Refactor and Cleanup Descriptor classes

- Define BaseDescriptor and generic read / write operations.

- Define descriptors: ESDescriptor, DecoderConfigDescriptor,
  DecoderSpecificInfoDescriptor, SLConfigDescriptor.
  DecoderSpecificInfoDescriptor and all other descriptors can now
  handle arbitrary length size, not limiting to 64 byte for
  DecoderSpecificInfoDescriptor, which was placed to limit
  ESDescriptor length size to one byte.

- Now DecoderConfigDescriptor is able to handle reading and writing
  of all fields including buffer_size_db, which was not handled
  earlier.

Fixes #536.

Change-Id: Ia8a775f8bf6e90e3343a85f0e643bc44cd017c7a

packager/media/codecs/es_descriptor.cc

@@ -36,7 +36,7 @@ enum SLPredefinedTags {
 
 // The elementary stream size is specific by up to 4 bytes.
 // The MSB of a byte indicates if there are more bytes for the size.
-bool ReadESSize(BitReader* reader, uint32_t* size) {
+bool ReadDescriptorSize(BitReader* reader, size_t* size) {
   uint8_t msb;
   uint8_t byte;
 
@@ -54,147 +54,180 @@ bool ReadESSize(BitReader* reader, uint32_t* size) {
   return true;
 }
 
-// Descryptor Header Size: 1 byte tag and 1 byte size (we don't support
+void WriteDescriptorSize(size_t size, BufferWriter* writer) {
-// multi-bytes size for now).
+  std::vector<uint8_t> size_bytes;
-const size_t kHeaderSize = 2;
+  while (size > 0) {
-const size_t kMaxDecoderSpecificInfoSize = 64;
+    uint8_t byte = (size & 0x7F);
-const uint32_t kUnknownBitrate = 0;
+    size >>= 7;
-const size_t kBitsInByte = 8;
+    if (!size_bytes.empty())
+      byte |= 0x80;
+    size_bytes.push_back(byte);
+  }
+  for (auto iter = size_bytes.rbegin(); iter != size_bytes.rend(); iter++)
+    writer->AppendInt(*iter);
+}
+
+size_t CountDescriptorSize(size_t size) {
+  size_t num_bytes = 0;
+  while (size > 0) {
+    num_bytes++;
+    size >>= 7;
+  }
+  return num_bytes;
+}
 
 }  // namespace
 
-ESDescriptor::ESDescriptor()
+bool BaseDescriptor::Parse(const std::vector<uint8_t>& data) {
-    : esid_(0),
+  BitReader reader(data.data(), data.size());
-      object_type_(ObjectType::kForbidden),
+  return Read(&reader);
-      max_bitrate_(kUnknownBitrate),
+}
-      avg_bitrate_(kUnknownBitrate) {}
 
-ESDescriptor::~ESDescriptor() {}
+bool BaseDescriptor::Read(BitReader* reader) {
-
-bool ESDescriptor::Parse(const std::vector<uint8_t>& data) {
-  BitReader reader(&data[0], data.size());
   uint8_t tag;
-  uint32_t size;
+  RCHECK(reader->ReadBits(8, &tag));
-  uint8_t stream_dependency_flag;
+  if (tag != static_cast<uint8_t>(tag_)) {
-  uint8_t url_flag;
+    LOG(ERROR) << "Expecting tag " << static_cast<int>(tag_) << ", but seeing "
-  uint8_t ocr_stream_flag;
+               << static_cast<int>(tag);
-  uint16_t dummy;
+    return false;
+  }
+  RCHECK(ReadDescriptorSize(reader, &data_size_));
+  return ReadData(reader);
+}
 
-  RCHECK(reader.ReadBits(8, &tag));
+void BaseDescriptor::Write(BufferWriter* writer) {
-  RCHECK(tag == kESDescrTag);
+  // Compute and update descriptor size.
-  RCHECK(ReadESSize(&reader, &size));
+  size_t size = ComputeSize();
+  size_t buffer_size_before_write = writer->Size();
 
-  RCHECK(reader.ReadBits(16, &esid_));  // ES_ID
+  WriteInternal(writer);
-  RCHECK(reader.ReadBits(1, &stream_dependency_flag));
-  RCHECK(reader.ReadBits(1, &url_flag));
-  RCHECK(!url_flag);  // We don't support url flag
-  RCHECK(reader.ReadBits(1, &ocr_stream_flag));
-  RCHECK(reader.ReadBits(5, &dummy));  // streamPriority
 
-  if (stream_dependency_flag)
+  DCHECK_EQ(size, writer->Size() - buffer_size_before_write);
-    RCHECK(reader.ReadBits(16, &dummy));  // dependsOn_ES_ID
+}
-  if (ocr_stream_flag)
-    RCHECK(reader.ReadBits(16, &dummy));  // OCR_ES_Id
 
-  RCHECK(ParseDecoderConfigDescriptor(&reader));
+size_t BaseDescriptor::ComputeSize() {
+  data_size_ = ComputeDataSize();
+  return 1 + CountDescriptorSize(data_size_) + data_size_;
+}
 
+void BaseDescriptor::WriteHeader(BufferWriter* writer) {
+  writer->AppendInt(static_cast<uint8_t>(tag_));
+  WriteDescriptorSize(data_size_, writer);
+}
+
+bool DecoderSpecificInfoDescriptor::ReadData(BitReader* reader) {
+  data_.resize(data_size());
+  for (uint8_t& data_entry : data_)
+    RCHECK(reader->ReadBits(8, &data_entry));
   return true;
 }
 
-bool ESDescriptor::ParseDecoderConfigDescriptor(BitReader* reader) {
+void DecoderSpecificInfoDescriptor::WriteInternal(BufferWriter* writer) {
-  uint8_t tag;
+  WriteHeader(writer);
-  uint32_t size;
+  writer->AppendVector(data_);
-  uint32_t dummy;
+}
 
-  RCHECK(reader->ReadBits(8, &tag));
+size_t DecoderSpecificInfoDescriptor::ComputeDataSize() {
-  RCHECK(tag == kDecoderConfigDescrTag);
+  return data_.size();
-  RCHECK(ReadESSize(reader, &size));
+}
 
+bool DecoderConfigDescriptor::ReadData(BitReader* reader) {
   const size_t start_pos = reader->bit_position();
   RCHECK(reader->ReadBits(8, &object_type_));
-  RCHECK(reader->ReadBits(32, &dummy));
+
+  int stream_type;
+  RCHECK(reader->ReadBits(6, &stream_type));
+  if (stream_type != kAudioStreamType) {
+    LOG(ERROR) << "Seeing non audio stream type " << stream_type;
+    return false;
+  }
+
+  RCHECK(reader->SkipBits(2));  // Skip |upStream| and |reserved|.
+  RCHECK(reader->ReadBits(24, &buffer_size_db_));
   RCHECK(reader->ReadBits(32, &max_bitrate_));
   RCHECK(reader->ReadBits(32, &avg_bitrate_));
   const size_t fields_bits = reader->bit_position() - start_pos;
 
-  const bool has_child_tags = size * kBitsInByte > fields_bits;
+  const size_t kBitsInByte = 8;
+  const bool has_child_tags = data_size() * kBitsInByte > fields_bits;
+  decoder_specific_info_descriptor_ = DecoderSpecificInfoDescriptor();
   if (has_child_tags)
-    RCHECK(ParseDecoderSpecificInfo(reader));
+    RCHECK(decoder_specific_info_descriptor_.Read(reader));
 
   return true;
 }
 
-bool ESDescriptor::ParseDecoderSpecificInfo(BitReader* reader) {
+void DecoderConfigDescriptor::WriteInternal(BufferWriter* writer) {
-  DCHECK(reader);
+  WriteHeader(writer);
-  uint8_t tag;
-  uint32_t size;
-
-  RCHECK(reader->ReadBits(8, &tag));
-  RCHECK(tag == kDecoderSpecificInfoTag);
-  RCHECK(ReadESSize(reader, &size));
-
-  decoder_specific_info_.resize(size);
-  for (uint32_t i = 0; i < size; ++i)
-    RCHECK(reader->ReadBits(8, &decoder_specific_info_[i]));
-  return true;
-}
-
-void ESDescriptor::Write(BufferWriter* writer) const {
-  DCHECK(writer);
-  CHECK_LT(decoder_specific_info_.size(), kMaxDecoderSpecificInfoSize);
-
-  const std::vector<uint8_t> kEmptyDecodingBufferSize(3, 0);
-  const uint8_t kNoEsFlags = 0;
-
-  const uint8_t decoder_specific_info_size =
-    static_cast<uint8_t>(decoder_specific_info_.size());
 
+  writer->AppendInt(static_cast<uint8_t>(object_type_));
   // 6 bit stream type. The last bit is reserved with 1.
   const uint8_t stream_type = (kAudioStreamType << 2) | 1;
-  const uint8_t decoder_config_size =
-    static_cast<uint8_t>(decoder_specific_info_size + kHeaderSize +
-       sizeof(uint8_t) +  // object_type_.
-       sizeof(stream_type) +
-       kEmptyDecodingBufferSize.size() +
-       sizeof(kUnknownBitrate) * 2);
-
-  const uint8_t sl_config_size = sizeof(uint8_t);  // predefined.
-  const uint8_t es_size = decoder_config_size + kHeaderSize + sl_config_size +
-                          kHeaderSize + sizeof(esid_) + sizeof(kNoEsFlags);
-
-  writer->AppendInt(static_cast<uint8_t>(kESDescrTag));
-  writer->AppendInt(es_size);
-  writer->AppendInt(esid_);
-  writer->AppendInt(kNoEsFlags);
-
-  writer->AppendInt(static_cast<uint8_t>(kDecoderConfigDescrTag));
-  writer->AppendInt(decoder_config_size);
-  writer->AppendInt(static_cast<uint8_t>(object_type_));
   writer->AppendInt(stream_type);
-  writer->AppendVector(kEmptyDecodingBufferSize);
+  writer->AppendNBytes(buffer_size_db_, 3);
   writer->AppendInt(max_bitrate_);
   writer->AppendInt(avg_bitrate_);
+  decoder_specific_info_descriptor_.Write(writer);
+}
 
-  writer->AppendInt(static_cast<uint8_t>(kDecoderSpecificInfoTag));
+size_t DecoderConfigDescriptor::ComputeDataSize() {
-  writer->AppendInt(decoder_specific_info_size);
+  // object_type (1 byte), stream_type (1 byte), decoding_buffer_size (3 bytes),
-  writer->AppendVector(decoder_specific_info_);
+  // max_bitrate (4 bytes), avg_bitrate (4 bytes).
+  const size_t data_size_without_children = 1 + 1 + 3 + 4 + 4;
+  return data_size_without_children +
+         decoder_specific_info_descriptor_.ComputeSize();
+}
 
-  writer->AppendInt(static_cast<uint8_t>(kSLConfigTag));
+bool SLConfigDescriptor::ReadData(BitReader* reader) {
-  writer->AppendInt(sl_config_size);
+  return true;
+}
+
+void SLConfigDescriptor::WriteInternal(BufferWriter* writer) {
+  WriteHeader(writer);
   writer->AppendInt(static_cast<uint8_t>(kSLPredefinedMP4));
 }
 
-size_t ESDescriptor::ComputeSize() const {
+size_t SLConfigDescriptor::ComputeDataSize() {
-  // A bit magical. Refer to ESDescriptor::Write for details.
+  return 1;
-  const uint8_t decoder_specific_info_size =
+}
-    static_cast<uint8_t>(decoder_specific_info_.size());
+
-  const uint8_t decoder_config_size = decoder_specific_info_size + kHeaderSize +
+bool ESDescriptor::ReadData(BitReader* reader) {
-                                      sizeof(uint8_t) * 5 +
+  bool stream_dependency_flag;
-                                      sizeof(uint32_t) * 2;
+  bool url_flag;
-  const uint8_t sl_config_size = sizeof(uint8_t);
+  bool ocr_stream_flag;
-  const uint8_t es_size = decoder_config_size + kHeaderSize + sl_config_size +
+  RCHECK(reader->ReadBits(16, &esid_));
-                          kHeaderSize + sizeof(esid_) + sizeof(uint8_t);
+  RCHECK(reader->ReadBits(1, &stream_dependency_flag));
-  return es_size + kHeaderSize;
+  RCHECK(reader->ReadBits(1, &url_flag));
+  RCHECK(!url_flag);  // We don't support url flag
+  RCHECK(reader->ReadBits(1, &ocr_stream_flag));
+  RCHECK(reader->SkipBits(5));  // streamPriority
+
+  if (stream_dependency_flag)
+    RCHECK(reader->SkipBits(16));  // dependsOn_ES_ID
+  if (ocr_stream_flag)
+    RCHECK(reader->SkipBits(16));  // OCR_ES_Id
+
+  return decoder_config_descriptor_.Read(reader);
+  // Skip the parsing of |sl_config_descriptor_| intentionally as we do not care
+  // about the data.
+}
+
+void ESDescriptor::WriteInternal(BufferWriter* writer) {
+  WriteHeader(writer);
+
+  writer->AppendInt(esid_);
+  const uint8_t kNoEsFlags = 0;
+  writer->AppendInt(kNoEsFlags);
+
+  decoder_config_descriptor_.Write(writer);
+  sl_config_descriptor_.Write(writer);
+}
+
+size_t ESDescriptor::ComputeDataSize() {
+  // esid (2 bytes), es_flags (1 byte).
+  const size_t data_size_without_children = 2 + 1;
+  return data_size_without_children + decoder_config_descriptor_.ComputeSize() +
+         sl_config_descriptor_.ComputeSize();
 }
 
 }  // namespace media

packager/media/codecs/es_descriptor.h

@@ -28,22 +28,88 @@ enum class ObjectType : uint8_t {
   kDTSL = 0xAB,                // DTS-HD Master Audio
 };
 
-/// This class parses object type and decoder specific information from an
+enum class DescriptorTag {
-/// elementary stream descriptor, which is usually contained in an esds
+  kForbidden = 0,
-/// box. Please refer to ISO 14496 Part 1 7.2.6.5 for more details.
+  kES = 0x03,
-class ESDescriptor {
+  kDecoderConfig = 0x04,
+  kDecoderSpecificInfo = 0x05,
+  kSLConfig = 0x06,
+};
+
+/// Defines the base Descriptor object as defined in ISO 14496-1:2004 Systems
+/// section 7.2.2.2. All descriptors inherit from either BaseDescriptor.
+class BaseDescriptor {
  public:
-  ESDescriptor();
+  explicit BaseDescriptor(DescriptorTag tag) : tag_(tag) {}
-  ~ESDescriptor();
 
+  /// Parse the descriptor from input data.
+  /// @param data contains the descriptor data.
   bool Parse(const std::vector<uint8_t>& data);
-  void Write(BufferWriter* writer) const;
-  size_t ComputeSize() const;
 
-  uint16_t esid() const { return esid_; }
+  /// Read the descriptor.
-  void set_esid(uint16_t esid) { esid_ = esid; }
+  /// @param reader points to a BitReader object.
+  bool Read(BitReader* reader);
 
-  uint32_t max_bitrate() const {return max_bitrate_; }
+  /// Write the descriptor to buffer. This function calls ComputeSize internally
+  /// to compute and update descriptor size.
+  /// @param writer points to a BufferWriter object which wraps the buffer for
+  ///        writing.
+  void Write(BufferWriter* writer);
+
+  /// Compute the size of this descriptor. It will also update descriptor size.
+  /// @return The size of result descriptor including child descriptors.
+  size_t ComputeSize();
+
+ protected:
+  /// Write descriptor header.
+  void WriteHeader(BufferWriter* writer);
+
+  /// @return descriptor data size without header in bytes.
+  size_t data_size() const { return data_size_; }
+
+ private:
+  // Read the descriptor data (header is already read).
+  virtual bool ReadData(BitReader* reader) = 0;
+  // Write the descriptor. The descriptor data size should already be updated.
+  virtual void WriteInternal(BufferWriter* writer) = 0;
+  // Compute the data size, with child descriptors included.
+  virtual size_t ComputeDataSize() = 0;
+
+  DescriptorTag tag_ = DescriptorTag::kForbidden;
+  size_t data_size_ = 0;
+};
+
+/// Implements DecoderSpecificInfo descriptor according to ISO
+/// 14496-1:2004 7.2.6.7 DecoderSpecificInfo.
+class DecoderSpecificInfoDescriptor : public BaseDescriptor {
+ public:
+  DecoderSpecificInfoDescriptor()
+      : BaseDescriptor(DescriptorTag::kDecoderSpecificInfo) {}
+
+  const std::vector<uint8_t>& data() const { return data_; }
+
+  void set_data(const std::vector<uint8_t>& data) { data_ = data; }
+
+ private:
+  bool ReadData(BitReader* reader) override;
+  void WriteInternal(BufferWriter* writer) override;
+  size_t ComputeDataSize() override;
+
+  std::vector<uint8_t> data_;
+};
+
+/// Implements DecoderConfig descriptor according to ISO 14496-1:2004 7.2.6.6
+/// DecoderConfigDescriptor.
+class DecoderConfigDescriptor : public BaseDescriptor {
+ public:
+  DecoderConfigDescriptor() : BaseDescriptor(DescriptorTag::kDecoderConfig) {}
+
+  uint32_t buffer_size_db() const { return buffer_size_db_; }
+  void set_buffer_size_db(uint32_t buffer_size_db) {
+    buffer_size_db_ = buffer_size_db;
+  }
+
+  uint32_t max_bitrate() const { return max_bitrate_; }
   void set_max_bitrate(uint32_t max_bitrate) { max_bitrate_ = max_bitrate; }
 
   uint32_t avg_bitrate() const { return avg_bitrate_; }
@@ -52,20 +118,13 @@ class ESDescriptor {
   ObjectType object_type() const { return object_type_; }
   void set_object_type(ObjectType object_type) { object_type_ = object_type; }
 
-  const std::vector<uint8_t>& decoder_specific_info() const {
-    return decoder_specific_info_;
-  }
-  void set_decoder_specific_info(
-      const std::vector<uint8_t>& decoder_specific_info) {
-    decoder_specific_info_ = decoder_specific_info;
-  }
-
   /// @return true if the stream is AAC.
   bool IsAAC() const {
     return object_type_ == ObjectType::kISO_14496_3 ||
            object_type_ == ObjectType::kISO_13818_7_AAC_LC;
   }
 
+  /// @return true if the stream is DTS.
   bool IsDTS() const {
     return object_type_ == ObjectType::kDTSC ||
            object_type_ == ObjectType::kDTSE ||
@@ -73,22 +132,66 @@ class ESDescriptor {
            object_type_ == ObjectType::kDTSL;
   }
 
+  const DecoderSpecificInfoDescriptor& decoder_specific_info_descriptor()
+      const {
+    return decoder_specific_info_descriptor_;
+  }
+
+  DecoderSpecificInfoDescriptor* mutable_decoder_specific_info_descriptor() {
+    return &decoder_specific_info_descriptor_;
+  }
+
  private:
-  enum Tag {
+  bool ReadData(BitReader* reader) override;
-    kESDescrTag = 0x03,
+  void WriteInternal(BufferWriter* writer) override;
-    kDecoderConfigDescrTag = 0x04,
+  size_t ComputeDataSize() override;
-    kDecoderSpecificInfoTag = 0x05,
-    kSLConfigTag = 0x06,
-  };
 
-  bool ParseDecoderConfigDescriptor(BitReader* reader);
+  ObjectType object_type_ = ObjectType::kForbidden;
-  bool ParseDecoderSpecificInfo(BitReader* reader);
+  uint32_t buffer_size_db_ = 0;
+  uint32_t max_bitrate_ = 0;
+  uint32_t avg_bitrate_ = 0;
+  DecoderSpecificInfoDescriptor decoder_specific_info_descriptor_;
+};
 
-  uint16_t esid_;  // Elementary Stream ID.
+/// Implements SLConfig descriptor according to ISO 14496-1:2004 7.2.6.8
-  ObjectType object_type_;
+/// SLConfigDescriptor.
-  uint32_t max_bitrate_;
+class SLConfigDescriptor : public BaseDescriptor {
-  uint32_t avg_bitrate_;
+ public:
-  std::vector<uint8_t> decoder_specific_info_;
+  SLConfigDescriptor() : BaseDescriptor(DescriptorTag::kSLConfig) {}
+
+ private:
+  bool ReadData(BitReader* reader) override;
+  void WriteInternal(BufferWriter* writer) override;
+  size_t ComputeDataSize() override;
+};
+
+/// This class parses object type and decoder specific information from an
+/// elementary stream descriptor, which is usually contained in an esds
+/// box. Please refer to ISO 14496 Part 1 7.2.6.5 for more details.
+class ESDescriptor : public BaseDescriptor {
+ public:
+  ESDescriptor() : BaseDescriptor(DescriptorTag::kES) {}
+
+  uint16_t esid() const { return esid_; }
+  void set_esid(uint16_t esid) { esid_ = esid; }
+
+  const DecoderConfigDescriptor& decoder_config_descriptor() const {
+    return decoder_config_descriptor_;
+  }
+
+  DecoderConfigDescriptor* mutable_decoder_config_descriptor() {
+    return &decoder_config_descriptor_;
+  }
+
+ private:
+  bool ReadData(BitReader* reader) override;
+  void WriteInternal(BufferWriter* writer) override;
+  size_t ComputeDataSize() override;
+
+  uint16_t esid_ = 0;  // Elementary Stream ID.
+
+  DecoderConfigDescriptor decoder_config_descriptor_;
+  SLConfigDescriptor sl_config_descriptor_;
 };
 
 }  // namespace media

packager/media/codecs/es_descriptor_unittest.cc

@@ -7,7 +7,10 @@
 #include <gmock/gmock.h>
 #include <gtest/gtest.h>
 
+#include "packager/media/base/buffer_writer.h"
+
 using ::testing::ElementsAre;
+using ::testing::ElementsAreArray;
 
 namespace shaka {
 namespace media {
@@ -38,11 +41,21 @@ TEST(ESDescriptorTest, SingleByteLengthTest) {
   std::vector<uint8_t> data(std::begin(kBuffer), std::end(kBuffer));
 
   ESDescriptor es_desc;
-  EXPECT_EQ(es_desc.object_type(), ObjectType::kForbidden);
+  const DecoderConfigDescriptor& decoder_config_descriptor =
+      es_desc.decoder_config_descriptor();
+  EXPECT_EQ(decoder_config_descriptor.object_type(), ObjectType::kForbidden);
   EXPECT_TRUE(es_desc.Parse(data));
 
-  EXPECT_EQ(es_desc.object_type(), ObjectType::kISO_14496_3);
+  EXPECT_EQ(decoder_config_descriptor.object_type(), ObjectType::kISO_14496_3);
-  EXPECT_THAT(es_desc.decoder_specific_info(), ElementsAre(0x12, 0x10));
+  EXPECT_THAT(
+      decoder_config_descriptor.decoder_specific_info_descriptor().data(),
+      ElementsAre(0x12, 0x10));
+
+  BufferWriter writer;
+  es_desc.Write(&writer);
+  EXPECT_THAT(
+      std::vector<uint8_t>(writer.Buffer(), writer.Buffer() + writer.Size()),
+      ElementsAreArray(kBuffer));
 }
 
 TEST(ESDescriptorTest, NonAACTest) {
@@ -73,9 +86,19 @@ TEST(ESDescriptorTest, NonAACTest) {
   ESDescriptor es_desc;
   EXPECT_TRUE(es_desc.Parse(data));
 
-  EXPECT_EQ(static_cast<int>(es_desc.object_type()), 0x66);
+  const DecoderConfigDescriptor& decoder_config_descriptor =
-  EXPECT_NE(es_desc.object_type(), ObjectType::kISO_14496_3);
+      es_desc.decoder_config_descriptor();
-  EXPECT_THAT(es_desc.decoder_specific_info(), ElementsAre(0x12, 0x10));
+  EXPECT_EQ(static_cast<int>(decoder_config_descriptor.object_type()), 0x66);
+  EXPECT_NE(decoder_config_descriptor.object_type(), ObjectType::kISO_14496_3);
+  EXPECT_THAT(
+      decoder_config_descriptor.decoder_specific_info_descriptor().data(),
+      ElementsAre(0x12, 0x10));
+
+  BufferWriter writer;
+  es_desc.Write(&writer);
+  EXPECT_THAT(
+      std::vector<uint8_t>(writer.Buffer(), writer.Buffer() + writer.Size()),
+      ElementsAreArray(kBuffer));
 }
 
 TEST(ESDescriptorTest, NonAACWithoutDecoderSpecificInfoTagTest) {
@@ -102,42 +125,157 @@ TEST(ESDescriptorTest, NonAACWithoutDecoderSpecificInfoTagTest) {
   ESDescriptor es_desc;
   EXPECT_TRUE(es_desc.Parse(data));
 
-  EXPECT_EQ(static_cast<int>(es_desc.object_type()), 0x6b);
+  const DecoderConfigDescriptor& decoder_config_descriptor =
-  EXPECT_EQ(es_desc.max_bitrate(), 0x28500u);
+      es_desc.decoder_config_descriptor();
-  EXPECT_EQ(es_desc.avg_bitrate(), 0x27100u);
+  EXPECT_EQ(static_cast<int>(decoder_config_descriptor.object_type()), 0x6b);
-  EXPECT_THAT(es_desc.decoder_specific_info(), ElementsAre());
+  EXPECT_EQ(decoder_config_descriptor.max_bitrate(), 0x28500u);
+  EXPECT_EQ(decoder_config_descriptor.avg_bitrate(), 0x27100u);
+  EXPECT_THAT(
+      decoder_config_descriptor.decoder_specific_info_descriptor().data(),
+      ElementsAre());
 }
 
-TEST(ESDescriptorTest, MultiByteLengthTest) {
+// https://github.com/google/shaka-packager/issues/536.
+TEST(ESDescriptorTest, Issue536) {
   // clang-format off
-  const uint8_t kBuffer[] = {
+  const uint8_t kInput[] = {
-      // ESDescriptor tag with two bytes size.
+      // ESDescriptor tag with size.
-      0x03, 0x80, 0x1b,
+      0x03, 0x80, 0x80, 0x80, 0x70,
         // ESDescriptor fields.
-        0x00, 0x01, 0x00,
+        0x00, 0x00, 0x00,
-        // DecoderConfigDescriptor tag with three bytes size.
+        // DecoderConfigDescriptor tag with size.
-        0x04, 0x80, 0x80, 0x14,
+        0x04, 0x80, 0x80, 0x80, 0x62,
           // Object Type.
           0x40,
           // Three 4-byte fields: dummy, max bitrate, avg bitrate.
-          0x15, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+          0x15, 0x00, 0x30, 0x00, 0x00, 0x01, 0xF4, 0x00,
-          // DecoderSpecificInfo tag with four bytes size.
+          0x00, 0x01, 0xF4, 0x00,
-          0x05, 0x80, 0x80, 0x80, 0x02,
+          // DecoderSpecificInfo tag with size.
+          0x05, 0x80, 0x80, 0x80, 0x50,
             // DecoderSpecificInfo fields.
-            0x12, 0x10,
+            0x11, 0x90, 0x08, 0xC4, 0x00, 0x00, 0x20, 0x00,
-        // SLConfig tag with one byte size.
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+        // SLConfig tag with size.
+        0x06, 0x80, 0x80, 0x80, 0x01,
+          // SLConfig fields.
+          0x02,
+  };
+  const uint8_t kOutput[] = {
+      // ESDescriptor tag with size.
+      0x03, 0x67,
+        // ESDescriptor fields.
+        0x00, 0x00, 0x00,
+        // DecoderConfigDescriptor tag with size.
+        0x04, 0x5F,
+          // Object Type.
+          0x40,
+          // Three 4-byte fields: dummy, max bitrate, avg bitrate.
+          0x15, 0x00, 0x30, 0x00, 0x00, 0x01, 0xF4, 0x00,
+          0x00, 0x01, 0xF4, 0x00,
+          // DecoderSpecificInfo tag with size.
+          0x05, 0x50,
+            // DecoderSpecificInfo fields.
+            0x11, 0x90, 0x08, 0xC4, 0x00, 0x00, 0x20, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+            0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+        // SLConfig tag with size.
         0x06, 0x01,
           // SLConfig fields.
           0x02,
   };
   // clang-format on
-  std::vector<uint8_t> data(std::begin(kBuffer), std::end(kBuffer));
+  std::vector<uint8_t> data(std::begin(kInput), std::end(kInput));
 
   ESDescriptor es_desc;
   EXPECT_TRUE(es_desc.Parse(data));
 
-  EXPECT_EQ(es_desc.object_type(), ObjectType::kISO_14496_3);
+  BufferWriter writer;
-  EXPECT_THAT(es_desc.decoder_specific_info(), ElementsAre(0x12, 0x10));
+  es_desc.Write(&writer);
+  EXPECT_THAT(
+      std::vector<uint8_t>(writer.Buffer(), writer.Buffer() + writer.Size()),
+      ElementsAreArray(kOutput));
+}
+
+class DescriptorLengthTest : public testing::Test {
+ public:
+  void TestReadWrite(const std::vector<uint8_t>& input,
+                     const std::vector<uint8_t>& expected_output) {
+    DecoderSpecificInfoDescriptor desc;
+    EXPECT_TRUE(desc.Parse(input));
+
+    BufferWriter writer;
+    desc.Write(&writer);
+    EXPECT_THAT(
+        std::vector<uint8_t>(writer.Buffer(), writer.Buffer() + writer.Size()),
+        ElementsAreArray(expected_output));
+  }
+};
+
+// Use DecoderSpecificInfo descriptor for length testing.
+
+TEST_F(DescriptorLengthTest, OneByteLengthData) {
+  const uint8_t kBuffer[] = {0x05, 0x02, 0x12, 0x10};
+  std::vector<uint8_t> data(std::begin(kBuffer), std::end(kBuffer));
+  TestReadWrite(data, data);
+}
+
+TEST_F(DescriptorLengthTest, TwoBytesLengthForOneByteLengthData) {
+  const uint8_t kInput[] = {0x05, 0x80, 0x02, 0x12, 0x10};
+  const uint8_t kOutput[] = {0x05, 0x02, 0x12, 0x10};
+  std::vector<uint8_t> input(std::begin(kInput), std::end(kInput));
+  std::vector<uint8_t> output(std::begin(kOutput), std::end(kOutput));
+  TestReadWrite(input, output);
+}
+
+TEST_F(DescriptorLengthTest, ThreeBytesLengthForOneByteLengthData) {
+  const uint8_t kInput[] = {0x05, 0x80, 0x80, 0x02, 0x12, 0x10};
+  const uint8_t kOutput[] = {0x05, 0x02, 0x12, 0x10};
+  std::vector<uint8_t> input(std::begin(kInput), std::end(kInput));
+  std::vector<uint8_t> output(std::begin(kOutput), std::end(kOutput));
+  TestReadWrite(input, output);
+}
+
+TEST_F(DescriptorLengthTest, FourBytesLengthForOneByteLengthData) {
+  const uint8_t kInput[] = {0x05, 0x80, 0x80, 0x80, 0x02, 0x12, 0x10};
+  const uint8_t kOutput[] = {0x05, 0x02, 0x12, 0x10};
+  std::vector<uint8_t> input(std::begin(kInput), std::end(kInput));
+  std::vector<uint8_t> output(std::begin(kOutput), std::end(kOutput));
+  TestReadWrite(input, output);
+}
+
+TEST_F(DescriptorLengthTest, TwoBytesLengthData) {
+  const uint8_t kBuffer[] = {
+      0x05, 0x81, 0x02, 0x12, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00,
+  };
+  std::vector<uint8_t> input(std::begin(kBuffer), std::end(kBuffer));
+  std::vector<uint8_t> output(std::begin(kBuffer), std::end(kBuffer));
+  TestReadWrite(input, output);
 }
 
 }  // namespace media

packager/media/formats/mp4/box_definitions.cc

@@ -1602,9 +1602,11 @@ bool ElementaryStreamDescriptor::ReadWriteInternal(BoxBuffer* buffer) {
     std::vector<uint8_t> data;
     RCHECK(buffer->ReadWriteVector(&data, buffer->BytesLeft()));
     RCHECK(es_descriptor.Parse(data));
-    if (es_descriptor.IsAAC()) {
+    if (es_descriptor.decoder_config_descriptor().IsAAC()) {
       RCHECK(aac_audio_specific_config.Parse(
-          es_descriptor.decoder_specific_info()));
+          es_descriptor.decoder_config_descriptor()
+              .decoder_specific_info_descriptor()
+              .data()));
     }
   } else {
     DCHECK(buffer->writer());
@@ -1615,8 +1617,10 @@ bool ElementaryStreamDescriptor::ReadWriteInternal(BoxBuffer* buffer) {
 
 size_t ElementaryStreamDescriptor::ComputeSizeInternal() {
   // This box is optional. Skip it if not initialized.
-  if (es_descriptor.object_type() == ObjectType::kForbidden)
+  if (es_descriptor.decoder_config_descriptor().object_type() ==
+      ObjectType::kForbidden) {
     return 0;
+  }
   return HeaderSize() + es_descriptor.ComputeSize();
 }
 

packager/media/formats/mp4/box_definitions_comparison.h

@@ -256,11 +256,24 @@ inline bool operator==(const VideoSampleEntry& lhs,
          lhs.codec_configuration == rhs.codec_configuration;
 }
 
-inline bool operator==(const ESDescriptor& lhs, const ESDescriptor& rhs) {
+inline bool operator==(const DecoderSpecificInfoDescriptor& lhs,
-  return lhs.esid() == rhs.esid() && lhs.object_type() == rhs.object_type() &&
+                       const DecoderSpecificInfoDescriptor& rhs) {
+  return lhs.data() == rhs.data();
+}
+
+inline bool operator==(const DecoderConfigDescriptor& lhs,
+                       const DecoderConfigDescriptor& rhs) {
+  return lhs.buffer_size_db() == rhs.buffer_size_db() &&
          lhs.max_bitrate() == rhs.max_bitrate() &&
          lhs.avg_bitrate() == rhs.avg_bitrate() &&
-         lhs.decoder_specific_info() == rhs.decoder_specific_info();
+         lhs.object_type() == rhs.object_type() &&
+         lhs.decoder_specific_info_descriptor() ==
+             rhs.decoder_specific_info_descriptor();
+}
+
+inline bool operator==(const ESDescriptor& lhs, const ESDescriptor& rhs) {
+  return lhs.esid() == rhs.esid() &&
+         lhs.decoder_config_descriptor() == rhs.decoder_config_descriptor();
 }
 
 inline bool operator==(const ElementaryStreamDescriptor& lhs,

packager/media/formats/mp4/box_definitions_unittest.cc

@@ -373,11 +373,14 @@ class BoxDefinitionsTestGeneral : public testing::Test {
   void Fill(ElementaryStreamDescriptor* esds) {
     const uint8_t kDecoderSpecificInfo[] = {18, 16};
     esds->es_descriptor.set_esid(1);
-    esds->es_descriptor.set_object_type(ObjectType::kISO_14496_3);
+    esds->es_descriptor.mutable_decoder_config_descriptor()->set_object_type(
+        ObjectType::kISO_14496_3);
     std::vector<uint8_t> decoder_specific_info(
         kDecoderSpecificInfo,
         kDecoderSpecificInfo + sizeof(kDecoderSpecificInfo));
-    esds->es_descriptor.set_decoder_specific_info(decoder_specific_info);
+    esds->es_descriptor.mutable_decoder_config_descriptor()
+        ->mutable_decoder_specific_info_descriptor()
+        ->set_data(decoder_specific_info);
   }
 
   void Modify(ElementaryStreamDescriptor* esds) {

packager/media/formats/mp4/mp4_media_parser.cc

@@ -392,11 +392,13 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       std::vector<uint8_t> codec_config;
 
       switch (actual_format) {
-        case FOURCC_mp4a:
+        case FOURCC_mp4a: {
-          max_bitrate = entry.esds.es_descriptor.max_bitrate();
+          const DecoderConfigDescriptor& decoder_config =
-          avg_bitrate = entry.esds.es_descriptor.avg_bitrate();
+              entry.esds.es_descriptor.decoder_config_descriptor();
+          max_bitrate = decoder_config.max_bitrate();
+          avg_bitrate = decoder_config.avg_bitrate();
 
-          codec = ObjectTypeToCodec(entry.esds.es_descriptor.object_type());
+          codec = ObjectTypeToCodec(decoder_config.object_type());
           if (codec == kCodecAAC) {
             const AACAudioSpecificConfig& aac_audio_specific_config =
                 entry.esds.aac_audio_specific_config;
@@ -404,18 +406,19 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
             sampling_frequency =
                 aac_audio_specific_config.GetSamplesPerSecond();
             audio_object_type = aac_audio_specific_config.GetAudioObjectType();
-            codec_config = entry.esds.es_descriptor.decoder_specific_info();
+            codec_config =
+                decoder_config.decoder_specific_info_descriptor().data();
           } else if (codec == kUnknownCodec) {
             // Intentionally not to fail in the parser as there may be multiple
             // streams in the source content, which allows the supported stream
             // to be packaged. An error will be returned if the unsupported
             // stream is passed to the muxer.
             LOG(WARNING) << "Unsupported audio object type "
-                         << static_cast<int>(
+                         << static_cast<int>(decoder_config.object_type())
-                                entry.esds.es_descriptor.object_type())
                          << " in stsd.es_desriptor.";
           }
           break;
+        }
         case FOURCC_dtsc:
           FALLTHROUGH_INTENDED;
         case FOURCC_dtse:

packager/media/formats/mp4/mp4_muxer.cc

@@ -444,15 +444,17 @@ bool MP4Muxer::GenerateAudioTrak(const AudioStreamInfo* audio_info,
   audio.format =
       CodecToFourCC(audio_info->codec(), H26xStreamFormat::kUnSpecified);
   switch(audio_info->codec()){
-    case kCodecAAC:
+    case kCodecAAC: {
-      audio.esds.es_descriptor.set_object_type(
-          ObjectType::kISO_14496_3);  // MPEG4 AAC.
       audio.esds.es_descriptor.set_esid(track_id);
-      audio.esds.es_descriptor.set_decoder_specific_info(
+      DecoderConfigDescriptor* decoder_config =
+          audio.esds.es_descriptor.mutable_decoder_config_descriptor();
+      decoder_config->set_object_type(ObjectType::kISO_14496_3);  // MPEG4 AAC.
+      decoder_config->set_max_bitrate(audio_info->max_bitrate());
+      decoder_config->set_avg_bitrate(audio_info->avg_bitrate());
+      decoder_config->mutable_decoder_specific_info_descriptor()->set_data(
           audio_info->codec_config());
-      audio.esds.es_descriptor.set_max_bitrate(audio_info->max_bitrate());
-      audio.esds.es_descriptor.set_avg_bitrate(audio_info->avg_bitrate());
       break;
+    }
     case kCodecDTSC:
     case kCodecDTSH:
     case kCodecDTSL:

packager/media/formats/wvm/wvm_media_parser.cc

@@ -716,7 +716,9 @@ bool WvmMediaParser::ParseIndexEntry() {
                 "Could not extract AudioSpecificConfig from ES_Descriptor";
             return false;
           }
-          audio_codec_config = descriptor.decoder_specific_info();
+          audio_codec_config = descriptor.decoder_config_descriptor()
+                                   .decoder_specific_info_descriptor()
+                                   .data();
           break;
         }
         case Audio_EC3SpecificData: