Support multiple media tracks in MSE / ChunkDemuxer

media/filters/media_source_state.h

 // Copyright (c) 2016 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef MEDIA_FILTERS_MEDIA_SOURCE_STATE_H_
 #define MEDIA_FILTERS_MEDIA_SOURCE_STATE_H_
 
 #include <list>
 
 #include "base/bind.h"
 #include "base/macros.h"
+#include "media/base/audio_codecs.h"
 #include "media/base/demuxer.h"
 #include "media/base/demuxer_stream.h"
 #include "media/base/media_export.h"
 #include "media/base/media_log.h"
 #include "media/base/stream_parser.h"
 #include "media/base/stream_parser_buffer.h"
+#include "media/base/video_codecs.h"
 
 namespace media {
 
 using base::TimeDelta;
 
 class ChunkDemuxerStream;
 class FrameProcessor;
 
 // Contains state belonging to a source id.
 class MEDIA_EXPORT MediaSourceState {
  public:
   // Callback signature used to create ChunkDemuxerStreams.
   typedef base::Callback<ChunkDemuxerStream*(DemuxerStream::Type)>
       CreateDemuxerStreamCB;
 
   typedef base::Callback<void(ChunkDemuxerStream*, const TextTrackConfig&)>
       NewTextTrackCB;
 
   MediaSourceState(std::unique_ptr<StreamParser> stream_parser,
                    std::unique_ptr<FrameProcessor> frame_processor,
                    const CreateDemuxerStreamCB& create_demuxer_stream_cb,
                    const scoped_refptr<MediaLog>& media_log);
 
   ~MediaSourceState();
 
   void Init(const StreamParser::InitCB& init_cb,
-            bool allow_audio,
-            bool allow_video,
+            const std::string& expected_codecs,
             const StreamParser::EncryptedMediaInitDataCB&
                 encrypted_media_init_data_cb,
             const NewTextTrackCB& new_text_track_cb);
 
   // Appends new data to the StreamParser.
   // Returns true if the data was successfully appended. Returns false if an
   // error occurred. |*timestamp_offset| is used and possibly updated by the
   // append. |append_window_start| and |append_window_end| correspond to the MSE
   // spec's similarly named source buffer attributes that are used in coded
   // frame processing.
@@ skipping 73 matching lines @@
     UNINITIALIZED = 0,
     PENDING_PARSER_CONFIG,
     PENDING_PARSER_INIT,
     PARSER_INITIALIZED
   };
 
   // Called by the |stream_parser_| when a new initialization segment is
   // encountered.
   // Returns true on a successful call. Returns false if an error occurred while
   // processing decoder configurations.
-  bool OnNewConfigs(bool allow_audio,
-                    bool allow_video,
+  bool OnNewConfigs(std::string expected_codecs,
                     std::unique_ptr<MediaTracks> tracks,
                     const StreamParser::TextTrackConfigMap& text_configs);
 
   // Called by the |stream_parser_| at the beginning of a new media segment.
   void OnNewMediaSegment();
 
   // Called by the |stream_parser_| at the end of a media segment.
   void OnEndOfMediaSegment();
 
   // Called by the |stream_parser_| when new buffers have been parsed.
   // It processes the new buffers using |frame_processor_|, which includes
   // appending the processed frames to associated demuxer streams for each
   // frame's track.
   // Returns true on a successful call. Returns false if an error occurred while
   // processing the buffers.
   bool OnNewBuffers(const StreamParser::BufferQueueMap& buffer_queue_map);
 
   void OnSourceInitDone(const StreamParser::InitParameters& params);
 
-  // EstimateVideoDataSize uses some heuristics to estimate the size of the
-  // video size in the chunk of muxed audio/video data without parsing it.
-  // This is used by EvictCodedFrames algorithm, which happens before Append
-  // (and therefore before parsing is performed) to prepare space for new data.
-  size_t EstimateVideoDataSize(size_t muxed_data_chunk_size) const;
+  // Sets memory limits for all demuxer streams.
+  void SetStreamMemoryLimits();
 
   // Tracks the number of MEDIA_LOGs emitted for segments missing expected audio
   // or video blocks. Useful to prevent log spam.
   int num_missing_track_logs_ = 0;
 
   CreateDemuxerStreamCB create_demuxer_stream_cb_;
   NewTextTrackCB new_text_track_cb_;
 
   // During Append(), if OnNewBuffers() coded frame processing updates the
   // timestamp offset then |*timestamp_offset_during_append_| is also updated
   // so Append()'s caller can know the new offset. This pointer is only non-NULL
   // during the lifetime of an Append() call.
   TimeDelta* timestamp_offset_during_append_;
 
   // During Append(), coded frame processing triggered by OnNewBuffers()
   // requires these two attributes. These are only valid during the lifetime of
   // an Append() call.
   TimeDelta append_window_start_during_append_;
   TimeDelta append_window_end_during_append_;
 
   // Keeps track of whether a media segment is being parsed.
   bool parsing_media_segment_;
 
   // Valid only while |parsing_media_segment_| is true. These flags enable
-  // warning when at least one frame for each A/V track is not in a parsed media
-  // segment.
-  bool media_segment_contained_audio_frame_;
-  bool media_segment_contained_video_frame_;
+  // warning when the parsed media segment doesn't have frames for some track.
+  std::map<StreamParser::TrackId, bool> media_segment_has_data_for_track_;
 
   // The object used to parse appended data.
   std::unique_ptr<StreamParser> stream_parser_;
 
-  ChunkDemuxerStream* audio_;  // Not owned by |this|.
-  ChunkDemuxerStream* video_;  // Not owned by |this|.
+  // Note that ChunkDemuxerStreams are created and owned by the parent
+  // ChunkDemuxer. They are not owned by |this|.
+  using DemuxerStreamMap = std::map<StreamParser::TrackId, ChunkDemuxerStream*>;
+  DemuxerStreamMap audio_streams_;
+  DemuxerStreamMap video_streams_;
 
   typedef std::map<StreamParser::TrackId, ChunkDemuxerStream*> TextStreamMap;
   TextStreamMap text_stream_map_;  // |this| owns the map's stream pointers.
 
   std::unique_ptr<FrameProcessor> frame_processor_;
   scoped_refptr<MediaLog> media_log_;
   StreamParser::InitCB init_cb_;
 
   State state_;
 
   // During Append(), OnNewConfigs() will trigger the initialization segment
   // received algorithm. Note, the MSE spec explicitly disallows this algorithm
   // during an Abort(), since Abort() is allowed only to emit coded frames, and
   // only if the parser is PARSING_MEDIA_SEGMENT (not an INIT segment). So we
   // also have a flag here that indicates if Append is in progress and we can
   // invoke this callback.
   Demuxer::MediaTracksUpdatedCB init_segment_received_cb_;
   bool append_in_progress_ = false;
+  bool first_init_segment_received_ = false;
+
+  std::vector<AudioCodec> expected_audio_codecs_;
+  std::vector<VideoCodec> expected_video_codecs_;
 
   // Indicates that timestampOffset should be updated automatically during
   // OnNewBuffers() based on the earliest end timestamp of the buffers provided.
   // TODO(wolenetz): Refactor this function while integrating April 29, 2014
   // changes to MSE spec. See http://crbug.com/371499.
   bool auto_update_timestamp_offset_;
 
   DISALLOW_COPY_AND_ASSIGN(MediaSourceState);
 };
 
 }  // namespace media
 
 #endif  // MEDIA_FILTERS_MEDIA_SOURCE_STATE_H_