Render inband text tracks in the media pipeline

media/filters/ffmpeg_demuxer.cc

Index: media/filters/ffmpeg_demuxer.cc
diff --git a/media/filters/ffmpeg_demuxer.cc b/media/filters/ffmpeg_demuxer.cc
index 30c8101baeaa1349c9b2e3e91ea4fbba513a258d..55e3c6e5f4a8b9756b349b0890df41ef95632c27 100644
--- a/media/filters/ffmpeg_demuxer.cc
+++ b/media/filters/ffmpeg_demuxer.cc
@@ -11,11 +11,9 @@
 #include "base/bind.h"
 #include "base/callback.h"
 #include "base/callback_helpers.h"
-#include "base/command_line.h"
 #include "base/memory/scoped_ptr.h"
 #include "base/message_loop/message_loop.h"
 #include "base/metrics/sparse_histogram.h"
-#include "base/stl_util.h"
 #include "base/strings/string_util.h"
 #include "base/strings/stringprintf.h"
 #include "base/task_runner_util.h"
@@ -26,7 +24,6 @@
 #include "media/base/decrypt_config.h"
 #include "media/base/limits.h"
 #include "media/base/media_log.h"
-#include "media/base/media_switches.h"
 #include "media/base/video_decoder_config.h"
 #include "media/ffmpeg/ffmpeg_common.h"
 #include "media/filters/ffmpeg_glue.h"
@@ -64,6 +61,9 @@ FFmpegDemuxerStream::FFmpegDemuxerStream(
       AVStreamToVideoDecoderConfig(stream, &video_config_, true);
       is_encrypted = video_config_.is_encrypted();
       break;
+    case AVMEDIA_TYPE_SUBTITLE:
+      type_ = TEXT;
+      break;
     default:
       NOTREACHED();
       break;
@@ -110,31 +110,62 @@ void FFmpegDemuxerStream::EnqueuePacket(ScopedAVPacket packet) {
     LOG(ERROR) << "Format conversion failed.";
   }
 
-  // Get side data if any. For now, the only type of side_data is VP8 Alpha. We
-  // keep this generic so that other side_data types in the future can be
-  // handled the same way as well.
-  av_packet_split_side_data(packet.get());
-  int side_data_size = 0;
-  uint8* side_data = av_packet_get_side_data(
-      packet.get(),
-      AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
-      &side_data_size);
-
-  // If a packet is returned by FFmpeg's av_parser_parse2() the packet will
-  // reference inner memory of FFmpeg.  As such we should transfer the packet
-  // into memory we control.
   scoped_refptr<DecoderBuffer> buffer;
-  if (side_data_size > 0) {
+
+  // Get side data if any. For now, the only types of side_data are VP8 Alpha,
+  // and WebVTT id and settings. We keep this generic so that other side_data
+  // types in the future can be handled the same way as well.
+  av_packet_split_side_data(packet.get());
+  if (type() == DemuxerStream::TEXT) {
+    int id_size = 0;
+    uint8* id_data = av_packet_get_side_data(
+        packet.get(),
+        AV_PKT_DATA_WEBVTT_IDENTIFIER,
+        &id_size);
+
+    int settings_size = 0;
+    uint8* settings_data = av_packet_get_side_data(
+        packet.get(),
+        AV_PKT_DATA_WEBVTT_SETTINGS,
+        &settings_size);
+
+    // The DecoderBuffer only supports a single side data item. In the case of
+    // a WebVTT cue, we can have potentially two side data items. In order to
+    // avoid disrupting DecoderBuffer any more than we need to, we copy both
+    // side data items onto a single one, and terminate each with a NUL marker.
+    std::vector<uint8> side_data;
+    side_data.reserve(id_size + 1 + settings_size + 1);
+    side_data.insert(side_data.end(),
+                     id_data, id_data + id_size);
+    side_data.push_back(0);
+    side_data.insert(side_data.end(),
+                     settings_data, settings_data + settings_size);
+    side_data.push_back(0);
+
     buffer = DecoderBuffer::CopyFrom(packet.get()->data, packet.get()->size,
-                                     side_data, side_data_size);
+                                     side_data.data(), side_data.size());
   } else {
-    buffer = DecoderBuffer::CopyFrom(packet.get()->data, packet.get()->size);
+    int side_data_size = 0;
+    uint8* side_data = av_packet_get_side_data(
+        packet.get(),
+        AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
+        &side_data_size);
+
+    // If a packet is returned by FFmpeg's av_parser_parse2() the packet will
+    // reference inner memory of FFmpeg.  As such we should transfer the packet
+    // into memory we control.
+    if (side_data_size > 0) {
+      buffer = DecoderBuffer::CopyFrom(packet.get()->data, packet.get()->size,
+                                       side_data, side_data_size);
+    } else {
+      buffer = DecoderBuffer::CopyFrom(packet.get()->data, packet.get()->size);
+    }
   }
 
   if ((type() == DemuxerStream::AUDIO && audio_config_.is_encrypted()) ||
       (type() == DemuxerStream::VIDEO && video_config_.is_encrypted())) {
     scoped_ptr<DecryptConfig> config(WebMCreateDecryptConfig(
-        packet->data,  packet->size,
+        packet->data, packet->size,
         reinterpret_cast<const uint8*>(encryption_key_id_.data()),
         encryption_key_id_.size()));
     if (!config)
@@ -229,6 +260,14 @@ VideoDecoderConfig FFmpegDemuxerStream::video_decoder_config() {
   return video_config_;
 }
 
+TextTrackConfig FFmpegDemuxerStream::text_track_config() {
+  DCHECK(message_loop_->BelongsToCurrentThread());
+  CHECK_EQ(type_, TEXT);
+  return TextTrackConfig(GetTextKind(),
+                         GetMetadata("title"),
+                         GetMetadata("language"));
+}
+
 FFmpegDemuxerStream::~FFmpegDemuxerStream() {
   DCHECK(!demuxer_);
   DCHECK(read_cb_.is_null());
@@ -272,6 +311,27 @@ bool FFmpegDemuxerStream::HasAvailableCapacity() {
   return buffer_queue_.IsEmpty() || buffer_queue_.Duration() < kCapacity;
 }
 
+TextKind FFmpegDemuxerStream::GetTextKind() const {
+  DCHECK_EQ(type_, DemuxerStream::TEXT);
+
+  if (stream_->disposition & AV_DISPOSITION_CAPTIONS)
+    return kTextCaptions;
+
+  if (stream_->disposition & AV_DISPOSITION_DESCRIPTIONS)
+    return kTextDescriptions;
+
+  if (stream_->disposition & AV_DISPOSITION_METADATA)
+    return kTextMetadata;
+
+  return kTextSubtitles;
+}
+
+std::string FFmpegDemuxerStream::GetMetadata(const char* key) const {
+  const AVDictionaryEntry* entry =
+      av_dict_get(stream_->metadata, key, NULL, 0);
+  return (entry == NULL || entry->value == NULL) ? "" : entry->value;
+}
+
 // static
 base::TimeDelta FFmpegDemuxerStream::ConvertStreamTimestamp(
     const AVRational& time_base, int64 timestamp) {
@@ -288,6 +348,7 @@ FFmpegDemuxer::FFmpegDemuxer(
     const scoped_refptr<base::MessageLoopProxy>& message_loop,
     DataSource* data_source,
     const NeedKeyCB& need_key_cb,
+    bool text_enabled,
     const scoped_refptr<MediaLog>& media_log)
     : host_(NULL),
       message_loop_(message_loop),
@@ -300,6 +361,7 @@ FFmpegDemuxer::FFmpegDemuxer(
       bitrate_(0),
       start_time_(kNoTimestamp()),
       audio_disabled_(false),
+      text_enabled_(text_enabled),
       duration_known_(false),
       url_protocol_(data_source, BindToLoop(message_loop_, base::Bind(
           &FFmpegDemuxer::OnDataSourceError, base::Unretained(this)))),
@@ -404,6 +466,22 @@ base::TimeDelta FFmpegDemuxer::GetStartTime() const {
   return start_time_;
 }
 
+void FFmpegDemuxer::AddTextStreams() {
+  DCHECK(message_loop_->BelongsToCurrentThread());
+
+  for (StreamVector::size_type idx = 0; idx < streams_.size(); ++idx) {
+    FFmpegDemuxerStream* stream = streams_[idx];
+    if (stream == NULL || stream->type() != DemuxerStream::TEXT)
+      continue;
+
+    TextKind kind = stream->GetTextKind();
+    std::string title = stream->GetMetadata("title");
+    std::string language = stream->GetMetadata("language");
+
+    host_->AddTextStream(stream, kind, title, language);
+  }
+}
+
 // Helper for calculating the bitrate of the media based on information stored
 // in |format_context| or failing that the size and duration of the media.
 //
@@ -522,6 +600,10 @@ void FFmpegDemuxer::OnFindStreamInfoDone(const PipelineStatusCB& status_cb,
       if (!video_config.IsValidConfig())
         continue;
       video_stream = stream;
+    } else if (codec_type == AVMEDIA_TYPE_SUBTITLE) {
+      if (codec_context->codec_id != AV_CODEC_ID_WEBVTT || !text_enabled_) {
+        continue;
+      }
     } else {
       continue;
     }
@@ -542,6 +624,9 @@ void FFmpegDemuxer::OnFindStreamInfoDone(const PipelineStatusCB& status_cb,
     return;
   }
 
+  if (text_enabled_)
+    AddTextStreams();
+
   if (format_context->duration != static_cast<int64_t>(AV_NOPTS_VALUE)) {
     // If there is a duration value in the container use that to find the
     // maximum between it and the duration from A/V streams.