planar.
   kSampleFormatPlanarF32,  // Float 32-bit planar.
   kSampleFormatPlanarS32,  // Signed 32-bit planar.
+  kSampleFormatS24,        // Signed 24-bit.
 
   kSampleFormatMin = kUnknownSampleFormat,
-  kSampleFormatMax = kSampleFormatPlanarS32,
+  kSampleFormatMax = kSampleFormatS24,
 };
 
 enum VideoCodec {
diff --git a/media/base/audio_buffer.cc b/media/base/audio_buffer.cc
index a497640..168d371 100644
--- a/media/base/audio_buffer.cc
+++ b/media/base/audio_buffer.cc
@@ -315,6 +315,7 @@ void ReadFramesInterleaved(const std::vector<uint8*>& channel_data,
       InterleaveAndConvert<int16, Dest>(
           channel_data, frames_to_copy * channel_count, trim_start, dest_data);
       break;
+    case kSampleFormatS24:
     case kSampleFormatS32:
       InterleaveAndConvert<int32, Dest>(
           channel_data, frames_to_copy * channel_count, trim_start, dest_data);
@@ -402,6 +403,7 @@ void AudioBuffer::TrimRange(int start, int end) {
         break;
       case kSampleFormatU8:
       case kSampleFormatS16:
+      case kSampleFormatS24:
       case kSampleFormatS32:
       case kSampleFormatF32: {
         // Interleaved data can be shifted all at once.
diff --git a/media/base/audio_buffer_unittest.cc b/media/base/audio_buffer_unittest.cc
index b48220d..fdb89d3 100644
--- a/media/base/audio_buffer_unittest.cc
+++ b/media/base/audio_buffer_unittest.cc
@@ -463,6 +463,7 @@ static scoped_refptr<AudioBuffer> MakeReadFramesInterleavedTestBuffer(
                                     1,
                                     frames,
                                     base::TimeDelta::FromSeconds(0));
+    case kSampleFormatS24:
     case kSampleFormatS32:
       return MakeAudioBuffer<int32>(kSampleFormatS32,
                                     channel_layout,
diff --git a/media/base/sample_format.cc b/media/base/sample_format.cc
index dd0bc1d..de91283 100644
--- a/media/base/sample_format.cc
+++ b/media/base/sample_format.cc
@@ -17,6 +17,7 @@ int SampleFormatToBytesPerChannel(SampleFormat sample_format) {
     case kSampleFormatS16:
     case kSampleFormatPlanarS16:
       return 2;
+    case kSampleFormatS24:
     case kSampleFormatS32:
     case kSampleFormatF32:
     case kSampleFormatPlanarF32:
@@ -36,6 +37,8 @@ const char* SampleFormatToString(SampleFormat sample_format) {
       return "Unsigned 8-bit with bias of 128";
     case kSampleFormatS16:
       return "Signed 16-bit";
+    case kSampleFormatS24:
+      return "Signed 24-bit";
     case kSampleFormatS32:
       return "Signed 32-bit";
     case kSampleFormatF32:
diff --git a/media/base/sample_format.h b/media/base/sample_format.h
index 0260f169..53d0cb9 100644
--- a/media/base/sample_format.h
+++ b/media/base/sample_format.h
@@ -22,9 +22,10 @@ enum SampleFormat {
   kSampleFormatPlanarS16,  // Signed 16-bit planar.
   kSampleFormatPlanarF32,  // Float 32-bit planar.
   kSampleFormatPlanarS32,  // Signed 32-bit planar.
+  kSampleFormatS24,        // Signed 24-bit.
 
   // Must always be equal to largest value ever logged.
-  kSampleFormatMax = kSampleFormatPlanarS32,
+  kSampleFormatMax = kSampleFormatS24,
 };
 
 // Returns the number of bytes used per channel for the specified
diff --git a/media/cast/test/fake_media_source.cc b/media/cast/test/fake_media_source.cc
index b8820b7..24327d0 100644
--- a/media/cast/test/fake_media_source.cc
+++ b/media/cast/test/fake_media_source.cc
@@ -470,18 +470,14 @@ void FakeMediaSource::DecodeAudio(ScopedAVPacket packet) {
       audio_sent_ts_->SetBaseTimestamp(base_ts);
     }
 
-    scoped_refptr<AudioBuffer> buffer =
-        AudioBuffer::CopyFrom(
-            AVSampleFormatToSampleFormat(
-                av_audio_context()->sample_fmt),
-            ChannelLayoutToChromeChannelLayout(
-                av_audio_context()->channel_layout,
-                av_audio_context()->channels),
-            av_audio_context()->channels,
-            av_audio_context()->sample_rate,
-            frames_read,
-            &avframe->data[0],
-            PtsToTimeDelta(avframe->pkt_pts, av_audio_stream()->time_base));
+    scoped_refptr<AudioBuffer> buffer = AudioBuffer::CopyFrom(
+        AVSampleFormatToSampleFormat(av_audio_context()->sample_fmt,
+                                     av_audio_context()->codec_id),
+        ChannelLayoutToChromeChannelLayout(av_audio_context()->channel_layout,
+                                           av_audio_context()->channels),
+        av_audio_context()->channels, av_audio_context()->sample_rate,
+        frames_read, &avframe->data[0],
+        PtsToTimeDelta(avframe->pkt_pts, av_audio_stream()->time_base));
     audio_algo_.EnqueueBuffer(buffer);
     av_frame_unref(avframe);
   } while (packet_temp.size > 0);
diff --git a/media/ffmpeg/ffmpeg_common.cc b/media/ffmpeg/ffmpeg_common.cc
index c9d8669..772b41d 100644
--- a/media/ffmpeg/ffmpeg_common.cc
+++ b/media/ffmpeg/ffmpeg_common.cc
@@ -74,6 +74,7 @@ static AudioCodec CodecIDToAudioCodec(AVCodecID codec_id) {
     case AV_CODEC_ID_PCM_U8:
     case AV_CODEC_ID_PCM_S16LE:
     case AV_CODEC_ID_PCM_S24LE:
+    case AV_CODEC_ID_PCM_S32LE:
     case AV_CODEC_ID_PCM_F32LE:
       return kCodecPCM;
     case AV_CODEC_ID_PCM_S16BE:
@@ -117,8 +118,10 @@ static AVCodecID AudioCodecToCodecID(AudioCodec audio_codec,
           return AV_CODEC_ID_PCM_U8;
         case kSampleFormatS16:
           return AV_CODEC_ID_PCM_S16LE;
-        case kSampleFormatS32:
+        case kSampleFormatS24:
           return AV_CODEC_ID_PCM_S24LE;
+        case kSampleFormatS32:
+          return AV_CODEC_ID_PCM_S32LE;
         case kSampleFormatF32:
           return AV_CODEC_ID_PCM_F32LE;
         default:
@@ -244,14 +247,18 @@ static int VideoCodecProfileToProfileID(VideoCodecProfile profile) {
   return FF_PROFILE_UNKNOWN;
 }
 
-SampleFormat AVSampleFormatToSampleFormat(AVSampleFormat sample_format) {
+SampleFormat AVSampleFormatToSampleFormat(AVSampleFormat sample_format,
+                                          AVCodecID codec_id) {
   switch (sample_format) {
     case AV_SAMPLE_FMT_U8:
       return kSampleFormatU8;
     case AV_SAMPLE_FMT_S16:
       return kSampleFormatS16;
     case AV_SAMPLE_FMT_S32:
-      return kSampleFormatS32;
+      if (codec_id == AV_CODEC_ID_PCM_S24LE)
+        return kSampleFormatS24;
+      else
+        return kSampleFormatS32;
     case AV_SAMPLE_FMT_FLT:
       return kSampleFormatF32;
     case AV_SAMPLE_FMT_S16P:
@@ -272,6 +279,8 @@ static AVSampleFormat SampleFormatToAVSampleFormat(SampleFormat sample_format) {
       return AV_SAMPLE_FMT_U8;
     case kSampleFormatS16:
       return AV_SAMPLE_FMT_S16;
+    // pcm_s24le is treated as a codec with sample format s32 in ffmpeg
+    case kSampleFormatS24:
     case kSampleFormatS32:
       return AV_SAMPLE_FMT_S32;
     case kSampleFormatF32:
@@ -293,8 +302,8 @@ bool AVCodecContextToAudioDecoderConfig(const AVCodecContext* codec_context,
 
   AudioCodec codec = CodecIDToAudioCodec(codec_context->codec_id);
 
-  SampleFormat sample_format =
-      AVSampleFormatToSampleFormat(codec_context->sample_fmt);
+  SampleFormat sample_format = AVSampleFormatToSampleFormat(
+      codec_context->sample_fmt, codec_context->codec_id);
 
   ChannelLayout channel_layout = ChannelLayoutToChromeChannelLayout(
       codec_context->channel_layout, codec_context->channels);
diff --git a/media/ffmpeg/ffmpeg_common.h b/media/ffmpeg/ffmpeg_common.h
index 7938a32..955ee64 100644
--- a/media/ffmpeg/ffmpeg_common.h
+++ b/media/ffmpeg/ffmpeg_common.h
@@ -123,7 +123,7 @@ MEDIA_EXPORT AVCodecID VideoCodecToCodecID(VideoCodec video_codec);
 
 // Converts FFmpeg's audio sample format to Chrome's SampleFormat.
 MEDIA_EXPORT SampleFormat
-AVSampleFormatToSampleFormat(AVSampleFormat sample_format);
+AVSampleFormatToSampleFormat(AVSampleFormat sample_format, AVCodecID codec_id);
 
 // Converts FFmpeg's pixel formats to its corresponding supported video format.
 MEDIA_EXPORT VideoPixelFormat
diff --git a/media/ffmpeg/ffmpeg_common_unittest.cc b/media/ffmpeg/ffmpeg_common_unittest.cc
index aa8a06d..9a73587 100644
--- a/media/ffmpeg/ffmpeg_common_unittest.cc
+++ b/media/ffmpeg/ffmpeg_common_unittest.cc
@@ -161,17 +161,24 @@ TEST_F(FFmpegCommonTest, VerifyFormatSizes) {
   for (AVSampleFormat format = AV_SAMPLE_FMT_NONE;
        format < AV_SAMPLE_FMT_NB;
        format = static_cast<AVSampleFormat>(format + 1)) {
-    SampleFormat sample_format = AVSampleFormatToSampleFormat(format);
-    if (sample_format == kUnknownSampleFormat) {
-      // This format not supported, so skip it.
-      continue;
+    std::vector<AVCodecID> codec_ids(1, AV_CODEC_ID_NONE);
+    if (format == AV_SAMPLE_FMT_S32)
+      codec_ids.push_back(AV_CODEC_ID_PCM_S24LE);
+    for (const auto& codec_id : codec_ids) {
+      SampleFormat sample_format =
+          AVSampleFormatToSampleFormat(format, codec_id);
+      if (sample_format == kUnknownSampleFormat) {
+        // This format not supported, so skip it.
+        continue;
+      }
+
+      // Have FFMpeg compute the size of a buffer of 1 channel / 1 frame
+      // with 1 byte alignment to make sure the sizes match.
+      int single_buffer_size =
+          av_samples_get_buffer_size(NULL, 1, 1, format, 1);
+      int bytes_per_channel = SampleFormatToBytesPerChannel(sample_format);
+      EXPECT_EQ(bytes_per_channel, single_buffer_size);
     }
-
-    // Have FFMpeg compute the size of a buffer of 1 channel / 1 frame
-    // with 1 byte alignment to make sure the sizes match.
-    int single_buffer_size = av_samples_get_buffer_size(NULL, 1, 1, format, 1);
-    int bytes_per_channel = SampleFormatToBytesPerChannel(sample_format);
-    EXPECT_EQ(bytes_per_channel, single_buffer_size);
   }
 }
 
diff --git a/media/filters/ffmpeg_audio_decoder.cc b/media/filters/ffmpeg_audio_decoder.cc
index 206f971..2ad55cd 100644
--- a/media/filters/ffmpeg_audio_decoder.cc
+++ b/media/filters/ffmpeg_audio_decoder.cc
@@ -58,7 +58,8 @@ static int GetAudioBuffer(struct AVCodecContext* s, AVFrame* frame, int flags) {
   // data, use the values supplied by FFmpeg (ignoring the current settings).
   // FFmpegDecode() gets to determine if the buffer is useable or not.
   AVSampleFormat format = static_cast<AVSampleFormat>(frame->format);
-  SampleFormat sample_format = AVSampleFormatToSampleFormat(format);
+  SampleFormat sample_format =
+      AVSampleFormatToSampleFormat(format, s->codec_id);
   int channels = DetermineChannels(frame);
   if (channels <= 0 || channels >= limits::kMaxChannels) {
     DLOG(ERROR) << "Requested number of channels (" << channels
diff --git a/media/mojo/interfaces/media_types.mojom b/media/mojo/interfaces/media_types.mojom
index 78784f9..befd80a 100644
--- a/media/mojo/interfaces/media_types.mojom
+++ b/media/mojo/interfaces/media_types.mojom
@@ -86,7 +86,8 @@ enum SampleFormat {
   PlanarS16,
   PlanarF32,
   PlanarS32,
-  Max = PlanarS32,
+  S24,
+  Max = S24,
 };
 
 // See media/base/video_types.h for descriptions.
diff --git a/media/test/data/sfx_s32le.wav b/media/test/data/sfx_s32le.wav
new file mode 100644
index 0000000..5f670a9
--- /dev/null
+++ b/media/test/data/sfx_s32le.wav
diff --git a/media/test/pipeline_integration_test.cc b/media/test/pipeline_integration_test.cc
index b19b32c..69da770 100644
--- a/media/test/pipeline_integration_test.cc
+++ b/media/test/pipeline_integration_test.cc
@@ -909,6 +909,14 @@ TEST_F(PipelineIntegrationTest, BasicPlaybackLive) {
             demuxer_->GetTimelineOffset());
 }
 
+TEST_F(PipelineIntegrationTest, S32PlaybackHashed) {
+  ASSERT_EQ(PIPELINE_OK, Start("sfx_s32le.wav", kHashed));
+  Play();
+  ASSERT_TRUE(WaitUntilOnEnded());
+  EXPECT_HASH_EQ(std::string(kNullVideoHash), GetVideoHash());
+  EXPECT_HASH_EQ("3.03,2.86,2.99,3.31,3.57,4.06,", GetAudioHash());
+}
+
 TEST_F(PipelineIntegrationTest, F32PlaybackHashed) {
   ASSERT_EQ(PIPELINE_OK, Start("sfx_f32le.wav", kHashed));
   Play();