media/formats/mp4/mp4_stream_parser.cc

   1 // Copyright 2014 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "media/formats/mp4/mp4_stream_parser.h"
   6
   7 #include "base/callback.h"
   8 #include "base/callback_helpers.h"
   9 #include "base/logging.h"
  10 #include "base/time/time.h"
  11 #include "media/base/audio_decoder_config.h"
  12 #include "media/base/stream_parser_buffer.h"
  13 #include "media/base/text_track_config.h"
  14 #include "media/base/video_decoder_config.h"
  15 #include "media/base/video_util.h"
  16 #include "media/formats/mp4/box_definitions.h"
  17 #include "media/formats/mp4/box_reader.h"
  18 #include "media/formats/mp4/es_descriptor.h"
  19 #include "media/formats/mp4/rcheck.h"
  20 #include "media/formats/mpeg/adts_constants.h"
  21
  22 namespace media {
  23 namespace mp4 {
  24
  25 MP4StreamParser::MP4StreamParser(const std::set<int>& audio_object_types,
  26                                  bool has_sbr)
  27     : state_(kWaitingForInit),
  28       moof_head_(0),
  29       mdat_tail_(0),
  30       highest_end_offset_(0),
  31       has_audio_(false),
  32       has_video_(false),
  33       audio_track_id_(0),
  34       video_track_id_(0),
  35       audio_object_types_(audio_object_types),
  36       has_sbr_(has_sbr),
  37       is_audio_track_encrypted_(false),
  38       is_video_track_encrypted_(false) {
  39 }
  40
  41 MP4StreamParser::~MP4StreamParser() {}
  42
  43 void MP4StreamParser::Init(
  44     const InitCB& init_cb,
  45     const NewConfigCB& config_cb,
  46     const NewBuffersCB& new_buffers_cb,
  47     bool /* ignore_text_tracks */,
  48     const EncryptedMediaInitDataCB& encrypted_media_init_data_cb,
  49     const NewMediaSegmentCB& new_segment_cb,
  50     const base::Closure& end_of_segment_cb,
  51     const LogCB& log_cb) {
  52   DCHECK_EQ(state_, kWaitingForInit);
  53   DCHECK(init_cb_.is_null());
  54   DCHECK(!init_cb.is_null());
  55   DCHECK(!config_cb.is_null());
  56   DCHECK(!new_buffers_cb.is_null());
  57   DCHECK(!encrypted_media_init_data_cb.is_null());
  58   DCHECK(!end_of_segment_cb.is_null());
  59
  60   ChangeState(kParsingBoxes);
  61   init_cb_ = init_cb;
  62   config_cb_ = config_cb;
  63   new_buffers_cb_ = new_buffers_cb;
  64   encrypted_media_init_data_cb_ = encrypted_media_init_data_cb;
  65   new_segment_cb_ = new_segment_cb;
  66   end_of_segment_cb_ = end_of_segment_cb;
  67   log_cb_ = log_cb;
  68 }
  69
  70 void MP4StreamParser::Reset() {
  71   queue_.Reset();
  72   runs_.reset();
  73   moof_head_ = 0;
  74   mdat_tail_ = 0;
  75 }
  76
  77 void MP4StreamParser::Flush() {
  78   DCHECK_NE(state_, kWaitingForInit);
  79   Reset();
  80   ChangeState(kParsingBoxes);
  81 }
  82
  83 bool MP4StreamParser::Parse(const uint8* buf, int size) {
  84   DCHECK_NE(state_, kWaitingForInit);
  85
  86   if (state_ == kError)
  87     return false;
  88
  89   queue_.Push(buf, size);
  90
  91   BufferQueue audio_buffers;
  92   BufferQueue video_buffers;
  93
  94   bool result = false;
  95   bool err = false;
  96
  97   do {
  98     switch (state_) {
  99       case kWaitingForInit:
 100       case kError:
 101         NOTREACHED();
 102         return false;
 103
 104       case kParsingBoxes:
 105         result = ParseBox(&err);
 106         break;
 107
 108       case kWaitingForSampleData:
 109         result = HaveEnoughDataToEnqueueSamples();
 110         if (result)
 111           ChangeState(kEmittingSamples);
 112         break;
 113
 114       case kEmittingSamples:
 115         result = EnqueueSample(&audio_buffers, &video_buffers, &err);
 116         if (result) {
 117           int64 max_clear = runs_->GetMaxClearOffset() + moof_head_;
 118           err = !ReadAndDiscardMDATsUntil(max_clear);
 119         }
 120         break;
 121     }
 122   } while (result && !err);
 123
 124   if (!err)
 125     err = !SendAndFlushSamples(&audio_buffers, &video_buffers);
 126
 127   if (err) {
 128     DLOG(ERROR) << "Error while parsing MP4";
 129     moov_.reset();
 130     Reset();
 131     ChangeState(kError);
 132     return false;
 133   }
 134
 135   return true;
 136 }
 137
 138 bool MP4StreamParser::ParseBox(bool* err) {
 139   const uint8* buf;
 140   int size;
 141   queue_.Peek(&buf, &size);
 142   if (!size) return false;
 143
 144   scoped_ptr<BoxReader> reader(
 145       BoxReader::ReadTopLevelBox(buf, size, log_cb_, err));
 146   if (reader.get() == NULL) return false;
 147
 148   if (reader->type() == FOURCC_MOOV) {
 149     *err = !ParseMoov(reader.get());
 150   } else if (reader->type() == FOURCC_MOOF) {
 151     moof_head_ = queue_.head();
 152     *err = !ParseMoof(reader.get());
 153
 154     // Set up first mdat offset for ReadMDATsUntil().
 155     mdat_tail_ = queue_.head() + reader->size();
 156
 157     // Return early to avoid evicting 'moof' data from queue. Auxiliary info may
 158     // be located anywhere in the file, including inside the 'moof' itself.
 159     // (Since 'default-base-is-moof' is mandated, no data references can come
 160     // before the head of the 'moof', so keeping this box around is sufficient.)
 161     return !(*err);
 162   } else {
 163     MEDIA_LOG(DEBUG, log_cb_) << "Skipping unrecognized top-level box: "
 164                               << FourCCToString(reader->type());
 165   }
 166
 167   queue_.Pop(reader->size());
 168   return !(*err);
 169 }
 170
 171
 172 bool MP4StreamParser::ParseMoov(BoxReader* reader) {
 173   moov_.reset(new Movie);
 174   RCHECK(moov_->Parse(reader));
 175   runs_.reset();
 176
 177   has_audio_ = false;
 178   has_video_ = false;
 179
 180   AudioDecoderConfig audio_config;
 181   VideoDecoderConfig video_config;
 182
 183   for (std::vector<Track>::const_iterator track = moov_->tracks.begin();
 184        track != moov_->tracks.end(); ++track) {
 185     // TODO(strobe): Only the first audio and video track present in a file are
 186     // used. (Track selection is better accomplished via Source IDs, though, so
 187     // adding support for track selection within a stream is low-priority.)
 188     const SampleDescription& samp_descr =
 189         track->media.information.sample_table.description;
 190
 191     // TODO(strobe): When codec reconfigurations are supported, detect and send
 192     // a codec reconfiguration for fragments using a sample description index
 193     // different from the previous one
 194     size_t desc_idx = 0;
 195     for (size_t t = 0; t < moov_->extends.tracks.size(); t++) {
 196       const TrackExtends& trex = moov_->extends.tracks[t];
 197       if (trex.track_id == track->header.track_id) {
 198         desc_idx = trex.default_sample_description_index;
 199         break;
 200       }
 201     }
 202     RCHECK(desc_idx > 0);
 203     desc_idx -= 1;  // BMFF descriptor index is one-based
 204
 205     if (track->media.handler.type == kAudio && !audio_config.IsValidConfig()) {
 206       RCHECK(!samp_descr.audio_entries.empty());
 207
 208       // It is not uncommon to find otherwise-valid files with incorrect sample
 209       // description indices, so we fail gracefully in that case.
 210       if (desc_idx >= samp_descr.audio_entries.size())
 211         desc_idx = 0;
 212       const AudioSampleEntry& entry = samp_descr.audio_entries[desc_idx];
 213       const AAC& aac = entry.esds.aac;
 214
 215       if (!(entry.format == FOURCC_MP4A ||
 216             (entry.format == FOURCC_ENCA &&
 217              entry.sinf.format.format == FOURCC_MP4A))) {
 218         MEDIA_LOG(ERROR, log_cb_) << "Unsupported audio format 0x" << std::hex
 219                                   << entry.format << " in stsd box.";
 220         return false;
 221       }
 222
 223       uint8 audio_type = entry.esds.object_type;
 224       DVLOG(1) << "audio_type " << std::hex << static_cast<int>(audio_type);
 225       if (audio_object_types_.find(audio_type) == audio_object_types_.end()) {
 226         MEDIA_LOG(ERROR, log_cb_) << "audio object type 0x" << std::hex
 227                                   << audio_type
 228                                   << " does not match what is specified in the"
 229                                   << " mimetype.";
 230         return false;
 231       }
 232
 233       AudioCodec codec = kUnknownAudioCodec;
 234       ChannelLayout channel_layout = CHANNEL_LAYOUT_NONE;
 235       int sample_per_second = 0;
 236       std::vector<uint8> extra_data;
 237       // Check if it is MPEG4 AAC defined in ISO 14496 Part 3 or
 238       // supported MPEG2 AAC varients.
 239       if (ESDescriptor::IsAAC(audio_type)) {
 240         codec = kCodecAAC;
 241         channel_layout = aac.GetChannelLayout(has_sbr_);
 242         sample_per_second = aac.GetOutputSamplesPerSecond(has_sbr_);
 243 #if defined(OS_ANDROID)
 244         extra_data = aac.codec_specific_data();
 245 #endif
 246       } else {
 247         MEDIA_LOG(ERROR, log_cb_) << "Unsupported audio object type 0x"
 248                                   << std::hex << audio_type << " in esds.";
 249         return false;
 250       }
 251
 252       SampleFormat sample_format;
 253       if (entry.samplesize == 8) {
 254         sample_format = kSampleFormatU8;
 255       } else if (entry.samplesize == 16) {
 256         sample_format = kSampleFormatS16;
 257       } else if (entry.samplesize == 32) {
 258         sample_format = kSampleFormatS32;
 259       } else {
 260         LOG(ERROR) << "Unsupported sample size.";
 261         return false;
 262       }
 263
 264       is_audio_track_encrypted_ = entry.sinf.info.track_encryption.is_encrypted;
 265       DVLOG(1) << "is_audio_track_encrypted_: " << is_audio_track_encrypted_;
 266       audio_config.Initialize(
 267           codec, sample_format, channel_layout, sample_per_second,
 268           extra_data.size() ? &extra_data[0] : NULL, extra_data.size(),
 269           is_audio_track_encrypted_, false, base::TimeDelta(),
 270           0);
 271       has_audio_ = true;
 272       audio_track_id_ = track->header.track_id;
 273     }
 274     if (track->media.handler.type == kVideo && !video_config.IsValidConfig()) {
 275       RCHECK(!samp_descr.video_entries.empty());
 276       if (desc_idx >= samp_descr.video_entries.size())
 277         desc_idx = 0;
 278       const VideoSampleEntry& entry = samp_descr.video_entries[desc_idx];
 279
 280       if (!entry.IsFormatValid()) {
 281         MEDIA_LOG(ERROR, log_cb_) << "Unsupported video format 0x" << std::hex
 282                                   << entry.format << " in stsd box.";
 283         return false;
 284       }
 285
 286       // TODO(strobe): Recover correct crop box
 287       gfx::Size coded_size(entry.width, entry.height);
 288       gfx::Rect visible_rect(coded_size);
 289       gfx::Size natural_size = GetNaturalSize(visible_rect.size(),
 290                                               entry.pixel_aspect.h_spacing,
 291                                               entry.pixel_aspect.v_spacing);
 292       is_video_track_encrypted_ = entry.sinf.info.track_encryption.is_encrypted;
 293       DVLOG(1) << "is_video_track_encrypted_: " << is_video_track_encrypted_;
 294       video_config.Initialize(kCodecH264, H264PROFILE_MAIN,  VideoFrame::YV12,
 295                               coded_size, visible_rect, natural_size,
 296                               // No decoder-specific buffer needed for AVC;
 297                               // SPS/PPS are embedded in the video stream
 298                               NULL, 0, is_video_track_encrypted_, false);
 299       has_video_ = true;
 300       video_track_id_ = track->header.track_id;
 301     }
 302   }
 303
 304   RCHECK(config_cb_.Run(audio_config, video_config, TextTrackConfigMap()));
 305
 306   StreamParser::InitParameters params(kInfiniteDuration());
 307   if (moov_->extends.header.fragment_duration > 0) {
 308     params.duration = TimeDeltaFromRational(
 309         moov_->extends.header.fragment_duration, moov_->header.timescale);
 310   } else if (moov_->header.duration > 0 &&
 311              moov_->header.duration != kuint64max) {
 312     params.duration =
 313         TimeDeltaFromRational(moov_->header.duration, moov_->header.timescale);
 314   }
 315
 316   if (!init_cb_.is_null())
 317     base::ResetAndReturn(&init_cb_).Run(params);
 318
 319   if (!moov_->pssh.empty())
 320     OnEncryptedMediaInitData(moov_->pssh);
 321
 322   return true;
 323 }
 324
 325 bool MP4StreamParser::ParseMoof(BoxReader* reader) {
 326   RCHECK(moov_.get());  // Must already have initialization segment
 327   MovieFragment moof;
 328   RCHECK(moof.Parse(reader));
 329   if (!runs_)
 330     runs_.reset(new TrackRunIterator(moov_.get(), log_cb_));
 331   RCHECK(runs_->Init(moof));
 332   RCHECK(ComputeHighestEndOffset(moof));
 333
 334   if (!moof.pssh.empty())
 335     OnEncryptedMediaInitData(moof.pssh);
 336
 337   new_segment_cb_.Run();
 338   ChangeState(kWaitingForSampleData);
 339   return true;
 340 }
 341
 342 void MP4StreamParser::OnEncryptedMediaInitData(
 343     const std::vector<ProtectionSystemSpecificHeader>& headers) {
 344   // TODO(strobe): ensure that the value of init_data (all PSSH headers
 345   // concatenated in arbitrary order) matches the EME spec.
 346   // See https://www.w3.org/Bugs/Public/show_bug.cgi?id=17673.
 347   size_t total_size = 0;
 348   for (size_t i = 0; i < headers.size(); i++)
 349     total_size += headers[i].raw_box.size();
 350
 351   std::vector<uint8> init_data(total_size);
 352   size_t pos = 0;
 353   for (size_t i = 0; i < headers.size(); i++) {
 354     memcpy(&init_data[pos], &headers[i].raw_box[0],
 355            headers[i].raw_box.size());
 356     pos += headers[i].raw_box.size();
 357   }
 358   encrypted_media_init_data_cb_.Run(EmeInitDataType::CENC, init_data);
 359 }
 360
 361 bool MP4StreamParser::PrepareAVCBuffer(
 362     const AVCDecoderConfigurationRecord& avc_config,
 363     std::vector<uint8>* frame_buf,
 364     std::vector<SubsampleEntry>* subsamples) const {
 365   // Convert the AVC NALU length fields to Annex B headers, as expected by
 366   // decoding libraries. Since this may enlarge the size of the buffer, we also
 367   // update the clear byte count for each subsample if encryption is used to
 368   // account for the difference in size between the length prefix and Annex B
 369   // start code.
 370   RCHECK(AVC::ConvertFrameToAnnexB(avc_config.length_size, frame_buf));
 371   if (!subsamples->empty()) {
 372     const int nalu_size_diff = 4 - avc_config.length_size;
 373     size_t expected_size = runs_->sample_size() +
 374         subsamples->size() * nalu_size_diff;
 375     RCHECK(frame_buf->size() == expected_size);
 376     for (size_t i = 0; i < subsamples->size(); i++)
 377       (*subsamples)[i].clear_bytes += nalu_size_diff;
 378   }
 379
 380   if (runs_->is_keyframe()) {
 381     // If this is a keyframe, we (re-)inject SPS and PPS headers at the start of
 382     // a frame. If subsample info is present, we also update the clear byte
 383     // count for that first subsample.
 384     RCHECK(AVC::InsertParamSetsAnnexB(avc_config, frame_buf, subsamples));
 385   }
 386
 387   DCHECK(AVC::IsValidAnnexB(*frame_buf, *subsamples));
 388   return true;
 389 }
 390
 391 bool MP4StreamParser::PrepareAACBuffer(
 392     const AAC& aac_config, std::vector<uint8>* frame_buf,
 393     std::vector<SubsampleEntry>* subsamples) const {
 394   // Append an ADTS header to every audio sample.
 395   RCHECK(aac_config.ConvertEsdsToADTS(frame_buf));
 396
 397   // As above, adjust subsample information to account for the headers. AAC is
 398   // not required to use subsample encryption, so we may need to add an entry.
 399   if (subsamples->empty()) {
 400     subsamples->push_back(SubsampleEntry(
 401         kADTSHeaderMinSize, frame_buf->size() - kADTSHeaderMinSize));
 402   } else {
 403     (*subsamples)[0].clear_bytes += kADTSHeaderMinSize;
 404   }
 405   return true;
 406 }
 407
 408 bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
 409                                     BufferQueue* video_buffers,
 410                                     bool* err) {
 411   DCHECK_EQ(state_, kEmittingSamples);
 412
 413   if (!runs_->IsRunValid()) {
 414     // Flush any buffers we've gotten in this chunk so that buffers don't
 415     // cross NewSegment() calls
 416     *err = !SendAndFlushSamples(audio_buffers, video_buffers);
 417     if (*err)
 418       return false;
 419
 420     // Remain in kEmittingSamples state, discarding data, until the end of
 421     // the current 'mdat' box has been appended to the queue.
 422     if (!queue_.Trim(mdat_tail_))
 423       return false;
 424
 425     ChangeState(kParsingBoxes);
 426     end_of_segment_cb_.Run();
 427     return true;
 428   }
 429
 430   if (!runs_->IsSampleValid()) {
 431     runs_->AdvanceRun();
 432     return true;
 433   }
 434
 435   DCHECK(!(*err));
 436
 437   const uint8* buf;
 438   int buf_size;
 439   queue_.Peek(&buf, &buf_size);
 440   if (!buf_size) return false;
 441
 442   bool audio = has_audio_ && audio_track_id_ == runs_->track_id();
 443   bool video = has_video_ && video_track_id_ == runs_->track_id();
 444
 445   // Skip this entire track if it's not one we're interested in
 446   if (!audio && !video) {
 447     runs_->AdvanceRun();
 448     return true;
 449   }
 450
 451   // Attempt to cache the auxiliary information first. Aux info is usually
 452   // placed in a contiguous block before the sample data, rather than being
 453   // interleaved. If we didn't cache it, this would require that we retain the
 454   // start of the segment buffer while reading samples. Aux info is typically
 455   // quite small compared to sample data, so this pattern is useful on
 456   // memory-constrained devices where the source buffer consumes a substantial
 457   // portion of the total system memory.
 458   if (runs_->AuxInfoNeedsToBeCached()) {
 459     queue_.PeekAt(runs_->aux_info_offset() + moof_head_, &buf, &buf_size);
 460     if (buf_size < runs_->aux_info_size()) return false;
 461     *err = !runs_->CacheAuxInfo(buf, buf_size);
 462     return !*err;
 463   }
 464
 465   queue_.PeekAt(runs_->sample_offset() + moof_head_, &buf, &buf_size);
 466   if (buf_size < runs_->sample_size()) return false;
 467
 468   scoped_ptr<DecryptConfig> decrypt_config;
 469   std::vector<SubsampleEntry> subsamples;
 470   if (runs_->is_encrypted()) {
 471     decrypt_config = runs_->GetDecryptConfig();
 472     if (!decrypt_config) {
 473       *err = true;
 474       return false;
 475     }
 476     subsamples = decrypt_config->subsamples();
 477   }
 478
 479   std::vector<uint8> frame_buf(buf, buf + runs_->sample_size());
 480   if (video) {
 481     if (!PrepareAVCBuffer(runs_->video_description().avcc,
 482                           &frame_buf, &subsamples)) {
 483       MEDIA_LOG(ERROR, log_cb_) << "Failed to prepare AVC sample for decode";
 484       *err = true;
 485       return false;
 486     }
 487   }
 488
 489   if (audio) {
 490     if (ESDescriptor::IsAAC(runs_->audio_description().esds.object_type) &&
 491         !PrepareAACBuffer(runs_->audio_description().esds.aac,
 492                           &frame_buf, &subsamples)) {
 493       MEDIA_LOG(ERROR, log_cb_) << "Failed to prepare AAC sample for decode";
 494       *err = true;
 495       return false;
 496     }
 497   }
 498
 499   if (decrypt_config) {
 500     if (!subsamples.empty()) {
 501     // Create a new config with the updated subsamples.
 502     decrypt_config.reset(new DecryptConfig(
 503         decrypt_config->key_id(),
 504         decrypt_config->iv(),
 505         subsamples));
 506     }
 507     // else, use the existing config.
 508   } else if ((audio && is_audio_track_encrypted_) ||
 509              (video && is_video_track_encrypted_)) {
 510     // The media pipeline requires a DecryptConfig with an empty |iv|.
 511     // TODO(ddorwin): Refactor so we do not need a fake key ID ("1");
 512     decrypt_config.reset(
 513         new DecryptConfig("1", "", std::vector<SubsampleEntry>()));
 514   }
 515
 516   StreamParserBuffer::Type buffer_type = audio ? DemuxerStream::AUDIO :
 517       DemuxerStream::VIDEO;
 518
 519   // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 520   // type and allow multiple tracks for same media type, if applicable. See
 521   // https://crbug.com/341581.
 522   //
 523   // NOTE: MPEG's "random access point" concept is equivalent to the
 524   // downstream code's "is keyframe" concept.
 525   scoped_refptr<StreamParserBuffer> stream_buf =
 526       StreamParserBuffer::CopyFrom(&frame_buf[0], frame_buf.size(),
 527                                    runs_->is_random_access_point(),
 528                                    buffer_type, 0);
 529
 530   if (decrypt_config)
 531     stream_buf->set_decrypt_config(decrypt_config.Pass());
 532
 533   stream_buf->set_duration(runs_->duration());
 534   stream_buf->set_timestamp(runs_->cts());
 535   stream_buf->SetDecodeTimestamp(runs_->dts());
 536
 537   DVLOG(3) << "Pushing frame: aud=" << audio
 538            << ", key=" << runs_->is_keyframe()
 539            << ", rap=" << runs_->is_random_access_point()
 540            << ", dur=" << runs_->duration().InMilliseconds()
 541            << ", dts=" << runs_->dts().InMilliseconds()
 542            << ", cts=" << runs_->cts().InMilliseconds()
 543            << ", size=" << runs_->sample_size();
 544
 545   if (audio) {
 546     audio_buffers->push_back(stream_buf);
 547   } else {
 548     video_buffers->push_back(stream_buf);
 549   }
 550
 551   runs_->AdvanceSample();
 552   return true;
 553 }
 554
 555 bool MP4StreamParser::SendAndFlushSamples(BufferQueue* audio_buffers,
 556                                           BufferQueue* video_buffers) {
 557   if (audio_buffers->empty() && video_buffers->empty())
 558     return true;
 559
 560   TextBufferQueueMap empty_text_map;
 561   bool success = new_buffers_cb_.Run(*audio_buffers,
 562                                      *video_buffers,
 563                                      empty_text_map);
 564   audio_buffers->clear();
 565   video_buffers->clear();
 566   return success;
 567 }
 568
 569 bool MP4StreamParser::ReadAndDiscardMDATsUntil(int64 max_clear_offset) {
 570   bool err = false;
 571   int64 upper_bound = std::min(max_clear_offset, queue_.tail());
 572   while (mdat_tail_ < upper_bound) {
 573     const uint8* buf = NULL;
 574     int size = 0;
 575     queue_.PeekAt(mdat_tail_, &buf, &size);
 576
 577     FourCC type;
 578     int box_sz;
 579     if (!BoxReader::StartTopLevelBox(buf, size, log_cb_,
 580                                      &type, &box_sz, &err))
 581       break;
 582
 583     if (type != FOURCC_MDAT) {
 584       MEDIA_LOG(DEBUG, log_cb_) << "Unexpected box type while parsing MDATs: "
 585                                 << FourCCToString(type);
 586     }
 587     mdat_tail_ += box_sz;
 588   }
 589   queue_.Trim(std::min(mdat_tail_, upper_bound));
 590   return !err;
 591 }
 592
 593 void MP4StreamParser::ChangeState(State new_state) {
 594   DVLOG(2) << "Changing state: " << new_state;
 595   state_ = new_state;
 596 }
 597
 598 bool MP4StreamParser::HaveEnoughDataToEnqueueSamples() {
 599   DCHECK_EQ(state_, kWaitingForSampleData);
 600   // For muxed content, make sure we have data up to |highest_end_offset_|
 601   // so we can ensure proper enqueuing behavior. Otherwise assume we have enough
 602   // data and allow per sample offset checks to meter sample enqueuing.
 603   // TODO(acolwell): Fix trun box handling so we don't have to special case
 604   // muxed content.
 605   return !(has_audio_ && has_video_ &&
 606            queue_.tail() < highest_end_offset_ + moof_head_);
 607 }
 608
 609 bool MP4StreamParser::ComputeHighestEndOffset(const MovieFragment& moof) {
 610   highest_end_offset_ = 0;
 611
 612   TrackRunIterator runs(moov_.get(), log_cb_);
 613   RCHECK(runs.Init(moof));
 614
 615   while (runs.IsRunValid()) {
 616     int64 aux_info_end_offset = runs.aux_info_offset() + runs.aux_info_size();
 617     if (aux_info_end_offset > highest_end_offset_)
 618       highest_end_offset_ = aux_info_end_offset;
 619
 620     while (runs.IsSampleValid()) {
 621       int64 sample_end_offset = runs.sample_offset() + runs.sample_size();
 622       if (sample_end_offset > highest_end_offset_)
 623         highest_end_offset_ = sample_end_offset;
 624
 625       runs.AdvanceSample();
 626     }
 627     runs.AdvanceRun();
 628   }
 629
 630   return true;
 631 }
 632
 633 }  // namespace mp4
 634 }  // namespace media