Allow other text sample entries than wvtt

Before this change, the media file will fail to be parsed if it contains
non-wvtt text streams. This change allows demuxing of video and audio
streams in these media files even though the text stream is ignored.

Fixes #74.

Change-Id: I7ef108a51125c4965fe48c39efe9638f3a0e4759

packager/media/formats/mp4/box_definitions.cc

@@ -586,7 +586,7 @@ bool SampleDescription::ReadWriteInternal(BoxBuffer* buffer) {
       count = audio_entries.size();
       break;
     case kText:
-      count = wvtt_entries.size();
+      count = text_entries.size();
       break;
     default:
       NOTIMPLEMENTED() << "SampleDecryption type " << type
@@ -609,8 +609,8 @@ bool SampleDescription::ReadWriteInternal(BoxBuffer* buffer) {
       RCHECK(reader->ReadAllChildren(&audio_entries));
       RCHECK(audio_entries.size() == count);
     } else if (type == kText) {
-      RCHECK(reader->ReadAllChildren(&wvtt_entries));
-      RCHECK(wvtt_entries.size() == count);
+      RCHECK(reader->ReadAllChildren(&text_entries));
+      RCHECK(text_entries.size() == count);
     }
   } else {
     DCHECK_LT(0u, count);
@@ -622,7 +622,7 @@ bool SampleDescription::ReadWriteInternal(BoxBuffer* buffer) {
         RCHECK(buffer->ReadWriteChild(&audio_entries[i]));
     } else if (type == kText) {
       for (uint32_t i = 0; i < count; ++i)
-        RCHECK(buffer->ReadWriteChild(&wvtt_entries[i]));
+        RCHECK(buffer->ReadWriteChild(&text_entries[i]));
     } else {
       NOTIMPLEMENTED();
     }
@@ -638,6 +638,9 @@ uint32_t SampleDescription::ComputeSizeInternal() {
   } else if (type == kAudio) {
     for (uint32_t i = 0; i < audio_entries.size(); ++i)
       box_size += audio_entries[i].ComputeSize();
+  } else if (type == kText) {
+    for (uint32_t i = 0; i < text_entries.size(); ++i)
+      box_size += text_entries[i].ComputeSize();
   }
   return box_size;
 }
@@ -1589,25 +1592,37 @@ uint32_t WebVTTSourceLabelBox::ComputeSizeInternal() {
   return HeaderSize() + source_label.size();
 }
 
-WVTTSampleEntry::WVTTSampleEntry() {}
-WVTTSampleEntry::~WVTTSampleEntry() {}
+TextSampleEntry::TextSampleEntry() : format(FOURCC_NULL) {}
+TextSampleEntry::~TextSampleEntry() {}
 
-FourCC WVTTSampleEntry::BoxType() const {
-  return FOURCC_wvtt;
+FourCC TextSampleEntry::BoxType() const {
+  if (format == FOURCC_NULL) {
+    LOG(ERROR) << "TextSampleEntry should be parsed according to the "
+               << "handler type recovered in its Media ancestor.";
+  }
+  return format;
 }
 
-bool WVTTSampleEntry::ReadWriteInternal(BoxBuffer* buffer) {
+bool TextSampleEntry::ReadWriteInternal(BoxBuffer* buffer) {
+  if (buffer->Reading()) {
+    DCHECK(buffer->reader());
+    format = buffer->reader()->type();
+  } else {
+    RCHECK(ReadWriteHeaderInternal(buffer));
+  }
+  RCHECK(buffer->IgnoreBytes(6) &&  // reserved for SampleEntry.
+         buffer->ReadWriteUInt16(&data_reference_index));
+
+  if (format == FOURCC_wvtt) {
     // TODO(rkuroiwa): Handle the optional MPEG4BitRateBox.
-  RCHECK(ReadWriteHeaderInternal(buffer) &&
-         buffer->IgnoreBytes(6) &&  // reserved for SampleEntry.
-         buffer->ReadWriteUInt16(&data_reference_index) &&
-         buffer->PrepareChildren() &&
+    RCHECK(buffer->PrepareChildren() &&
            buffer->ReadWriteChild(&config) &&
            buffer->ReadWriteChild(&label));
+  }
   return true;
 }
 
-uint32_t WVTTSampleEntry::ComputeSizeInternal() {
+uint32_t TextSampleEntry::ComputeSizeInternal() {
   // 6 for the (anonymous) reserved bytes for SampleEntry class.
   return HeaderSize() + 6 + sizeof(data_reference_index) +
          config.ComputeSize() + label.ComputeSize();
@@ -1762,7 +1777,7 @@ bool MediaInformation::ReadWriteInternal(BoxBuffer* buffer) {
       RCHECK(buffer->ReadWriteChild(&smhd));
       break;
     case kText:
-      RCHECK(buffer->ReadWriteChild(&sthd));
+      RCHECK(buffer->TryReadWriteChild(&sthd));
       break;
     default:
       NOTIMPLEMENTED();

packager/media/formats/mp4/box_definitions.h

@@ -374,11 +374,16 @@ struct WebVTTSourceLabelBox : Box {
   std::string source_label;
 };
 
-struct WVTTSampleEntry : Box {
-  DECLARE_BOX_METHODS(WVTTSampleEntry);
+struct TextSampleEntry : Box {
+  DECLARE_BOX_METHODS(TextSampleEntry);
 
+  // Specifies fourcc of this sample entry. It needs to be set on write, e.g.
+  // set to 'wvtt' to write WVTTSampleEntry; On read, it is recovered from box
+  // header.
+  FourCC format;
   uint16_t data_reference_index;
 
+  // Sub boxes for wvtt text sample entry.
   WebVTTConfigurationBox config;
   WebVTTSourceLabelBox label;
   // Optional MPEG4BitRateBox.
@@ -392,7 +397,7 @@ struct SampleDescription : FullBox {
   // SampleEntry struct, std::vector<SampleEntry> sample_entries.
   std::vector<VideoSampleEntry> video_entries;
   std::vector<AudioSampleEntry> audio_entries;
-  std::vector<WVTTSampleEntry> wvtt_entries;
+  std::vector<TextSampleEntry> text_entries;
 };
 
 struct DecodingTime {

packager/media/formats/mp4/box_definitions_comparison.h

@@ -270,8 +270,7 @@ inline bool operator==(const WebVTTSourceLabelBox& lhs,
   return lhs.source_label == rhs.source_label;
 }
 
-inline bool operator==(const WVTTSampleEntry& lhs,
-                       const WVTTSampleEntry& rhs) {
+inline bool operator==(const TextSampleEntry& lhs, const TextSampleEntry& rhs) {
   return lhs.config == rhs.config && lhs.label == rhs.label;
 }
 

packager/media/formats/mp4/box_definitions_unittest.cc

@@ -54,7 +54,7 @@ class BoxDefinitionsTestGeneral : public testing::Test {
     return true;
   }
 
-  // FourCC for VideoSampleEntry is not a constant, e.g. could be avc1, or encv.
+  // FourCC for VideoSampleEntry is fixed, e.g. could be avc1, or encv.
   bool ReadBack(VideoSampleEntry* video) {
     scoped_ptr<BoxReader> reader(CreateReader());
     std::vector<VideoSampleEntry> video_entries;
@@ -64,7 +64,7 @@ class BoxDefinitionsTestGeneral : public testing::Test {
     return true;
   }
 
-  // FourCC for AudioSampleEntry is not a constant, e.g. could be mp4a, or enca.
+  // FourCC for AudioSampleEntry is fixed, e.g. could be mp4a, or enca.
   bool ReadBack(AudioSampleEntry* audio) {
     scoped_ptr<BoxReader> reader(CreateReader());
     std::vector<AudioSampleEntry> audio_entries;
@@ -74,6 +74,16 @@ class BoxDefinitionsTestGeneral : public testing::Test {
     return true;
   }
 
+  // FourCC for TextSampleEntry is fixed, e.g. could be text, or wvtt.
+  bool ReadBack(TextSampleEntry* text) {
+    scoped_ptr<BoxReader> reader(CreateReader());
+    std::vector<TextSampleEntry> text_entries;
+    RCHECK(reader->ReadAllChildren(&text_entries));
+    RCHECK(text_entries.size() == 1);
+    *text = text_entries[0];
+    return true;
+  }
+
   // SampleDescription cannot parse on its own. Its type parameter should
   // be preset before scanning the box.
   bool ReadBack(SampleDescription* stsd) {
@@ -429,12 +439,13 @@ class BoxDefinitionsTestGeneral : public testing::Test {
     vlab->source_label = "another_label";
   }
 
-  void Fill(WVTTSampleEntry* wvtt) {
+  void Fill(TextSampleEntry* wvtt) {
+    wvtt->format = FOURCC_wvtt;
     Fill(&wvtt->config);
     Fill(&wvtt->label);
   }
 
-  void Modify(WVTTSampleEntry* wvtt) {
+  void Modify(TextSampleEntry* wvtt) {
     Modify(&wvtt->config);
     Modify(&wvtt->label);
   }
@@ -953,7 +964,7 @@ typedef testing::Types<FileType,
                        AudioSampleEntry,
                        WebVTTConfigurationBox,
                        WebVTTSourceLabelBox,
-                       WVTTSampleEntry,
+                       TextSampleEntry,
                        SampleDescription,
                        DecodingTimeToSample,
                        CompositionTimeToSample,