Adding VP8 Alpha support in Media Source

Adding support for playback of VP8 videos with Alpha Channel through the
Media Source API.

BUG=242357
TEST=VP8 Alpha Streams play properly via Media Source

Review URL: https://chromiumcodereview.appspot.com/15342004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@201496 0039d316-1c4b-4281-b951-d872f2087c98

11 files changed, 133 insertions, 54 deletions

diff --git a/media/base/decoder_buffer.cc b/media/base/decoder_buffer.cc
index aec4521..d3acc26 100644
--- a/media/base/decoder_buffer.cc
+++ b/media/base/decoder_buffer.cc
@@ -15,30 +15,20 @@ DecoderBuffer::DecoderBuffer(int size)
   Initialize();
 }
 
-DecoderBuffer::DecoderBuffer(const uint8* data, int size)
-    : size_(size),
-      side_data_size_(0) {
-  if (!data) {
-    CHECK_EQ(size_, 0);
-    return;
-  }
-
-  Initialize();
-  memcpy(data_.get(), data, size_);
-}
-
 DecoderBuffer::DecoderBuffer(const uint8* data, int size,
                              const uint8* side_data, int side_data_size)
     : size_(size),
       side_data_size_(side_data_size) {
   if (!data) {
     CHECK_EQ(size_, 0);
+    CHECK(!side_data);
     return;
   }
 
   Initialize();
   memcpy(data_.get(), data, size_);
-  memcpy(side_data_.get(), side_data, side_data_size_);
+  if (side_data)
+    memcpy(side_data_.get(), side_data, side_data_size_);
 }
 
 DecoderBuffer::~DecoderBuffer() {}
@@ -60,7 +50,7 @@ scoped_refptr<DecoderBuffer> DecoderBuffer::CopyFrom(const uint8* data,
                                                      int data_size) {
   // If you hit this CHECK you likely have a bug in a demuxer. Go fix it.
   CHECK(data);
-  return make_scoped_refptr(new DecoderBuffer(data, data_size));
+  return make_scoped_refptr(new DecoderBuffer(data, data_size, NULL, 0));
 }
 
 // static
@@ -77,7 +67,7 @@ scoped_refptr<DecoderBuffer> DecoderBuffer::CopyFrom(const uint8* data,
 
 // static
 scoped_refptr<DecoderBuffer> DecoderBuffer::CreateEOSBuffer() {
-  return make_scoped_refptr(new DecoderBuffer(NULL, 0));
+  return make_scoped_refptr(new DecoderBuffer(NULL, 0, NULL, 0));
 }
 
 base::TimeDelta DecoderBuffer::GetTimestamp() const {
diff --git a/media/base/decoder_buffer.h b/media/base/decoder_buffer.h
index 168ab2c..d1c20dd 100644
--- a/media/base/decoder_buffer.h
+++ b/media/base/decoder_buffer.h
@@ -89,7 +89,6 @@ class MEDIA_EXPORT DecoderBuffer
   // Allocates a buffer of size |size| >= 0 and copies |data| into it.  Buffer
   // will be padded and aligned as necessary.  If |data| is NULL then |data_| is
   // set to NULL and |buffer_size_| to 0.
-  DecoderBuffer(const uint8* data, int size);
   DecoderBuffer(const uint8* data, int size,
                 const uint8* side_data, int side_data_size);
   virtual ~DecoderBuffer();
diff --git a/media/base/stream_parser_buffer.cc b/media/base/stream_parser_buffer.cc
index 547cf27..f1dd6fd 100644
--- a/media/base/stream_parser_buffer.cc
+++ b/media/base/stream_parser_buffer.cc
@@ -10,13 +10,21 @@
 namespace media {
 
 scoped_refptr<StreamParserBuffer> StreamParserBuffer::CreateEOSBuffer() {
-  return make_scoped_refptr(new StreamParserBuffer(NULL, 0, false));
+  return make_scoped_refptr(new StreamParserBuffer(NULL, 0, NULL, 0, false));
 }
 
 scoped_refptr<StreamParserBuffer> StreamParserBuffer::CopyFrom(
     const uint8* data, int data_size, bool is_keyframe) {
   return make_scoped_refptr(
-      new StreamParserBuffer(data, data_size, is_keyframe));
+      new StreamParserBuffer(data, data_size, NULL, 0, is_keyframe));
+}
+
+scoped_refptr<StreamParserBuffer> StreamParserBuffer::CopyFrom(
+    const uint8* data, int data_size,
+    const uint8* side_data, int side_data_size, bool is_keyframe) {
+  return make_scoped_refptr(
+      new StreamParserBuffer(data, data_size, side_data, side_data_size,
+                             is_keyframe));
 }
 
 base::TimeDelta StreamParserBuffer::GetDecodeTimestamp() const {
@@ -30,8 +38,9 @@ void StreamParserBuffer::SetDecodeTimestamp(const base::TimeDelta& timestamp) {
 }
 
 StreamParserBuffer::StreamParserBuffer(const uint8* data, int data_size,
-                                       bool is_keyframe)
-    : DecoderBuffer(data, data_size),
+                                       const uint8* side_data,
+                                       int side_data_size, bool is_keyframe)
+    : DecoderBuffer(data, data_size, side_data, side_data_size),
       is_keyframe_(is_keyframe),
       decode_timestamp_(kNoTimestamp()),
       config_id_(kInvalidConfigId) {
diff --git a/media/base/stream_parser_buffer.h b/media/base/stream_parser_buffer.h
index 90e4ca9..8899f11 100644
--- a/media/base/stream_parser_buffer.h
+++ b/media/base/stream_parser_buffer.h
@@ -18,6 +18,9 @@ class MEDIA_EXPORT StreamParserBuffer : public DecoderBuffer {
   static scoped_refptr<StreamParserBuffer> CreateEOSBuffer();
   static scoped_refptr<StreamParserBuffer> CopyFrom(
       const uint8* data, int data_size, bool is_keyframe);
+  static scoped_refptr<StreamParserBuffer> CopyFrom(
+      const uint8* data, int data_size,
+      const uint8* side_data, int side_data_size, bool is_keyframe);
   bool IsKeyframe() const { return is_keyframe_; }
 
   // Decode timestamp. If not explicitly set, or set to kNoTimestamp(), the
@@ -31,7 +34,9 @@ class MEDIA_EXPORT StreamParserBuffer : public DecoderBuffer {
   void SetConfigId(int config_id);
 
  private:
-  StreamParserBuffer(const uint8* data, int data_size, bool is_keyframe);
+  StreamParserBuffer(const uint8* data, int data_size,
+                     const uint8* side_data, int side_data_size,
+                     bool is_keyframe);
   virtual ~StreamParserBuffer();
 
   bool is_keyframe_;
diff --git a/media/filters/pipeline_integration_test.cc b/media/filters/pipeline_integration_test.cc
index 63a6892..ccbdd6d 100644
--- a/media/filters/pipeline_integration_test.cc
+++ b/media/filters/pipeline_integration_test.cc
@@ -54,6 +54,7 @@ static const int k640IsoFileDurationMs = 2737;
 static const int k640IsoCencFileDurationMs = 2736;
 static const int k1280IsoFileDurationMs = 2736;
 static const int kVP9WebMFileDurationMs = 2736;
+static const int kVP8AWebMFileDurationMs = 2700;
 
 // Note: Tests using this class only exercise the DecryptingDemuxerStream path.
 // They do not exercise the Decrypting{Audio|Video}Decoder path.
@@ -479,6 +480,23 @@ TEST_F(PipelineIntegrationTest, BasicPlayback_MediaSource_VideoOnly_VP9_WebM) {
   Stop();
 }
 
+TEST_F(PipelineIntegrationTest, BasicPlayback_MediaSource_VP8A_WebM) {
+  MockMediaSource source("bear-vp8a.webm", kVideoOnlyWebM, kAppendWholeFile);
+  StartPipelineWithMediaSource(&source);
+  source.EndOfStream();
+
+  EXPECT_EQ(1u, pipeline_->GetBufferedTimeRanges().size());
+  EXPECT_EQ(0, pipeline_->GetBufferedTimeRanges().start(0).InMilliseconds());
+  EXPECT_EQ(kVP8AWebMFileDurationMs,
+            pipeline_->GetBufferedTimeRanges().end(0).InMilliseconds());
+
+  Play();
+
+  ASSERT_TRUE(WaitUntilOnEnded());
+  source.Abort();
+  Stop();
+}
+
 TEST_F(PipelineIntegrationTest, MediaSource_ConfigChange_WebM) {
   MockMediaSource source("bear-320x240-16x9-aspect.webm", kWebM,
                          kAppendWholeFile);
diff --git a/media/webm/webm_cluster_parser.cc b/media/webm/webm_cluster_parser.cc
index ec89b67..41e193f 100644
--- a/media/webm/webm_cluster_parser.cc
+++ b/media/webm/webm_cluster_parser.cc
@@ -7,6 +7,7 @@
 #include <vector>
 
 #include "base/logging.h"
+#include "base/sys_byteorder.h"
 #include "media/base/buffers.h"
 #include "media/base/decrypt_config.h"
 #include "media/webm/webm_constants.h"
@@ -61,6 +62,8 @@ WebMClusterParser::WebMClusterParser(
       last_block_timecode_(-1),
       block_data_size_(-1),
       block_duration_(-1),
+      block_add_id_(-1),
+      block_additional_data_size_(-1),
       cluster_timecode_(-1),
       cluster_start_time_(kNoTimestamp()),
       cluster_ended_(false),
@@ -134,6 +137,10 @@ WebMParserClient* WebMClusterParser::OnListStart(int id) {
     block_data_.reset();
     block_data_size_ = -1;
     block_duration_ = -1;
+  } else if (id == kWebMIdBlockAdditions) {
+    block_add_id_ = -1;
+    block_additional_data_.reset();
+    block_additional_data_size_ = -1;
   }
 
   return this;
@@ -150,30 +157,41 @@ bool WebMClusterParser::OnListEnd(int id) {
   }
 
   bool result = ParseBlock(false, block_data_.get(), block_data_size_,
-                           block_duration_);
+                           block_additional_data_.get(),
+                           block_additional_data_size_, block_duration_);
   block_data_.reset();
   block_data_size_ = -1;
   block_duration_ = -1;
+  block_add_id_ = -1;
+  block_additional_data_.reset();
+  block_additional_data_size_ = -1;
   return result;
 }
 
 bool WebMClusterParser::OnUInt(int id, int64 val) {
-  if (id == kWebMIdTimecode) {
-    if (cluster_timecode_ != -1)
-      return false;
-
-    cluster_timecode_ = val;
-  } else if (id == kWebMIdBlockDuration) {
-    if (block_duration_ != -1)
-      return false;
-    block_duration_ = val;
+  int64* dst;
+  switch (id) {
+    case kWebMIdTimecode:
+      dst = &cluster_timecode_;
+      break;
+    case kWebMIdBlockDuration:
+      dst = &block_duration_;
+      break;
+    case kWebMIdBlockAddID:
+      dst = &block_add_id_;
+      break;
+    default:
+      return true;
   }
-
+  if (*dst != -1)
+    return false;
+  *dst = val;
   return true;
 }
 
 bool WebMClusterParser::ParseBlock(bool is_simple_block, const uint8* buf,
-                                   int size, int duration) {
+                                   int size, const uint8* additional,
+                                   int additional_size, int duration) {
   if (size < 4)
     return false;
 
@@ -201,32 +219,58 @@ bool WebMClusterParser::ParseBlock(bool is_simple_block, const uint8* buf,
   const uint8* frame_data = buf + 4;
   int frame_size = size - (frame_data - buf);
   return OnBlock(is_simple_block, track_num, timecode, duration, flags,
-                 frame_data, frame_size);
+                 frame_data, frame_size, additional, additional_size);
 }
 
 bool WebMClusterParser::OnBinary(int id, const uint8* data, int size) {
-  if (id == kWebMIdSimpleBlock)
-    return ParseBlock(true, data, size, -1);
-
-  if (id != kWebMIdBlock)
-    return true;
+  switch (id) {
+    case kWebMIdSimpleBlock:
+      return ParseBlock(true, data, size, NULL, -1, -1);
+
+    case kWebMIdBlock:
+      if (block_data_) {
+        MEDIA_LOG(log_cb_) << "More than 1 Block in a BlockGroup is not "
+                              "supported.";
+        return false;
+      }
+      block_data_.reset(new uint8[size]);
+      memcpy(block_data_.get(), data, size);
+      block_data_size_ = size;
+      return true;
+
+    case kWebMIdBlockAdditional: {
+      uint64 block_add_id = base::HostToNet64(block_add_id_);
+      if (block_additional_data_) {
+        // TODO(vigneshv): Technically, more than 1 BlockAdditional is allowed
+        // as per matroska spec. But for now we don't have a use case to
+        // support parsing of such files. Take a look at this again when such a
+        // case arises.
+        MEDIA_LOG(log_cb_) << "More than 1 BlockAdditional in a BlockGroup is "
+                              "not supported.";
+        return false;
+      }
+      // First 8 bytes of side_data in DecoderBuffer is the BlockAddID
+      // element's value in Big Endian format. This is done to mimic ffmpeg
+      // demuxer's behavior.
+      block_additional_data_size_ = size + sizeof(block_add_id);
+      block_additional_data_.reset(new uint8[block_additional_data_size_]);
+      memcpy(block_additional_data_.get(), &block_add_id,
+             sizeof(block_add_id));
+      memcpy(block_additional_data_.get() + 8, data, size);
+      return true;
+    }
 
-  if (block_data_) {
-    MEDIA_LOG(log_cb_) << "More than 1 Block in a BlockGroup is not supported.";
-    return false;
+    default:
+      return true;
   }
-
-  block_data_.reset(new uint8[size]);
-  memcpy(block_data_.get(), data, size);
-  block_data_size_ = size;
-  return true;
 }
 
 bool WebMClusterParser::OnBlock(bool is_simple_block, int track_num,
                                 int timecode,
                                 int  block_duration,
                                 int flags,
-                                const uint8* data, int size) {
+                                const uint8* data, int size,
+                                const uint8* additional, int additional_size) {
   DCHECK_GE(size, 0);
   if (cluster_timecode_ == -1) {
     MEDIA_LOG(log_cb_) << "Got a block before cluster timecode.";
@@ -279,7 +323,8 @@ bool WebMClusterParser::OnBlock(bool is_simple_block, int track_num,
       is_simple_block ? (flags & 0x80) != 0 : track->IsKeyframe(data, size);
 
   scoped_refptr<StreamParserBuffer> buffer =
-      StreamParserBuffer::CopyFrom(data, size, is_keyframe);
+      StreamParserBuffer::CopyFrom(data, size, additional, additional_size,
+                                   is_keyframe);
 
   // Every encrypted Block has a signal byte and IV prepended to it. Current
   // encrypted WebM request for comments specification is here
diff --git a/media/webm/webm_cluster_parser.h b/media/webm/webm_cluster_parser.h
index 63d8626..e156d47 100644
--- a/media/webm/webm_cluster_parser.h
+++ b/media/webm/webm_cluster_parser.h
@@ -110,9 +110,10 @@ class MEDIA_EXPORT WebMClusterParser : public WebMParserClient {
   virtual bool OnBinary(int id, const uint8* data, int size) OVERRIDE;
 
   bool ParseBlock(bool is_simple_block, const uint8* buf, int size,
-                  int duration);
+                  const uint8* additional, int additional_size, int duration);
   bool OnBlock(bool is_simple_block, int track_num, int timecode, int duration,
-               int flags, const uint8* data, int size);
+               int flags, const uint8* data, int size,
+               const uint8* additional, int additional_size);
 
   // Resets the Track objects associated with each text track.
   void ResetTextTracks();
@@ -133,6 +134,9 @@ class MEDIA_EXPORT WebMClusterParser : public WebMParserClient {
   scoped_ptr<uint8[]> block_data_;
   int block_data_size_;
   int64 block_duration_;
+  int64 block_add_id_;
+  scoped_ptr<uint8[]> block_additional_data_;
+  int block_additional_data_size_;
 
   int64 cluster_timecode_;
   base::TimeDelta cluster_start_time_;
diff --git a/media/webm/webm_constants.h b/media/webm/webm_constants.h
index 5cfd5e6..cda45e0 100644
--- a/media/webm/webm_constants.h
+++ b/media/webm/webm_constants.h
@@ -14,6 +14,7 @@ namespace media {
 // This is a subset of the IDs in the Matroska spec.
 // http://www.matroska.org/technical/specs/index.html
 const int kWebMIdAESSettingsCipherMode = 0x47E8;
+const int kWebMIdAlphaMode = 0x53C0;
 const int kWebMIdAspectRatioType = 0x54B3;
 const int kWebMIdAttachedFile = 0x61A7;
 const int kWebMIdAttachmentLink = 0x7446;
diff --git a/media/webm/webm_parser.cc b/media/webm/webm_parser.cc
index fb9f899..30e5c1b5 100644
--- a/media/webm/webm_parser.cc
+++ b/media/webm/webm_parser.cc
@@ -179,6 +179,7 @@ static const ElementIdInfo kTrackTranslateIds[] = {
 static const ElementIdInfo kVideoIds[] = {
   {UINT, kWebMIdFlagInterlaced},
   {UINT, kWebMIdStereoMode},
+  {UINT, kWebMIdAlphaMode},
   {UINT, kWebMIdPixelWidth},
   {UINT, kWebMIdPixelHeight},
   {UINT, kWebMIdPixelCropBottom},
diff --git a/media/webm/webm_video_client.cc b/media/webm/webm_video_client.cc
index b4e729e..1d0cbcb 100644
--- a/media/webm/webm_video_client.cc
+++ b/media/webm/webm_video_client.cc
@@ -27,6 +27,7 @@ void WebMVideoClient::Reset() {
   display_width_ = -1;
   display_height_ = -1;
   display_unit_ = -1;
+  alpha_mode_ = -1;
 }
 
 bool WebMVideoClient::InitializeConfig(
@@ -47,6 +48,9 @@ bool WebMVideoClient::InitializeConfig(
     return false;
   }
 
+  VideoFrame::Format format =
+      (alpha_mode_ == 1) ? VideoFrame::YV12A : VideoFrame::YV12;
+
   if (pixel_width_ <= 0 || pixel_height_ <= 0)
     return false;
 
@@ -93,9 +97,8 @@ bool WebMVideoClient::InitializeConfig(
   }
 
   config->Initialize(
-      video_codec, profile, VideoFrame::YV12, coded_size,
-      visible_rect, natural_size, extra_data, extra_data_size,
-      is_encrypted, true);
+      video_codec, profile, format, coded_size, visible_rect, natural_size,
+      extra_data, extra_data_size, is_encrypted, true);
   return config->IsValidConfig();
 }
 
@@ -130,6 +133,9 @@ bool WebMVideoClient::OnUInt(int id, int64 val) {
     case kWebMIdDisplayUnit:
       dst = &display_unit_;
       break;
+    case kWebMIdAlphaMode:
+      dst = &alpha_mode_;
+      break;
     default:
       return true;
   }
diff --git a/media/webm/webm_video_client.h b/media/webm/webm_video_client.h
index eb6be6d..d1872ba 100644
--- a/media/webm/webm_video_client.h
+++ b/media/webm/webm_video_client.h
@@ -51,6 +51,7 @@ class WebMVideoClient : public WebMParserClient {
   int64 display_width_;
   int64 display_height_;
   int64 display_unit_;
+  int64 alpha_mode_;
 
   DISALLOW_COPY_AND_ASSIGN(WebMVideoClient);
 };