Consider text tracks in the frame processor for new media segments.

media/filters/source_buffer_stream.cc

Index: media/filters/source_buffer_stream.cc
diff --git a/media/filters/source_buffer_stream.cc b/media/filters/source_buffer_stream.cc
index 0ff0f3f0efb023e166248743947caa18f2af2b10..25bf138aae9d133956741e228073abb5dfb9ef16 100644
--- a/media/filters/source_buffer_stream.cc
+++ b/media/filters/source_buffer_stream.cc
@@ -447,6 +447,7 @@ bool SourceBufferStream::Append(const BufferQueue& buffers) {
 
   DCHECK(!buffers.empty());
   DCHECK(media_segment_start_time_ != kNoTimestamp());
+  DCHECK(media_segment_start_time_ <= buffers.front()->timestamp());

[acolwell GONE FROM CHROMIUM, 2014/04/03 00:26:52]

I was wrong. This should be using GetDecodeTimestamp() not timestamp(). I
believe this will prevent the failures in the SourceBufferStream unit tests.

[DaleCurtis, 2014/04/08 20:42:30]

Ah, good call, this fixed the issue.

   DCHECK(!end_of_stream_);
 
   // New media segments must begin with a keyframe.