VPCodecConfiguration box should inherit from FullBox instead of Box

Change-Id: If53c6cab009bdeadc37a1bac34272056a2b7d1dc

packager/app/test/testdata/bear-320x240-vp9-cenc-golden.mpd

@@ -3,14 +3,14 @@
 <MPD xmlns="urn:mpeg:dash:schema:mpd:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="urn:mpeg:dash:schema:mpd:2011 DASH-MPD.xsd" xmlns:cenc="urn:mpeg:cenc:2013" minBufferTime="PT2S" type="static" profiles="urn:mpeg:dash:profile:isoff-on-demand:2011" mediaPresentationDuration="PT2.7360000610351562S">
   <Period id="0">
     <AdaptationSet id="0" contentType="video" width="320" height="240" frameRate="1000000/34000" subsegmentAlignment="true" par="16:9">
-      <Representation id="0" bandwidth="210135" codecs="vp09.00.00.08.00.01.00.00" mimeType="video/mp4" sar="427:320">
+      <Representation id="0" bandwidth="210158" codecs="vp09.00.00.08.00.01.00.00" mimeType="video/mp4" sar="427:320">
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="31323334-3536-3738-3930-313233343536"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
           <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
         </ContentProtection>
         <BaseURL>output_video.mp4</BaseURL>
-        <SegmentBase indexRange="1039-1106" timescale="1000000">
+        <SegmentBase indexRange="1047-1114" timescale="1000000">
-          <Initialization range="0-1038"/>
+          <Initialization range="0-1046"/>
         </SegmentBase>
       </Representation>
     </AdaptationSet>

packager/app/test/testdata/bear-640x360-vp8-cenc-golden.mpd

@@ -3,14 +3,14 @@
 <MPD xmlns="urn:mpeg:dash:schema:mpd:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="urn:mpeg:dash:schema:mpd:2011 DASH-MPD.xsd" xmlns:cenc="urn:mpeg:cenc:2013" minBufferTime="PT2S" type="static" profiles="urn:mpeg:dash:profile:isoff-on-demand:2011" mediaPresentationDuration="PT2.7360000610351562S">
   <Period id="0">
     <AdaptationSet id="0" contentType="video" width="640" height="360" frameRate="1000000/33000" subsegmentAlignment="true" par="16:9">
-      <Representation id="0" bandwidth="342164" codecs="vp08.00.00.08.01.01.00.00" mimeType="video/mp4" sar="1:1">
+      <Representation id="0" bandwidth="342188" codecs="vp08.00.00.08.01.01.00.00" mimeType="video/mp4" sar="1:1">
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="31323334-3536-3738-3930-313233343536"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
           <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
         </ContentProtection>
         <BaseURL>output_video.mp4</BaseURL>
-        <SegmentBase indexRange="1007-1074" timescale="1000000">
+        <SegmentBase indexRange="1015-1082" timescale="1000000">
-          <Initialization range="0-1006"/>
+          <Initialization range="0-1014"/>
         </SegmentBase>
       </Representation>
     </AdaptationSet>

packager/media/formats/mp4/box_definitions.cc

@@ -1186,16 +1186,27 @@ uint32_t Metadata::ComputeSizeInternal() {
                          : HeaderSize() + handler.ComputeSize() + id3v2_size;
 }
 
-CodecConfigurationRecord::CodecConfigurationRecord() : box_type(FOURCC_NULL) {}
+CodecConfiguration::CodecConfiguration() : box_type(FOURCC_NULL) {}
-CodecConfigurationRecord::~CodecConfigurationRecord() {}
+CodecConfiguration::~CodecConfiguration() {}
-FourCC CodecConfigurationRecord::BoxType() const {
+
-  // CodecConfigurationRecord should be parsed according to format recovered in
+FourCC CodecConfiguration::BoxType() const {
+  // CodecConfiguration box should be parsed according to format recovered in
   // VideoSampleEntry. |box_type| is determined dynamically there.
   return box_type;
 }
 
-bool CodecConfigurationRecord::ReadWriteInternal(BoxBuffer* buffer) {
+bool CodecConfiguration::ReadWriteInternal(BoxBuffer* buffer) {
+  DCHECK_NE(box_type, FOURCC_NULL);
   RCHECK(ReadWriteHeaderInternal(buffer));
+
+  // VPCodecConfiguration box inherits from FullBox instead of Box. The extra 4
+  // bytes are handled here.
+  if (box_type == FOURCC_VPCC) {
+    uint32_t version_flags = 0;
+    RCHECK(buffer->ReadWriteUInt32(&version_flags));
+    RCHECK(version_flags == 0);
+  }
+
   if (buffer->Reading()) {
     RCHECK(buffer->ReadWriteVector(&data, buffer->BytesLeft()));
   } else {
@@ -1204,10 +1215,11 @@ bool CodecConfigurationRecord::ReadWriteInternal(BoxBuffer* buffer) {
   return true;
 }
 
-uint32_t CodecConfigurationRecord::ComputeSizeInternal() {
+uint32_t CodecConfiguration::ComputeSizeInternal() {
   if (data.empty())
     return 0;
-  return HeaderSize() + data.size();
+  DCHECK_NE(box_type, FOURCC_NULL);
+  return HeaderSize() + (box_type == FOURCC_VPCC ? 4 : 0) + data.size();
 }
 
 PixelAspectRatio::PixelAspectRatio() : h_spacing(0), v_spacing(0) {}
@@ -1310,37 +1322,50 @@ bool VideoSampleEntry::ReadWriteInternal(BoxBuffer* buffer) {
   }
 
   const FourCC actual_format = GetActualFormat();
-  switch (actual_format) {
+  if (buffer->Reading()) {
-    case FOURCC_AVC1:
+    codec_configuration.box_type = GetCodecConfigurationBoxType(actual_format);
-      codec_config_record.box_type = FOURCC_AVCC;
+  } else {
-      break;
+    DCHECK_EQ(codec_configuration.box_type,
-    case FOURCC_HEV1:
+              GetCodecConfigurationBoxType(actual_format));
-    case FOURCC_HVC1:
-      codec_config_record.box_type = FOURCC_HVCC;
-      break;
-    case FOURCC_VP08:
-    case FOURCC_VP09:
-    case FOURCC_VP10:
-      codec_config_record.box_type = FOURCC_VPCC;
-      break;
-    default:
-      LOG(ERROR) << FourCCToString(actual_format) << " is not supported.";
-      return false;
   }
-  RCHECK(buffer->ReadWriteChild(&codec_config_record));
+  DCHECK_NE(codec_configuration.box_type, FOURCC_NULL);
+
+  RCHECK(buffer->ReadWriteChild(&codec_configuration));
   RCHECK(buffer->TryReadWriteChild(&pixel_aspect));
   return true;
 }
 
 uint32_t VideoSampleEntry::ComputeSizeInternal() {
+  const FourCC actual_format = GetActualFormat();
+  if (actual_format == FOURCC_NULL)
+    return 0;
+  codec_configuration.box_type = GetCodecConfigurationBoxType(actual_format);
+  DCHECK_NE(codec_configuration.box_type, FOURCC_NULL);
   return HeaderSize() + sizeof(data_reference_index) + sizeof(width) +
          sizeof(height) + sizeof(kVideoResolution) * 2 +
          sizeof(kVideoFrameCount) + sizeof(kVideoDepth) +
          pixel_aspect.ComputeSize() + sinf.ComputeSize() +
-         codec_config_record.ComputeSize() + kCompressorNameSize + 6 + 4 + 16 +
+         codec_configuration.ComputeSize() + kCompressorNameSize + 6 + 4 + 16 +
          2;  // 6 + 4 bytes reserved, 16 + 2 bytes predefined.
 }
 
+FourCC VideoSampleEntry::GetCodecConfigurationBoxType(FourCC format) const {
+  switch (format) {
+    case FOURCC_AVC1:
+      return FOURCC_AVCC;
+    case FOURCC_HEV1:
+    case FOURCC_HVC1:
+      return FOURCC_HVCC;
+    case FOURCC_VP08:
+    case FOURCC_VP09:
+    case FOURCC_VP10:
+      return FOURCC_VPCC;
+    default:
+      LOG(ERROR) << FourCCToString(format) << " is not supported.";
+      return FOURCC_NULL;
+  }
+}
+
 ElementaryStreamDescriptor::ElementaryStreamDescriptor() {}
 ElementaryStreamDescriptor::~ElementaryStreamDescriptor() {}
 FourCC ElementaryStreamDescriptor::BoxType() const { return FOURCC_ESDS; }
@@ -1502,6 +1527,8 @@ bool AudioSampleEntry::ReadWriteInternal(BoxBuffer* buffer) {
 }
 
 uint32_t AudioSampleEntry::ComputeSizeInternal() {
+  if (GetActualFormat() == FOURCC_NULL)
+    return 0;
   return HeaderSize() + sizeof(data_reference_index) + sizeof(channelcount) +
          sizeof(samplesize) + sizeof(samplerate) + sinf.ComputeSize() +
          esds.ComputeSize() + ddts.ComputeSize() + dac3.ComputeSize() +

packager/media/formats/mp4/box_definitions.h

@@ -257,8 +257,14 @@ struct Metadata : FullBox {
   ID3v2 id3v2;
 };
 
-struct CodecConfigurationRecord : Box {
+// This defines a common structure for various CodecConfiguration boxes:
-  DECLARE_BOX_METHODS(CodecConfigurationRecord);
+// AVCConfiguration, HEVCConfiguration and VPCodecConfiguration.
+// Note that unlike the other two CodecConfiguration boxes, VPCodecConfiguration
+// box inherits from FullBox instead of Box, according to VP Codec ISO Media
+// File Format Binding specification. It will be handled properly in the
+// implementation.
+struct CodecConfiguration: Box {
+  DECLARE_BOX_METHODS(CodecConfiguration);
 
   FourCC box_type;
   // Contains full codec configuration record, including possible extension
@@ -279,6 +285,8 @@ struct VideoSampleEntry : Box {
   FourCC GetActualFormat() const {
     return format == FOURCC_ENCV ? sinf.format.format : format;
   }
+  // Returns the box type of codec configuration box from video format.
+  FourCC GetCodecConfigurationBoxType(FourCC format) const;
 
   FourCC format;
   uint16_t data_reference_index;
@@ -287,7 +295,7 @@ struct VideoSampleEntry : Box {
 
   PixelAspectRatio pixel_aspect;
   ProtectionSchemeInfo sinf;
-  CodecConfigurationRecord codec_config_record;
+  CodecConfiguration codec_configuration;
 };
 
 struct ElementaryStreamDescriptor : FullBox {

packager/media/formats/mp4/box_definitions_comparison.h

@@ -200,9 +200,9 @@ inline bool operator==(const Metadata& lhs, const Metadata& rhs) {
   return lhs.handler == rhs.handler && lhs.id3v2 == rhs.id3v2;
 }
 
-inline bool operator==(const CodecConfigurationRecord& lhs,
+inline bool operator==(const CodecConfiguration& lhs,
-                       const CodecConfigurationRecord& rhs) {
+                       const CodecConfiguration& rhs) {
-  return lhs.data == rhs.data;
+  return lhs.box_type == rhs.box_type && lhs.data == rhs.data;
 }
 
 inline bool operator==(const PixelAspectRatio& lhs,
@@ -216,7 +216,7 @@ inline bool operator==(const VideoSampleEntry& lhs,
          lhs.data_reference_index == rhs.data_reference_index &&
          lhs.width == rhs.width && lhs.height == rhs.height &&
          lhs.pixel_aspect == rhs.pixel_aspect && lhs.sinf == rhs.sinf &&
-         lhs.codec_config_record == rhs.codec_config_record;
+         lhs.codec_configuration == rhs.codec_configuration;
 }
 
 inline bool operator==(const ESDescriptor& lhs, const ESDescriptor& rhs) {

packager/media/formats/mp4/box_definitions_unittest.cc

@@ -328,24 +328,26 @@ class BoxDefinitionsTestGeneral : public testing::Test {
 
   void Modify(PixelAspectRatio* pasp) { pasp->v_spacing *= 8; }
 
-  void Fill(CodecConfigurationRecord* codec_config_record) {
+  void Fill(CodecConfiguration* codec_configuration) {
-    const uint8_t kAvccData[] = {
+    const uint8_t kCodecConfigurationData[] = {
         0x01, 0x64, 0x00, 0x1f, 0xff, 0xe1, 0x00, 0x18, 0x67, 0x64, 0x00,
         0x1f, 0xac, 0xd9, 0x40, 0x50, 0x05, 0xbb, 0x01, 0x10, 0x00, 0x00,
         0x3e, 0x90, 0x00, 0x0e, 0xa6, 0x00, 0xf1, 0x83, 0x19, 0x60, 0x01,
         0x00, 0x06, 0x68, 0xeb, 0xe3, 0xcb, 0x22, 0xc0};
-    codec_config_record->data.assign(kAvccData,
+    codec_configuration->data.assign(
-                                     kAvccData + arraysize(kAvccData));
+        kCodecConfigurationData,
+        kCodecConfigurationData + arraysize(kCodecConfigurationData));
   }
 
-  void Modify(CodecConfigurationRecord* codec_config_record) {
+  void Modify(CodecConfiguration* codec_configuration) {
-    const uint8_t kAvccData[] = {
+    const uint8_t kCodecConfigurationData[] = {
         0x01, 0x64, 0x00, 0x1e, 0xff, 0xe1, 0x00, 0x19, 0x67, 0x64, 0x00,
         0x1e, 0xac, 0xd9, 0x40, 0xa0, 0x2f, 0xf9, 0x70, 0x11, 0x00, 0x00,
         0x03, 0x03, 0xe9, 0x00, 0x00, 0xea, 0x60, 0x0f, 0x16, 0x2d, 0x96,
         0x01, 0x00, 0x05, 0x68, 0xeb, 0xec, 0xb2, 0x2c};
-    codec_config_record->data.assign(kAvccData,
+    codec_configuration->data.assign(
-                                     kAvccData + arraysize(kAvccData));
+        kCodecConfigurationData,
+        kCodecConfigurationData + arraysize(kCodecConfigurationData));
   }
 
   void Fill(VideoSampleEntry* encv) {
@@ -355,12 +357,12 @@ class BoxDefinitionsTestGeneral : public testing::Test {
     encv->height = 600;
     Fill(&encv->pixel_aspect);
     Fill(&encv->sinf);
-    Fill(&encv->codec_config_record);
+    Fill(&encv->codec_configuration);
   }
 
   void Modify(VideoSampleEntry* encv) {
     encv->height += 600;
-    Modify(&encv->codec_config_record);
+    Modify(&encv->codec_configuration);
   }
 
   void Fill(ElementaryStreamDescriptor* esds) {
@@ -913,11 +915,13 @@ class BoxDefinitionsTestGeneral : public testing::Test {
   bool IsOptional(const ProtectionSchemeInfo* box) { return true; }
   bool IsOptional(const EditList* box) { return true; }
   bool IsOptional(const Edit* box) { return true; }
-  bool IsOptional(const CodecConfigurationRecord* box) { return true; }
+  bool IsOptional(const CodecConfiguration* box) { return true; }
   bool IsOptional(const PixelAspectRatio* box) { return true; }
+  bool IsOptional(const VideoSampleEntry* box) { return true; }
   bool IsOptional(const ElementaryStreamDescriptor* box) { return true; }
   bool IsOptional(const AC3Specific* box) { return true; }
   bool IsOptional(const EC3Specific* box) { return true; }
+  bool IsOptional(const AudioSampleEntry* box) { return true; }
   // Recommended, but optional.
   bool IsOptional(const ProtectionSystemSpecificHeader* box) { return true; }
   bool IsOptional(const WebVTTSourceLabelBox* box) { return true; }
@@ -956,7 +960,6 @@ typedef testing::Types<FileType,
                        HandlerReference,
                        ID3v2,
                        Metadata,
-                       CodecConfigurationRecord,
                        PixelAspectRatio,
                        VideoSampleEntry,
                        ElementaryStreamDescriptor,
@@ -1094,6 +1097,38 @@ TEST_F(BoxDefinitionsTest, MediaHandlerType) {
   ASSERT_EQ(FOURCC_VIDE, media_readback.handler.handler_type);
 }
 
+TEST_F(BoxDefinitionsTest, AvcCodecConfiguration) {
+  CodecConfiguration codec_configuration;
+  Fill(&codec_configuration);
+  codec_configuration.box_type = FOURCC_AVCC;
+  codec_configuration.Write(this->buffer_.get());
+  // Should inherit from Box.
+  EXPECT_EQ(
+      8u, codec_configuration.ComputeSize() - codec_configuration.data.size());
+
+  CodecConfiguration codec_configuration_readback;
+  // BoxType should be provided before parsing the box.
+  codec_configuration_readback.box_type = FOURCC_AVCC;
+  ASSERT_TRUE(ReadBack(&codec_configuration_readback));
+  EXPECT_EQ(codec_configuration, codec_configuration_readback);
+}
+
+TEST_F(BoxDefinitionsTest, VPCodecConfiguration) {
+  CodecConfiguration codec_configuration;
+  Fill(&codec_configuration);
+  codec_configuration.box_type = FOURCC_VPCC;
+  codec_configuration.Write(this->buffer_.get());
+  // Should inherit from FullBox.
+  EXPECT_EQ(
+      12u, codec_configuration.ComputeSize() - codec_configuration.data.size());
+
+  CodecConfiguration codec_configuration_readback;
+  // BoxType should be provided before parsing the box.
+  codec_configuration_readback.box_type = FOURCC_VPCC;
+  ASSERT_TRUE(ReadBack(&codec_configuration_readback));
+  EXPECT_EQ(codec_configuration, codec_configuration_readback);
+}
+
 TEST_F(BoxDefinitionsTest, DTSSampleEntry) {
   AudioSampleEntry entry;
   entry.format = FOURCC_DTSE;

packager/media/formats/mp4/mp4_media_parser.cc

@@ -471,7 +471,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       switch (actual_format) {
         case FOURCC_AVC1: {
           AVCDecoderConfiguration avc_config;
-          if (!avc_config.Parse(entry.codec_config_record.data)) {
+          if (!avc_config.Parse(entry.codec_configuration.data)) {
             LOG(ERROR) << "Failed to parse avcc.";
             return false;
           }
@@ -508,7 +508,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
         case FOURCC_HEV1:
         case FOURCC_HVC1: {
           HEVCDecoderConfiguration hevc_config;
-          if (!hevc_config.Parse(entry.codec_config_record.data)) {
+          if (!hevc_config.Parse(entry.codec_configuration.data)) {
             LOG(ERROR) << "Failed to parse hevc.";
             return false;
           }
@@ -520,7 +520,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
         case FOURCC_VP09:
         case FOURCC_VP10: {
           VPCodecConfiguration vp_config;
-          if (!vp_config.Parse(entry.codec_config_record.data)) {
+          if (!vp_config.Parse(entry.codec_configuration.data)) {
             LOG(ERROR) << "Failed to parse vpcc.";
             return false;
           }
@@ -540,8 +540,8 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
           codec_string, track->media.header.language.code, coded_width,
           coded_height, pixel_width, pixel_height,
           0,  // trick_play_rate
-          nalu_length_size, entry.codec_config_record.data.data(),
+          nalu_length_size, entry.codec_configuration.data.data(),
-          entry.codec_config_record.data.size(), is_encrypted));
+          entry.codec_configuration.data.size(), is_encrypted));
     }
   }
 

packager/media/formats/mp4/mp4_muxer.cc

@@ -232,7 +232,7 @@ void MP4Muxer::GenerateVideoTrak(const VideoStreamInfo* video_info,
   video.format = VideoCodecToFourCC(video_info->codec());
   video.width = video_info->width();
   video.height = video_info->height();
-  video.codec_config_record.data = video_info->extra_data();
+  video.codec_configuration.data = video_info->extra_data();
   if (pixel_width != 1 || pixel_height != 1) {
     video.pixel_aspect.h_spacing = pixel_width;
     video.pixel_aspect.v_spacing = pixel_height;