Basic media tracks implementation for media stream parsers

This CL adds basic media::MediaTrack implementation, media::MediaTracks
collection that associates media tracks with audio/video configs, and
makes necessary changes in MSE stream parsers to pass around the
MediaTracks collection instead of individual audio/video configs.

BUG=249427, 249428

Review URL: https://codereview.chromium.org/1716503002

Cr-Commit-Position: refs/heads/master@{#378583}

21 files changed, 340 insertions, 74 deletions

diff --git a/media/base/BUILD.gn b/media/base/BUILD.gn
index 0e4bebc..6d37637 100644
--- a/media/base/BUILD.gn
+++ b/media/base/BUILD.gn
@@ -135,6 +135,10 @@ source_set("base") {
     "media_resources.h",
     "media_switches.cc",
     "media_switches.h",
+    "media_track.cc",
+    "media_track.h",
+    "media_tracks.cc",
+    "media_tracks.h",
     "media_util.cc",
     "media_util.h",
     "mime_util.cc",
diff --git a/media/base/media_track.cc b/media/base/media_track.cc
new file mode 100644
index 0000000..ad5178f
--- /dev/null
+++ b/media/base/media_track.cc
@@ -0,0 +1,18 @@
+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "media/base/media_track.h"
+
+namespace media {
+
+MediaTrack::MediaTrack(Type type,
+                       const std::string& id,
+                       const std::string& kind,
+                       const std::string& label,
+                       const std::string& lang)
+    : type_(type), id_(id), kind_(kind), label_(label), language_(lang) {}
+
+MediaTrack::~MediaTrack() {}
+
+}  // namespace media
diff --git a/media/base/media_track.h b/media/base/media_track.h
new file mode 100644
index 0000000..a5cb6d3
--- /dev/null
+++ b/media/base/media_track.h
@@ -0,0 +1,41 @@
+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef MEDIA_BASE_MEDIA_TRACK_H_
+#define MEDIA_BASE_MEDIA_TRACK_H_
+
+#include <string>
+
+#include "media/base/media_export.h"
+
+namespace media {
+
+class MEDIA_EXPORT MediaTrack {
+ public:
+  enum Type { Text, Audio, Video };
+  MediaTrack(Type type,
+             const std::string& id,
+             const std::string& kind,
+             const std::string& label,
+             const std::string& lang);
+  ~MediaTrack();
+
+  Type type() const { return type_; }
+
+  const std::string& id() const { return id_; }
+  const std::string& kind() const { return kind_; }
+  const std::string& label() const { return label_; }
+  const std::string& language() const { return language_; }
+
+ private:
+  Type type_;
+  std::string id_;
+  std::string kind_;
+  std::string label_;
+  std::string language_;
+};
+
+}  // namespace media
+
+#endif  // MEDIA_BASE_MEDIA_TRACK_H_
diff --git a/media/base/media_tracks.cc b/media/base/media_tracks.cc
new file mode 100644
index 0000000..832e5a8
--- /dev/null
+++ b/media/base/media_tracks.cc
@@ -0,0 +1,81 @@
+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "media/base/media_tracks.h"
+
+#include "base/bind.h"
+#include "media/base/audio_decoder_config.h"
+#include "media/base/video_decoder_config.h"
+
+namespace media {
+
+MediaTracks::MediaTracks() {}
+
+MediaTracks::~MediaTracks() {}
+
+void MediaTracks::AddAudioTrack(const AudioDecoderConfig& config,
+                                const std::string& id,
+                                const std::string& kind,
+                                const std::string& label,
+                                const std::string& language) {
+  DCHECK(config.IsValidConfig());
+  CHECK(audio_configs_.find(id) == audio_configs_.end());
+  scoped_ptr<MediaTrack> track = make_scoped_ptr(
+      new MediaTrack(MediaTrack::Audio, id, kind, label, language));
+  tracks_.push_back(std::move(track));
+  audio_configs_[id] = config;
+}
+
+void MediaTracks::AddVideoTrack(const VideoDecoderConfig& config,
+                                const std::string& id,
+                                const std::string& kind,
+                                const std::string& label,
+                                const std::string& language) {
+  DCHECK(config.IsValidConfig());
+  CHECK(video_configs_.find(id) == video_configs_.end());
+  scoped_ptr<MediaTrack> track = make_scoped_ptr(
+      new MediaTrack(MediaTrack::Video, id, kind, label, language));
+  tracks_.push_back(std::move(track));
+  video_configs_[id] = config;
+}
+
+const AudioDecoderConfig& MediaTracks::getAudioConfig(
+    const std::string& id) const {
+  auto it = audio_configs_.find(id);
+  if (it != audio_configs_.end())
+    return it->second;
+  static AudioDecoderConfig invalidConfig;
+  return invalidConfig;
+}
+
+const VideoDecoderConfig& MediaTracks::getVideoConfig(
+    const std::string& id) const {
+  auto it = video_configs_.find(id);
+  if (it != video_configs_.end())
+    return it->second;
+  static VideoDecoderConfig invalidConfig;
+  return invalidConfig;
+}
+
+const AudioDecoderConfig& MediaTracks::getFirstAudioConfig() const {
+  for (const auto& track : tracks()) {
+    if (track->type() == MediaTrack::Audio) {
+      return getAudioConfig(track->id());
+    }
+  }
+  static AudioDecoderConfig invalidConfig;
+  return invalidConfig;
+}
+
+const VideoDecoderConfig& MediaTracks::getFirstVideoConfig() const {
+  for (const auto& track : tracks()) {
+    if (track->type() == MediaTrack::Video) {
+      return getVideoConfig(track->id());
+    }
+  }
+  static VideoDecoderConfig invalidConfig;
+  return invalidConfig;
+}
+
+}  // namespace media
diff --git a/media/base/media_tracks.h b/media/base/media_tracks.h
new file mode 100644
index 0000000..4e0fbb0
--- /dev/null
+++ b/media/base/media_tracks.h
@@ -0,0 +1,62 @@
+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef MEDIA_BASE_MEDIA_TRACKS_H_
+#define MEDIA_BASE_MEDIA_TRACKS_H_
+
+#include <map>
+#include <string>
+#include <vector>
+
+#include "base/macros.h"
+#include "base/memory/scoped_ptr.h"
+#include "media/base/media_export.h"
+#include "media/base/media_track.h"
+
+namespace media {
+
+class AudioDecoderConfig;
+class VideoDecoderConfig;
+
+class MEDIA_EXPORT MediaTracks {
+ public:
+  typedef std::vector<scoped_ptr<MediaTrack>> MediaTracksCollection;
+
+  MediaTracks();
+  ~MediaTracks();
+
+  // Callers need to ensure that track id is unique.
+  void AddAudioTrack(const AudioDecoderConfig& config,
+                     const std::string& id,
+                     const std::string& kind,
+                     const std::string& label,
+                     const std::string& language);
+  // Callers need to ensure that track id is unique.
+  void AddVideoTrack(const VideoDecoderConfig& config,
+                     const std::string& id,
+                     const std::string& kind,
+                     const std::string& label,
+                     const std::string& language);
+
+  const MediaTracksCollection& tracks() const { return tracks_; }
+
+  const AudioDecoderConfig& getAudioConfig(const std::string& id) const;
+  const VideoDecoderConfig& getVideoConfig(const std::string& id) const;
+
+  // TODO(servolk): These are temporary helpers useful until all code paths are
+  // converted to properly handle multiple media tracks.
+  const AudioDecoderConfig& getFirstAudioConfig() const;
+  const VideoDecoderConfig& getFirstVideoConfig() const;
+
+ private:
+  MediaTracksCollection tracks_;
+  std::map<std::string, AudioDecoderConfig> audio_configs_;
+  std::map<std::string, VideoDecoderConfig> video_configs_;
+
+  DISALLOW_COPY_AND_ASSIGN(MediaTracks);
+};
+
+}  // namespace media
+
+#endif  // MEDIA_BASE_MEDIA_TRACKS_H_
diff --git a/media/base/stream_parser.h b/media/base/stream_parser.h
index 11740c9..4b9fa7f 100644
--- a/media/base/stream_parser.h
+++ b/media/base/stream_parser.h
@@ -25,10 +25,9 @@
 
 namespace media {
 
-class AudioDecoderConfig;
+class MediaTracks;
 class StreamParserBuffer;
 class TextTrackConfig;
-class VideoDecoderConfig;
 
 // Abstract interface for parsing media byte streams.
 class MEDIA_EXPORT StreamParser {
@@ -76,18 +75,17 @@ class MEDIA_EXPORT StreamParser {
   typedef base::Callback<void(const InitParameters& params)> InitCB;
 
   // Indicates when new stream configurations have been parsed.
-  // First parameter - The new audio configuration. If the config is not valid
-  //                   then it means that there isn't an audio stream.
-  // Second parameter - The new video configuration. If the config is not valid
-  //                    then it means that there isn't an audio stream.
-  // Third parameter - The new text tracks configuration.  If the map is empty,
-  //                   then no text tracks were parsed from the stream.
+  // First parameter - An object containing information about media tracks as
+  //                   well as audio/video decoder configs associated with each
+  //                   track.
+  // Second parameter - The new text tracks configuration.  If the map is empty,
+  //                    then no text tracks were parsed from the stream.
   // Return value - True if the new configurations are accepted.
   //                False if the new configurations are not supported
   //                and indicates that a parsing error should be signalled.
-  typedef base::Callback<bool(const AudioDecoderConfig&,
-                              const VideoDecoderConfig&,
-                              const TextTrackConfigMap&)> NewConfigCB;
+  typedef base::Callback<bool(scoped_ptr<MediaTracks>,
+                              const TextTrackConfigMap&)>
+      NewConfigCB;
 
   // New stream buffers have been parsed.
   // First parameter - A queue of newly parsed audio buffers.
diff --git a/media/blink/websourcebuffer_impl.cc b/media/blink/websourcebuffer_impl.cc
index ffb62c5..ea09bde 100644
--- a/media/blink/websourcebuffer_impl.cc
+++ b/media/blink/websourcebuffer_impl.cc
@@ -12,6 +12,7 @@
 #include "base/bind.h"
 #include "base/callback.h"
 #include "base/callback_helpers.h"
+#include "media/base/media_tracks.h"
 #include "media/base/timestamp_constants.h"
 #include "media/filters/chunk_demuxer.h"
 #include "third_party/WebKit/public/platform/WebSourceBufferClient.h"
@@ -158,8 +159,10 @@ void WebSourceBufferImpl::removedFromMediaSource() {
   client_ = NULL;
 }
 
-void WebSourceBufferImpl::InitSegmentReceived() {
+void WebSourceBufferImpl::InitSegmentReceived(const MediaTracks& tracks) {
   DVLOG(1) << __FUNCTION__;
+  // TODO(servolk): Implement passing MediaTrack info to blink level.
+  // https://crbug.com/249428
   client_->initializationSegmentReceived();
 }
 
diff --git a/media/blink/websourcebuffer_impl.h b/media/blink/websourcebuffer_impl.h
index 9111fd2..1dc54e6 100644
--- a/media/blink/websourcebuffer_impl.h
+++ b/media/blink/websourcebuffer_impl.h
@@ -16,6 +16,7 @@
 
 namespace media {
 class ChunkDemuxer;
+class MediaTracks;
 
 class WebSourceBufferImpl : public blink::WebSourceBuffer {
  public:
@@ -42,7 +43,7 @@ class WebSourceBufferImpl : public blink::WebSourceBuffer {
  private:
   // Demuxer callback handler to process an initialization segment received
   // during an append() call.
-  void InitSegmentReceived();
+  void InitSegmentReceived(const MediaTracks& tracks);
 
   std::string id_;
   ChunkDemuxer* demuxer_;  // Owned by WebMediaPlayerImpl.
diff --git a/media/filters/chunk_demuxer_unittest.cc b/media/filters/chunk_demuxer_unittest.cc
index c095f5a..bd2a87e 100644
--- a/media/filters/chunk_demuxer_unittest.cc
+++ b/media/filters/chunk_demuxer_unittest.cc
@@ -18,6 +18,7 @@
 #include "media/base/audio_decoder_config.h"
 #include "media/base/decoder_buffer.h"
 #include "media/base/decrypt_config.h"
+#include "media/base/media_tracks.h"
 #include "media/base/mock_demuxer_host.h"
 #include "media/base/mock_media_log.h"
 #include "media/base/test_data_util.h"
@@ -841,7 +842,7 @@ class ChunkDemuxerTest : public ::testing::Test {
     // it.
     if (stream_flags != 0) {
       ExpectInitMediaLogs(stream_flags);
-      EXPECT_CALL(*this, InitSegmentReceived());
+      EXPECT_CALL(*this, InitSegmentReceived(_));
     } else {
       // OnNewConfigs() requires at least one audio, video, or text track.
       EXPECT_MEDIA_LOG(StreamParsingFailed());
@@ -883,11 +884,11 @@ class ChunkDemuxerTest : public ::testing::Test {
     // incompatible with InSequence tests. Refactoring of the duration
     // set expectation to not be added during CreateInitDoneCB() could fix this.
     ExpectInitMediaLogs(audio_flags);
-    EXPECT_CALL(*this, InitSegmentReceived());
+    EXPECT_CALL(*this, InitSegmentReceived(_));
     AppendInitSegmentWithSourceId(audio_id, audio_flags);
 
     ExpectInitMediaLogs(video_flags);
-    EXPECT_CALL(*this, InitSegmentReceived());
+    EXPECT_CALL(*this, InitSegmentReceived(_));
     AppendInitSegmentWithSourceId(video_id, video_flags);
     return true;
   }
@@ -923,7 +924,7 @@ class ChunkDemuxerTest : public ::testing::Test {
     // Adding expectation prior to CreateInitDoneCB() here because InSequence
     // tests require init segment received before duration set.
     ExpectInitMediaLogs(HAS_AUDIO | HAS_VIDEO);
-    EXPECT_CALL(*this, InitSegmentReceived());
+    EXPECT_CALL(*this, InitSegmentReceived(_));
     demuxer_->Initialize(
         &host_, CreateInitDoneCB(base::TimeDelta::FromMilliseconds(2744),
                                  PIPELINE_OK), true);
@@ -949,7 +950,7 @@ class ChunkDemuxerTest : public ::testing::Test {
     // media/test/data/bear-320x240-manifest.js which were
     // generated from media/test/data/bear-640x360.webm and
     // media/test/data/bear-320x240.webm respectively.
-    EXPECT_CALL(*this, InitSegmentReceived());
+    EXPECT_CALL(*this, InitSegmentReceived(_));
     AppendData(bear2->data(), 4340);
 
     // Append a media segment that goes from [0.527000, 1.014000).
@@ -960,7 +961,7 @@ class ChunkDemuxerTest : public ::testing::Test {
 
     // Append initialization segment for bear1 & fill gap with [779-1197)
     // segment.
-    EXPECT_CALL(*this, InitSegmentReceived());
+    EXPECT_CALL(*this, InitSegmentReceived(_));
     AppendData(bear1->data(), 4370);
     EXPECT_MEDIA_LOG(WebMSimpleBlockDurationEstimated(23));
     EXPECT_MEDIA_LOG(GeneratedSplice(26000, 779000));
@@ -1308,7 +1309,7 @@ class ChunkDemuxerTest : public ::testing::Test {
 
     // Read a WebM file into memory and send the data to the demuxer.
     scoped_refptr<DecoderBuffer> buffer = ReadTestDataFile(filename);
-    EXPECT_CALL(*this, InitSegmentReceived());
+    EXPECT_CALL(*this, InitSegmentReceived(_));
     AppendDataInPieces(buffer->data(), buffer->data_size(), 512);
 
     // Verify that the timestamps on the first few packets match what we
@@ -1345,7 +1346,7 @@ class ChunkDemuxerTest : public ::testing::Test {
                void(EmeInitDataType init_data_type,
                     const std::vector<uint8_t>& init_data));
 
-  MOCK_METHOD0(InitSegmentReceived, void(void));
+  MOCK_METHOD1(InitSegmentReceived, void(const MediaTracks&));
 
   void Seek(base::TimeDelta seek_time) {
     demuxer_->StartWaitingForSeek(seek_time);
@@ -1547,7 +1548,7 @@ TEST_F(ChunkDemuxerTest, SingleTextTrackIdChange) {
   CreateInitSegmentWithAlternateTextTrackNum(HAS_TEXT | HAS_AUDIO | HAS_VIDEO,
                                              false, false,
                                              &info_tracks, &info_tracks_size);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   demuxer_->AppendData(kSourceId, info_tracks.get(), info_tracks_size,
                        append_window_start_for_next_append_,
                        append_window_end_for_next_append_,
@@ -1592,7 +1593,7 @@ TEST_F(ChunkDemuxerTest, InitSegmentSetsNeedRandomAccessPointFlag) {
       MuxedStreamInfo(kTextTrackNum, "25K 40K"));
   CheckExpectedRanges("{ [23,46) }");
 
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegment(HAS_TEXT | HAS_AUDIO | HAS_VIDEO);
   AppendMuxedCluster(MuxedStreamInfo(kAudioTrackNum, "46K 69K", 23),
                      MuxedStreamInfo(kVideoTrackNum, "60 90K", 30),
@@ -1616,7 +1617,7 @@ TEST_F(ChunkDemuxerTest, Shutdown_BeforeAllInitSegmentsAppended) {
   EXPECT_EQ(AddId("video", HAS_VIDEO), ChunkDemuxer::kOk);
 
   ExpectInitMediaLogs(HAS_AUDIO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegmentWithSourceId("audio", HAS_AUDIO);
 
   ShutdownDemuxer();
@@ -1635,7 +1636,7 @@ TEST_F(ChunkDemuxerTest, Shutdown_BeforeAllInitSegmentsAppendedText) {
       .Times(Exactly(1));
 
   ExpectInitMediaLogs(HAS_VIDEO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegmentWithSourceId("video_and_text", HAS_VIDEO | HAS_TEXT);
 
   ShutdownDemuxer();
@@ -2157,7 +2158,7 @@ TEST_F(ChunkDemuxerTest, AppendingInPieces) {
   dst += cluster_b->size();
 
   ExpectInitMediaLogs(HAS_AUDIO | HAS_VIDEO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendDataInPieces(buffer.get(), buffer_size);
 
   GenerateExpectedReads(0, 9);
@@ -2412,7 +2413,7 @@ TEST_F(ChunkDemuxerTest, MultipleHeaders) {
   AppendCluster(kDefaultFirstCluster());
 
   // Append another identical initialization segment.
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegment(HAS_AUDIO | HAS_VIDEO);
 
   AppendCluster(kDefaultSecondCluster());
@@ -2468,7 +2469,7 @@ TEST_F(ChunkDemuxerTest, AddIdFailures) {
   ASSERT_EQ(AddId(), ChunkDemuxer::kReachedIdLimit);
 
   ExpectInitMediaLogs(HAS_AUDIO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegmentWithSourceId(audio_id, HAS_AUDIO);
 
   // Adding an id after append should fail.
@@ -2712,7 +2713,7 @@ TEST_F(ChunkDemuxerTest, GetBufferedRanges_AudioIdOnly) {
 
   ASSERT_EQ(AddId(kSourceId, HAS_AUDIO), ChunkDemuxer::kOk);
   ExpectInitMediaLogs(HAS_AUDIO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegment(HAS_AUDIO);
 
   // Test a simple cluster.
@@ -2736,7 +2737,7 @@ TEST_F(ChunkDemuxerTest, GetBufferedRanges_VideoIdOnly) {
 
   ASSERT_EQ(AddId(kSourceId, HAS_VIDEO), ChunkDemuxer::kOk);
   ExpectInitMediaLogs(HAS_VIDEO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendInitSegment(HAS_VIDEO);
 
   // Test a simple cluster.
@@ -3406,7 +3407,7 @@ TEST_F(ChunkDemuxerTest, EmitBuffersDuringAbort) {
   //        PTS: 353788 (0x000565fc)  [= 90 kHz-Timestamp: 0:00:03.9309]
 
   scoped_refptr<DecoderBuffer> buffer = ReadTestDataFile("bear-1280x720.ts");
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendData(kSourceId, buffer->data(), buffer->data_size());
 
   // Confirm we're in the middle of parsing a media segment.
@@ -3454,7 +3455,7 @@ TEST_F(ChunkDemuxerTest, SeekCompleteDuringAbort) {
   //        PTS: 353788 (0x000565fc)  [= 90 kHz-Timestamp: 0:00:03.9309]
 
   scoped_refptr<DecoderBuffer> buffer = ReadTestDataFile("bear-1280x720.ts");
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   AppendData(kSourceId, buffer->data(), buffer->data_size());
 
   // Confirm we're in the middle of parsing a media segment.
@@ -4065,7 +4066,7 @@ TEST_F(ChunkDemuxerTest, AppendWindow_WebMFile_AudioOnly) {
   scoped_refptr<DecoderBuffer> buffer =
       ReadTestDataFile("bear-320x240-audio-only.webm");
   ExpectInitMediaLogs(HAS_AUDIO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   EXPECT_MEDIA_LOG(WebMSimpleBlockDurationEstimated(2));
   AppendDataInPieces(buffer->data(), buffer->data_size(), 128);
 
@@ -4092,7 +4093,7 @@ TEST_F(ChunkDemuxerTest, AppendWindow_AudioConfigUpdateRemovesPreroll) {
   scoped_refptr<DecoderBuffer> buffer =
       ReadTestDataFile("bear-320x240-audio-only.webm");
   ExpectInitMediaLogs(HAS_AUDIO);
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   EXPECT_MEDIA_LOG(WebMSimpleBlockDurationEstimated(2));
   AppendDataInPieces(buffer->data(), buffer->data_size(), 512);
   CheckExpectedRanges("{ }");
@@ -4103,7 +4104,7 @@ TEST_F(ChunkDemuxerTest, AppendWindow_AudioConfigUpdateRemovesPreroll) {
   // Read a second WebM with a different config in and append the data.
   scoped_refptr<DecoderBuffer> buffer2 =
       ReadTestDataFile("bear-320x240-audio-only-48khz.webm");
-  EXPECT_CALL(*this, InitSegmentReceived());
+  EXPECT_CALL(*this, InitSegmentReceived(_));
   EXPECT_MEDIA_LOG(WebMSimpleBlockDurationEstimated(21));
   EXPECT_CALL(host_, SetDuration(_)).Times(AnyNumber());
   ASSERT_TRUE(SetTimestampOffset(kSourceId, duration_1));
diff --git a/media/filters/media_source_state.cc b/media/filters/media_source_state.cc
index ede9605..8f2f87c 100644
--- a/media/filters/media_source_state.cc
+++ b/media/filters/media_source_state.cc
@@ -6,6 +6,8 @@
 
 #include "base/callback_helpers.h"
 #include "base/stl_util.h"
+#include "media/base/media_track.h"
+#include "media/base/media_tracks.h"
 #include "media/filters/chunk_demuxer.h"
 #include "media/filters/frame_processor.h"
 #include "media/filters/source_buffer_stream.h"
@@ -468,9 +470,13 @@ bool MediaSourceState::IsSeekWaitingForData() const {
 bool MediaSourceState::OnNewConfigs(
     bool allow_audio,
     bool allow_video,
-    const AudioDecoderConfig& audio_config,
-    const VideoDecoderConfig& video_config,
+    scoped_ptr<MediaTracks> tracks,
     const StreamParser::TextTrackConfigMap& text_configs) {
+  DCHECK(tracks.get());
+  media_tracks_ = std::move(tracks);
+  const AudioDecoderConfig& audio_config = media_tracks_->getFirstAudioConfig();
+  const VideoDecoderConfig& video_config = media_tracks_->getFirstVideoConfig();
+
   DVLOG(1) << "OnNewConfigs(" << allow_audio << ", " << allow_video << ", "
            << audio_config.IsValidConfig() << ", "
            << video_config.IsValidConfig() << ")";
@@ -638,7 +644,7 @@ bool MediaSourceState::OnNewConfigs(
 
   DVLOG(1) << "OnNewConfigs() : " << (success ? "success" : "failed");
   if (success)
-    init_segment_received_cb_.Run();
+    init_segment_received_cb_.Run(*media_tracks_);
 
   return success;
 }
diff --git a/media/filters/media_source_state.h b/media/filters/media_source_state.h
index 1a8e42a..1197015 100644
--- a/media/filters/media_source_state.h
+++ b/media/filters/media_source_state.h
@@ -28,7 +28,7 @@ class MEDIA_EXPORT MediaSourceState {
   typedef base::Callback<ChunkDemuxerStream*(DemuxerStream::Type)>
       CreateDemuxerStreamCB;
 
-  typedef base::Closure InitSegmentReceivedCB;
+  typedef base::Callback<void(const MediaTracks&)> InitSegmentReceivedCB;
 
   typedef base::Callback<void(ChunkDemuxerStream*, const TextTrackConfig&)>
       NewTextTrackCB;
@@ -124,8 +124,7 @@ class MEDIA_EXPORT MediaSourceState {
   // processing decoder configurations.
   bool OnNewConfigs(bool allow_audio,
                     bool allow_video,
-                    const AudioDecoderConfig& audio_config,
-                    const VideoDecoderConfig& video_config,
+                    scoped_ptr<MediaTracks> tracks,
                     const StreamParser::TextTrackConfigMap& text_configs);
 
   // Called by the |stream_parser_| at the beginning of a new media segment.
@@ -183,6 +182,8 @@ class MEDIA_EXPORT MediaSourceState {
   // The object used to parse appended data.
   scoped_ptr<StreamParser> stream_parser_;
 
+  scoped_ptr<MediaTracks> media_tracks_;
+
   ChunkDemuxerStream* audio_;  // Not owned by |this|.
   ChunkDemuxerStream* video_;  // Not owned by |this|.
 
diff --git a/media/formats/common/stream_parser_test_base.cc b/media/formats/common/stream_parser_test_base.cc
index de80bac..b42173f 100644
--- a/media/formats/common/stream_parser_test_base.cc
+++ b/media/formats/common/stream_parser_test_base.cc
@@ -7,6 +7,8 @@
 #include <utility>
 
 #include "base/bind.h"
+#include "media/base/media_track.h"
+#include "media/base/media_tracks.h"
 #include "media/base/test_data_util.h"
 #include "testing/gtest/include/gtest/gtest.h"
 
@@ -83,14 +85,13 @@ void StreamParserTestBase::OnInitDone(
 }
 
 bool StreamParserTestBase::OnNewConfig(
-    const AudioDecoderConfig& audio_config,
-    const VideoDecoderConfig& video_config,
+    scoped_ptr<MediaTracks> tracks,
     const StreamParser::TextTrackConfigMap& text_config) {
-  DVLOG(1) << __FUNCTION__ << "(" << audio_config.IsValidConfig() << ", "
-           << video_config.IsValidConfig() << ")";
-  EXPECT_TRUE(audio_config.IsValidConfig());
-  EXPECT_FALSE(video_config.IsValidConfig());
-  last_audio_config_ = audio_config;
+  DVLOG(1) << __FUNCTION__ << " media tracks count=" << tracks->tracks().size();
+  EXPECT_EQ(tracks->tracks().size(), 1u);
+  EXPECT_TRUE(tracks->getFirstAudioConfig().IsValidConfig());
+  EXPECT_FALSE(tracks->getFirstVideoConfig().IsValidConfig());
+  last_audio_config_ = tracks->getFirstAudioConfig();
   return true;
 }
 
diff --git a/media/formats/common/stream_parser_test_base.h b/media/formats/common/stream_parser_test_base.h
index 42cb8d1..bdbfe7a 100644
--- a/media/formats/common/stream_parser_test_base.h
+++ b/media/formats/common/stream_parser_test_base.h
@@ -57,8 +57,7 @@ class StreamParserTestBase {
                           size_t length,
                           size_t piece_size);
   void OnInitDone(const StreamParser::InitParameters& params);
-  bool OnNewConfig(const AudioDecoderConfig& audio_config,
-                   const VideoDecoderConfig& video_config,
+  bool OnNewConfig(scoped_ptr<MediaTracks> tracks,
                    const StreamParser::TextTrackConfigMap& text_config);
   bool OnNewBuffers(const StreamParser::BufferQueue& audio_buffers,
                     const StreamParser::BufferQueue& video_buffers,
diff --git a/media/formats/mp2t/mp2t_stream_parser.cc b/media/formats/mp2t/mp2t_stream_parser.cc
index 12c8a82..90a2b96 100644
--- a/media/formats/mp2t/mp2t_stream_parser.cc
+++ b/media/formats/mp2t/mp2t_stream_parser.cc
@@ -9,6 +9,7 @@
 #include "base/bind.h"
 #include "base/callback_helpers.h"
 #include "base/stl_util.h"
+#include "media/base/media_tracks.h"
 #include "media/base/stream_parser_buffer.h"
 #include "media/base/text_track_config.h"
 #include "media/base/timestamp_constants.h"
@@ -502,6 +503,21 @@ void Mp2tStreamParser::OnAudioConfigChanged(
   }
 }
 
+scoped_ptr<MediaTracks> GenerateMediaTrackInfo(
+    const AudioDecoderConfig& audio_config,
+    const VideoDecoderConfig& video_config) {
+  scoped_ptr<MediaTracks> media_tracks(new MediaTracks());
+  // TODO(servolk): Implement proper sourcing of media track info as described
+  // in crbug.com/590085
+  if (audio_config.IsValidConfig()) {
+    media_tracks->AddAudioTrack(audio_config, "audio", "main", "", "");
+  }
+  if (video_config.IsValidConfig()) {
+    media_tracks->AddVideoTrack(video_config, "video", "main", "", "");
+  }
+  return media_tracks;
+}
+
 bool Mp2tStreamParser::FinishInitializationIfNeeded() {
   // Nothing to be done if already initialized.
   if (is_initialized_)
@@ -521,9 +537,9 @@ bool Mp2tStreamParser::FinishInitializationIfNeeded() {
     return true;
 
   // Pass the config before invoking the initialization callback.
-  RCHECK(config_cb_.Run(queue_with_config.audio_config,
-                        queue_with_config.video_config,
-                        TextTrackConfigMap()));
+  scoped_ptr<MediaTracks> media_tracks = GenerateMediaTrackInfo(
+      queue_with_config.audio_config, queue_with_config.video_config);
+  RCHECK(config_cb_.Run(std::move(media_tracks), TextTrackConfigMap()));
   queue_with_config.is_config_sent = true;
 
   // For Mpeg2 TS, the duration is not known.
@@ -620,9 +636,9 @@ bool Mp2tStreamParser::EmitRemainingBuffers() {
     // Update the audio and video config if needed.
     BufferQueueWithConfig& queue_with_config = buffer_queue_chain_.front();
     if (!queue_with_config.is_config_sent) {
-      if (!config_cb_.Run(queue_with_config.audio_config,
-                          queue_with_config.video_config,
-                          TextTrackConfigMap()))
+      scoped_ptr<MediaTracks> media_tracks = GenerateMediaTrackInfo(
+          queue_with_config.audio_config, queue_with_config.video_config);
+      if (!config_cb_.Run(std::move(media_tracks), TextTrackConfigMap()))
         return false;
       queue_with_config.is_config_sent = true;
     }
diff --git a/media/formats/mp2t/mp2t_stream_parser_unittest.cc b/media/formats/mp2t/mp2t_stream_parser_unittest.cc
index ab1d230..553f0c4 100644
--- a/media/formats/mp2t/mp2t_stream_parser_unittest.cc
+++ b/media/formats/mp2t/mp2t_stream_parser_unittest.cc
@@ -15,6 +15,8 @@
 #include "base/time/time.h"
 #include "media/base/audio_decoder_config.h"
 #include "media/base/decoder_buffer.h"
+#include "media/base/media_track.h"
+#include "media/base/media_tracks.h"
 #include "media/base/stream_parser_buffer.h"
 #include "media/base/test_data_util.h"
 #include "media/base/text_track_config.h"
@@ -111,10 +113,12 @@ class Mp2tStreamParserTest : public testing::Test {
              << ", autoTimestampOffset=" << params.auto_update_timestamp_offset;
   }
 
-  bool OnNewConfig(const AudioDecoderConfig& ac,
-                   const VideoDecoderConfig& vc,
+  bool OnNewConfig(scoped_ptr<MediaTracks> tracks,
                    const StreamParser::TextTrackConfigMap& tc) {
-    DVLOG(1) << "OnNewConfig: audio=" << ac.IsValidConfig()
+    const AudioDecoderConfig& ac = tracks->getFirstAudioConfig();
+    const VideoDecoderConfig& vc = tracks->getFirstVideoConfig();
+    DVLOG(1) << "OnNewConfig: media tracks count=" << tracks->tracks().size()
+             << ", audio=" << ac.IsValidConfig()
              << ", video=" << vc.IsValidConfig();
     config_count_++;
     EXPECT_TRUE(ac.IsValidConfig());
diff --git a/media/formats/mp4/mp4_stream_parser.cc b/media/formats/mp4/mp4_stream_parser.cc
index 7becdf9..95abf83 100644
--- a/media/formats/mp4/mp4_stream_parser.cc
+++ b/media/formats/mp4/mp4_stream_parser.cc
@@ -14,6 +14,7 @@
 #include "base/time/time.h"
 #include "build/build_config.h"
 #include "media/base/audio_decoder_config.h"
+#include "media/base/media_tracks.h"
 #include "media/base/stream_parser_buffer.h"
 #include "media/base/text_track_config.h"
 #include "media/base/timestamp_constants.h"
@@ -353,7 +354,16 @@ bool MP4StreamParser::ParseMoov(BoxReader* reader) {
   if (!moov_->pssh.empty())
     OnEncryptedMediaInitData(moov_->pssh);
 
-  RCHECK(config_cb_.Run(audio_config, video_config, TextTrackConfigMap()));
+  scoped_ptr<MediaTracks> media_tracks(new MediaTracks());
+  // TODO(servolk): Implement proper sourcing of media track info as described
+  // in crbug.com/590085
+  if (audio_config.IsValidConfig()) {
+    media_tracks->AddAudioTrack(audio_config, "audio", "", "", "");
+  }
+  if (video_config.IsValidConfig()) {
+    media_tracks->AddVideoTrack(video_config, "video", "", "", "");
+  }
+  RCHECK(config_cb_.Run(std::move(media_tracks), TextTrackConfigMap()));
 
   StreamParser::InitParameters params(kInfiniteDuration());
   if (moov_->extends.header.fragment_duration > 0) {
diff --git a/media/formats/mp4/mp4_stream_parser_unittest.cc b/media/formats/mp4/mp4_stream_parser_unittest.cc
index 635acd1..1fd2b75 100644
--- a/media/formats/mp4/mp4_stream_parser_unittest.cc
+++ b/media/formats/mp4/mp4_stream_parser_unittest.cc
@@ -15,6 +15,8 @@
 #include "base/time/time.h"
 #include "media/base/audio_decoder_config.h"
 #include "media/base/decoder_buffer.h"
+#include "media/base/media_track.h"
+#include "media/base/media_tracks.h"
 #include "media/base/mock_media_log.h"
 #include "media/base/stream_parser_buffer.h"
 #include "media/base/test_data_util.h"
@@ -97,14 +99,14 @@ class MP4StreamParserTest : public testing::Test {
     EXPECT_EQ(expected_liveness, params.liveness);
   }
 
-  bool NewConfigF(const AudioDecoderConfig& ac,
-                  const VideoDecoderConfig& vc,
+  bool NewConfigF(scoped_ptr<MediaTracks> tracks,
                   const StreamParser::TextTrackConfigMap& tc) {
-    DVLOG(1) << "NewConfigF: audio=" << ac.IsValidConfig()
-             << ", video=" << vc.IsValidConfig();
     configs_received_ = true;
-    audio_decoder_config_ = ac;
-    video_decoder_config_ = vc;
+    audio_decoder_config_ = tracks->getFirstAudioConfig();
+    video_decoder_config_ = tracks->getFirstVideoConfig();
+    DVLOG(1) << "NewConfigF: track count=" << tracks->tracks().size()
+             << " audio=" << audio_decoder_config_.IsValidConfig()
+             << " video=" << video_decoder_config_.IsValidConfig();
     return true;
   }
 
diff --git a/media/formats/mpeg/mpeg_audio_stream_parser_base.cc b/media/formats/mpeg/mpeg_audio_stream_parser_base.cc
index ea2261f..2028495 100644
--- a/media/formats/mpeg/mpeg_audio_stream_parser_base.cc
+++ b/media/formats/mpeg/mpeg_audio_stream_parser_base.cc
@@ -7,6 +7,7 @@
 #include "base/bind.h"
 #include "base/callback_helpers.h"
 #include "base/message_loop/message_loop.h"
+#include "media/base/media_tracks.h"
 #include "media/base/stream_parser_buffer.h"
 #include "media/base/text_track_config.h"
 #include "media/base/timestamp_constants.h"
@@ -220,8 +221,11 @@ int MPEGAudioStreamParserBase::ParseFrame(const uint8_t* data,
     timestamp_helper_.reset(new AudioTimestampHelper(sample_rate));
     timestamp_helper_->SetBaseTimestamp(base_timestamp);
 
-    VideoDecoderConfig video_config;
-    if (!config_cb_.Run(config_, video_config, TextTrackConfigMap()))
+    scoped_ptr<MediaTracks> media_tracks(new MediaTracks());
+    if (config_.IsValidConfig()) {
+      media_tracks->AddAudioTrack(config_, "audio", "", "", "");
+    }
+    if (!config_cb_.Run(std::move(media_tracks), TextTrackConfigMap()))
       return -1;
 
     if (!init_cb_.is_null()) {
diff --git a/media/formats/webm/webm_stream_parser.cc b/media/formats/webm/webm_stream_parser.cc
index e36a972..3090768 100644
--- a/media/formats/webm/webm_stream_parser.cc
+++ b/media/formats/webm/webm_stream_parser.cc
@@ -9,6 +9,8 @@
 #include "base/callback.h"
 #include "base/callback_helpers.h"
 #include "base/logging.h"
+#include "media/base/media_track.h"
+#include "media/base/media_tracks.h"
 #include "media/base/timestamp_constants.h"
 #include "media/formats/webm/webm_cluster_parser.h"
 #include "media/formats/webm/webm_constants.h"
@@ -222,9 +224,16 @@ int WebMStreamParser::ParseInfoAndTracks(const uint8_t* data, int size) {
   if (video_config.is_encrypted())
     OnEncryptedMediaInitData(tracks_parser.video_encryption_key_id());
 
-  if (!config_cb_.Run(audio_config,
-                      video_config,
-                      tracks_parser.text_tracks())) {
+  scoped_ptr<MediaTracks> media_tracks(new MediaTracks());
+  // TODO(servolk): Implement proper sourcing of media track info as described
+  // in crbug.com/590085
+  if (audio_config.IsValidConfig()) {
+    media_tracks->AddAudioTrack(audio_config, "audio", "", "", "");
+  }
+  if (video_config.IsValidConfig()) {
+    media_tracks->AddVideoTrack(video_config, "video", "", "", "");
+  }
+  if (!config_cb_.Run(std::move(media_tracks), tracks_parser.text_tracks())) {
     DVLOG(1) << "New config data isn't allowed.";
     return -1;
   }
diff --git a/media/media.gyp b/media/media.gyp
index 825dc85..5fe1f159 100644
--- a/media/media.gyp
+++ b/media/media.gyp
@@ -363,6 +363,10 @@
         'base/media_resources.h',
         'base/media_switches.cc',
         'base/media_switches.h',
+        'base/media_track.cc',
+        'base/media_track.h',
+        'base/media_tracks.cc',
+        'base/media_tracks.h',
         'base/media_util.cc',
         'base/media_util.h',
         'base/mime_util.cc',
diff --git a/media/test/pipeline_integration_test.cc b/media/test/pipeline_integration_test.cc
index 80871f9..51a845b 100644
--- a/media/test/pipeline_integration_test.cc
+++ b/media/test/pipeline_integration_test.cc
@@ -21,6 +21,7 @@
 #include "media/base/media.h"
 #include "media/base/media_keys.h"
 #include "media/base/media_switches.h"
+#include "media/base/media_tracks.h"
 #include "media/base/test_data_util.h"
 #include "media/base/timestamp_constants.h"
 #include "media/cdm/aes_decryptor.h"
@@ -636,7 +637,7 @@ class MockMediaSource {
     return last_timestamp_offset_;
   }
 
-  MOCK_METHOD0(InitSegmentReceived, void(void));
+  MOCK_METHOD1(InitSegmentReceived, void(const MediaTracks&));
 
  private:
   scoped_refptr<DecoderBuffer> file_data_;
@@ -682,7 +683,7 @@ class PipelineIntegrationTestHost : public testing::Test,
 class PipelineIntegrationTest : public PipelineIntegrationTestHost {
  public:
   void StartPipelineWithMediaSource(MockMediaSource* source) {
-    EXPECT_CALL(*source, InitSegmentReceived()).Times(AtLeast(1));
+    EXPECT_CALL(*source, InitSegmentReceived(_)).Times(AtLeast(1));
     EXPECT_CALL(*this, OnMetadata(_))
         .Times(AtMost(1))
         .WillRepeatedly(SaveArg<0>(&metadata_));
@@ -726,7 +727,7 @@ class PipelineIntegrationTest : public PipelineIntegrationTestHost {
 
   void StartPipelineWithEncryptedMedia(MockMediaSource* source,
                                        FakeEncryptedMedia* encrypted_media) {
-    EXPECT_CALL(*source, InitSegmentReceived()).Times(AtLeast(1));
+    EXPECT_CALL(*source, InitSegmentReceived(_)).Times(AtLeast(1));
     EXPECT_CALL(*this, OnMetadata(_))
         .Times(AtMost(1))
         .WillRepeatedly(SaveArg<0>(&metadata_));