Full vtt in mp4 path

Change-Id: I7f49dbfed188ff7a451b0b4fbd6947590e8935d1

packager/app/packager_main.cc

@@ -82,7 +82,7 @@ const char kUsage[] =
     "  - input (in): Required input/source media file path or network stream\n"
     "    URL.\n"
     "  - stream_selector (stream): Required field with value 'audio',\n"
-    "    'video', or stream number (zero based).\n"
+    "    'video', 'text', or stream number (zero based).\n"
     "  - output (out): Required output file (single file) or initialization\n"
     "    file path (multiple file).\n"
     "  - segment_template (segment): Optional value which specifies the\n"
@@ -253,8 +253,8 @@ bool CreateRemuxJobs(const StreamDescriptorList& stream_descriptors,
     }
     stream_muxer_options.bandwidth = stream_iter->bandwidth;
 
-    // Handle text input.
-    if (stream_iter->stream_selector == "text") {
+    if (stream_iter->stream_selector == "text" &&
+        stream_iter->output_format != CONTAINER_MOV) {
       MediaInfo text_media_info;
       if (!StreamInfoToTextMediaInfo(*stream_iter, stream_muxer_options,
                                      &text_media_info)) {

packager/media/base/demuxer.cc

@@ -33,6 +33,7 @@ const size_t kQueuedSamplesLimit = 10000;
 const size_t kInvalidStreamIndex = static_cast<size_t>(-1);
 const size_t kBaseVideoOutputStreamIndex = 0x100;
 const size_t kBaseAudioOutputStreamIndex = 0x200;
+const size_t kBaseTextOutputStreamIndex = 0x300;
 
 std::string GetStreamLabel(size_t stream_index) {
   switch (stream_index) {
@@ -40,6 +41,8 @@ std::string GetStreamLabel(size_t stream_index) {
       return "video";
     case kBaseAudioOutputStreamIndex:
       return "audio";
+    case kBaseTextOutputStreamIndex:
+      return "text";
     default:
       return base::SizeTToString(stream_index);
   }
@@ -51,11 +54,13 @@ bool GetStreamIndex(const std::string& stream_label, size_t* stream_index) {
     *stream_index = kBaseVideoOutputStreamIndex;
   } else if (stream_label == "audio") {
     *stream_index = kBaseAudioOutputStreamIndex;
+  } else if (stream_label == "text") {
+    *stream_index = kBaseTextOutputStreamIndex;
   } else {
     // Expect stream_label to be a zero based stream id.
     if (!base::StringToSizeT(stream_label, stream_index)) {
       LOG(ERROR) << "Invalid argument --stream=" << stream_label << "; "
-                 << "should be 'audio', 'video', or a number";
+                 << "should be 'audio', 'video', 'text', or a number";
       return false;
     }
   }
@@ -221,6 +226,9 @@ void Demuxer::ParserInitEvent(
   bool audio_handler_set =
       output_handlers().find(kBaseAudioOutputStreamIndex) !=
       output_handlers().end();
+  bool text_handler_set =
+      output_handlers().find(kBaseTextOutputStreamIndex) !=
+      output_handlers().end();
   for (const std::shared_ptr<StreamInfo>& stream_info : stream_infos) {
     size_t stream_index = base_stream_index;
     if (video_handler_set && stream_info->stream_type() == kStreamVideo) {
@@ -233,6 +241,10 @@ void Demuxer::ParserInitEvent(
       // Only for the first audio stream.
       audio_handler_set = false;
     }
+    if (text_handler_set && stream_info->stream_type() == kStreamText) {
+      stream_index = kBaseTextOutputStreamIndex;
+      text_handler_set = false;
+    }
 
     const bool handler_set =
         output_handlers().find(stream_index) != output_handlers().end();

packager/media/event/muxer_listener_internal.cc

@@ -14,6 +14,7 @@
 #include "packager/media/base/audio_stream_info.h"
 #include "packager/media/base/muxer_options.h"
 #include "packager/media/base/protection_system_specific_info.h"
+#include "packager/media/base/text_stream_info.h"
 #include "packager/media/base/video_stream_info.h"
 #include "packager/media/codecs/ec3_audio_util.h"
 #include "packager/mpd/base/media_info.pb.h"
@@ -134,11 +135,29 @@ void AddAudioInfo(const AudioStreamInfo* audio_stream_info,
   }
 }
 
+void AddTextInfo(const TextStreamInfo& text_stream_info,
+                 MediaInfo* media_info) {
+  MediaInfo::TextInfo* text_info = media_info->mutable_text_info();
+  // For now, set everything as subtitle.
+  text_info->set_type(MediaInfo::TextInfo::SUBTITLE);
+  if (text_stream_info.codec_string() == "wvtt") {
+    text_info->set_format("vtt");
+  } else {
+    LOG(WARNING) << "Unhandled codec " << text_stream_info.codec_string()
+                 << " copying it as format.";
+    text_info->set_format(text_stream_info.codec_string());
+  }
+
+  text_info->set_language(text_stream_info.language());
+}
+
 void SetMediaInfoStreamInfo(const StreamInfo& stream_info,
                             MediaInfo* media_info) {
   if (stream_info.stream_type() == kStreamAudio) {
     AddAudioInfo(static_cast<const AudioStreamInfo*>(&stream_info),
                  media_info);
+  } else if (stream_info.stream_type() == kStreamText) {
+    AddTextInfo(static_cast<const TextStreamInfo&>(stream_info), media_info);
   } else {
     DCHECK_EQ(stream_info.stream_type(), kStreamVideo);
     AddVideoInfo(static_cast<const VideoStreamInfo*>(&stream_info),

packager/media/formats/mp4/mp4_muxer.cc

@@ -13,6 +13,7 @@
 #include "packager/media/base/fourccs.h"
 #include "packager/media/base/key_source.h"
 #include "packager/media/base/media_sample.h"
+#include "packager/media/base/text_stream_info.h"
 #include "packager/media/base/video_stream_info.h"
 #include "packager/media/codecs/es_descriptor.h"
 #include "packager/media/event/muxer_listener.h"
@@ -150,6 +151,10 @@ Status MP4Muxer::InitializeMuxer() {
         GenerateAudioTrak(static_cast<AudioStreamInfo*>(streams()[i].get()),
                           &trak, i + 1);
         break;
+      case kStreamText:
+        GenerateTextTrak(static_cast<TextStreamInfo*>(streams()[i].get()),
+                         &trak, i + 1);
+        break;
       default:
         NOTIMPLEMENTED() << "Not implemented for stream type: "
                          << streams()[i]->stream_type();
@@ -373,6 +378,31 @@ void MP4Muxer::GenerateAudioTrak(const AudioStreamInfo* audio_info,
   }
 }
 
+void MP4Muxer::GenerateTextTrak(const TextStreamInfo* text_info,
+                                Track* trak,
+                                uint32_t track_id) {
+  InitializeTrak(text_info, trak);
+
+  if (text_info->codec_string() == "wvtt") {
+    // Handle WebVTT.
+    TextSampleEntry webvtt;
+    webvtt.format = FOURCC_wvtt;
+    webvtt.config.config.assign(text_info->codec_config().begin(),
+                                text_info->codec_config().end());
+    // TODO(rkuroiwa): This should be the source file URI(s). Putting bogus
+    // string for now so that the box will be there for samples with overlapping
+    // cues.
+    webvtt.label.source_label = "source_label";
+    SampleDescription& sample_description =
+        trak->media.information.sample_table.description;
+    sample_description.type = kText;
+    sample_description.text_entries.push_back(webvtt);
+    return;
+  }
+  NOTIMPLEMENTED() << text_info->codec_string()
+                   << " handling not implemented yet.";
+}
+
 bool MP4Muxer::GetInitRangeStartAndEnd(uint32_t* start, uint32_t* end) {
   DCHECK(start && end);
   size_t range_offset = 0;

packager/media/formats/mp4/mp4_muxer.h

@@ -16,6 +16,7 @@ namespace media {
 
 class AudioStreamInfo;
 class StreamInfo;
+class TextStreamInfo;
 class VideoStreamInfo;
 
 namespace mp4 {
@@ -50,6 +51,9 @@ class MP4Muxer : public Muxer {
   void GenerateVideoTrak(const VideoStreamInfo* video_info,
                          Track* trak,
                          uint32_t track_id);
+  void GenerateTextTrak(const TextStreamInfo* video_info,
+                        Track* trak,
+                        uint32_t track_id);
 
   // Gets |start| and |end| initialization range. Returns true if there is an
   // init range and sets start-end byte-range-spec specified in RFC2616.

packager/mpd/base/mpd_builder.cc

@@ -707,8 +707,10 @@ Representation* AdaptationSet::AddRepresentation(const MediaInfo& media_info) {
   std::unique_ptr<Representation> representation(new Representation(
       media_info, mpd_options_, representation_id, std::move(listener)));
 
-  if (!representation->Init())
+  if (!representation->Init()) {
+    LOG(ERROR) << "Failed to initialize Representation.";
     return NULL;
+  }
 
   // For videos, record the width, height, and the frame rate to calculate the
   // max {width,height,framerate} required for DASH IOP.
@@ -1383,6 +1385,8 @@ std::string Representation::GetTextMimeType() const {
   if (media_info_.text_info().format() == "vtt") {
     if (media_info_.container_type() == MediaInfo::CONTAINER_TEXT) {
       return "text/vtt";
+    } else if (media_info_.container_type() == MediaInfo::CONTAINER_MP4) {
+      return "application/mp4";
     }
     LOG(ERROR) << "Failed to determine MIME type for VTT container: "
                << media_info_.container_type();

packager/mpd/base/mpd_utils.cc

@@ -23,6 +23,10 @@ std::string TextCodecString(const MediaInfo& media_info) {
       (media_info.container_type() == MediaInfo::CONTAINER_MP4)) {
     return "stpp";
   }
+  if (format == "vtt" &&
+      (media_info.container_type() == MediaInfo::CONTAINER_MP4)) {
+    return "wvtt";
+  }
 
   // Otherwise codec doesn't need to be specified, e.g. vtt and ttml+xml are
   // obvious from the mime type.