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_helpers.h"
   8 #include "base/logging.h"
   9 #include "base/time/time.h"
  10 #include "media/base/audio_decoder_config.h"
  11 #include "media/base/stream_parser_buffer.h"
  12 #include "media/base/text_track_config.h"
  13 #include "media/base/video_decoder_config.h"
  14 #include "media/base/video_util.h"
  15 #include "media/formats/mp4/box_definitions.h"
  16 #include "media/formats/mp4/box_reader.h"
  17 #include "media/formats/mp4/es_descriptor.h"
  18 #include "media/formats/mp4/rcheck.h"
  19 #include "media/formats/mpeg/adts_constants.h"
  20
  21 namespace media {
  22 namespace mp4 {
  23
  24 MP4StreamParser::MP4StreamParser(const std::set<int>& audio_object_types,
  25                                  bool has_sbr)
  26     : state_(kWaitingForInit),
  27       moof_head_(0),
  28       mdat_tail_(0),
  29       highest_end_offset_(0),
  30       has_audio_(false),
  31       has_video_(false),
  32       audio_track_id_(0),
  33       video_track_id_(0),
  34       audio_object_types_(audio_object_types),
  35       has_sbr_(has_sbr),
  36       is_audio_track_encrypted_(false),
  37       is_video_track_encrypted_(false),
  38       num_top_level_box_skipped_(0) {
  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 scoped_refptr<MediaLog>& media_log) {
  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   media_log_ = media_log;
  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, media_log_, 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     // TODO(wolenetz,chcunningham): Enforce more strict adherence to MSE byte
 164     // stream spec for ftyp and styp. See http://crbug.com/504514.
 165     DVLOG(2) << "Skipping unrecognized top-level box: "
 166              << FourCCToString(reader->type());
 167   }
 168
 169   queue_.Pop(reader->size());
 170   return !(*err);
 171 }
 172
 173 bool MP4StreamParser::ParseMoov(BoxReader* reader) {
 174   moov_.reset(new Movie);
 175   RCHECK(moov_->Parse(reader));
 176   runs_.reset();
 177
 178   has_audio_ = false;
 179   has_video_ = false;
 180
 181   AudioDecoderConfig audio_config;
 182   VideoDecoderConfig video_config;
 183
 184   for (std::vector<Track>::const_iterator track = moov_->tracks.begin();
 185        track != moov_->tracks.end(); ++track) {
 186     // TODO(strobe): Only the first audio and video track present in a file are
 187     // used. (Track selection is better accomplished via Source IDs, though, so
 188     // adding support for track selection within a stream is low-priority.)
 189     const SampleDescription& samp_descr =
 190         track->media.information.sample_table.description;
 191
 192     // TODO(strobe): When codec reconfigurations are supported, detect and send
 193     // a codec reconfiguration for fragments using a sample description index
 194     // different from the previous one
 195     size_t desc_idx = 0;
 196     for (size_t t = 0; t < moov_->extends.tracks.size(); t++) {
 197       const TrackExtends& trex = moov_->extends.tracks[t];
 198       if (trex.track_id == track->header.track_id) {
 199         desc_idx = trex.default_sample_description_index;
 200         break;
 201       }
 202     }
 203     RCHECK(desc_idx > 0);
 204     desc_idx -= 1;  // BMFF descriptor index is one-based
 205
 206     if (track->media.handler.type == kAudio && !audio_config.IsValidConfig()) {
 207       RCHECK(!samp_descr.audio_entries.empty());
 208
 209       // It is not uncommon to find otherwise-valid files with incorrect sample
 210       // description indices, so we fail gracefully in that case.
 211       if (desc_idx >= samp_descr.audio_entries.size())
 212         desc_idx = 0;
 213       const AudioSampleEntry& entry = samp_descr.audio_entries[desc_idx];
 214       const AAC& aac = entry.esds.aac;
 215
 216       if (!(entry.format == FOURCC_MP4A ||
 217             (entry.format == FOURCC_ENCA &&
 218              entry.sinf.format.format == FOURCC_MP4A))) {
 219         MEDIA_LOG(ERROR, media_log_) << "Unsupported audio format 0x"
 220                                      << std::hex << entry.format
 221                                      << " in stsd box.";
 222         return false;
 223       }
 224
 225       uint8 audio_type = entry.esds.object_type;
 226       DVLOG(1) << "audio_type " << std::hex << static_cast<int>(audio_type);
 227       if (audio_object_types_.find(audio_type) == audio_object_types_.end()) {
 228         MEDIA_LOG(ERROR, media_log_)
 229             << "audio object type 0x" << std::hex << audio_type
 230             << " does not match what is specified in the"
 231             << " mimetype.";
 232         return false;
 233       }
 234
 235       AudioCodec codec = kUnknownAudioCodec;
 236       ChannelLayout channel_layout = CHANNEL_LAYOUT_NONE;
 237       int sample_per_second = 0;
 238       std::vector<uint8> extra_data;
 239       // Check if it is MPEG4 AAC defined in ISO 14496 Part 3 or
 240       // supported MPEG2 AAC varients.
 241       if (ESDescriptor::IsAAC(audio_type)) {
 242         codec = kCodecAAC;
 243         channel_layout = aac.GetChannelLayout(has_sbr_);
 244         sample_per_second = aac.GetOutputSamplesPerSecond(has_sbr_);
 245 #if defined(OS_ANDROID)
 246         extra_data = aac.codec_specific_data();
 247 #endif
 248       } else {
 249         MEDIA_LOG(ERROR, media_log_) << "Unsupported audio object type 0x"
 250                                      << std::hex << audio_type << " in esds.";
 251         return false;
 252       }
 253
 254       SampleFormat sample_format;
 255       if (entry.samplesize == 8) {
 256         sample_format = kSampleFormatU8;
 257       } else if (entry.samplesize == 16) {
 258         sample_format = kSampleFormatS16;
 259       } else if (entry.samplesize == 32) {
 260         sample_format = kSampleFormatS32;
 261       } else {
 262         LOG(ERROR) << "Unsupported sample size.";
 263         return false;
 264       }
 265
 266       is_audio_track_encrypted_ = entry.sinf.info.track_encryption.is_encrypted;
 267       DVLOG(1) << "is_audio_track_encrypted_: " << is_audio_track_encrypted_;
 268       audio_config.Initialize(
 269           codec, sample_format, channel_layout, sample_per_second,
 270           extra_data.size() ? &extra_data[0] : NULL, extra_data.size(),
 271           is_audio_track_encrypted_, false, base::TimeDelta(),
 272           0);
 273       has_audio_ = true;
 274       audio_track_id_ = track->header.track_id;
 275     }
 276     if (track->media.handler.type == kVideo && !video_config.IsValidConfig()) {
 277       RCHECK(!samp_descr.video_entries.empty());
 278       if (desc_idx >= samp_descr.video_entries.size())
 279         desc_idx = 0;
 280       const VideoSampleEntry& entry = samp_descr.video_entries[desc_idx];
 281
 282       if (!entry.IsFormatValid()) {
 283         MEDIA_LOG(ERROR, media_log_) << "Unsupported video format 0x"
 284                                      << std::hex << entry.format
 285                                      << " in stsd box.";
 286         return false;
 287       }
 288
 289       // TODO(strobe): Recover correct crop box
 290       gfx::Size coded_size(entry.width, entry.height);
 291       gfx::Rect visible_rect(coded_size);
 292
 293       // If PASP is available, use the coded size and PASP to calculate the
 294       // natural size. Otherwise, use the size in track header for natural size.
 295       gfx::Size natural_size(visible_rect.size());
 296       if (entry.pixel_aspect.h_spacing != 1 ||
 297           entry.pixel_aspect.v_spacing != 1) {
 298         natural_size =
 299             GetNaturalSize(visible_rect.size(), entry.pixel_aspect.h_spacing,
 300                            entry.pixel_aspect.v_spacing);
 301       } else if (track->header.width && track->header.height) {
 302         // An even width makes things easier for YV12 and appears to be the
 303         // behavior expected by WebKit layout tests. See GetNaturalSize().
 304         natural_size =
 305             gfx::Size(track->header.width & ~1, track->header.height);
 306       }
 307
 308       is_video_track_encrypted_ = entry.sinf.info.track_encryption.is_encrypted;
 309       DVLOG(1) << "is_video_track_encrypted_: " << is_video_track_encrypted_;
 310       video_config.Initialize(kCodecH264, H264PROFILE_MAIN, PIXEL_FORMAT_YV12,
 311                               COLOR_SPACE_HD_REC709, coded_size, visible_rect,
 312                               natural_size,
 313                               // No decoder-specific buffer needed for AVC;
 314                               // SPS/PPS are embedded in the video stream
 315                               NULL, 0, is_video_track_encrypted_, false);
 316       has_video_ = true;
 317       video_track_id_ = track->header.track_id;
 318     }
 319   }
 320
 321   RCHECK(config_cb_.Run(audio_config, video_config, TextTrackConfigMap()));
 322
 323   StreamParser::InitParameters params(kInfiniteDuration());
 324   if (moov_->extends.header.fragment_duration > 0) {
 325     params.duration = TimeDeltaFromRational(
 326         moov_->extends.header.fragment_duration, moov_->header.timescale);
 327     params.liveness = DemuxerStream::LIVENESS_RECORDED;
 328   } else if (moov_->header.duration > 0 &&
 329              moov_->header.duration != kuint64max) {
 330     params.duration =
 331         TimeDeltaFromRational(moov_->header.duration, moov_->header.timescale);
 332     params.liveness = DemuxerStream::LIVENESS_RECORDED;
 333   } else {
 334     // In ISO/IEC 14496-12:2005(E), 8.30.2: ".. If an MP4 file is created in
 335     // real-time, such as used in live streaming, it is not likely that the
 336     // fragment_duration is known in advance and this (mehd) box may be
 337     // omitted."
 338     // TODO(wolenetz): Investigate gating liveness detection on timeline_offset
 339     // when it's populated. See http://crbug.com/312699
 340     params.liveness = DemuxerStream::LIVENESS_LIVE;
 341   }
 342
 343   DVLOG(1) << "liveness: " << params.liveness;
 344
 345   if (!init_cb_.is_null())
 346     base::ResetAndReturn(&init_cb_).Run(params);
 347
 348   if (!moov_->pssh.empty())
 349     OnEncryptedMediaInitData(moov_->pssh);
 350
 351   return true;
 352 }
 353
 354 bool MP4StreamParser::ParseMoof(BoxReader* reader) {
 355   RCHECK(moov_.get());  // Must already have initialization segment
 356   MovieFragment moof;
 357   RCHECK(moof.Parse(reader));
 358   if (!runs_)
 359     runs_.reset(new TrackRunIterator(moov_.get(), media_log_));
 360   RCHECK(runs_->Init(moof));
 361   RCHECK(ComputeHighestEndOffset(moof));
 362
 363   if (!moof.pssh.empty())
 364     OnEncryptedMediaInitData(moof.pssh);
 365
 366   new_segment_cb_.Run();
 367   ChangeState(kWaitingForSampleData);
 368   return true;
 369 }
 370
 371 void MP4StreamParser::OnEncryptedMediaInitData(
 372     const std::vector<ProtectionSystemSpecificHeader>& headers) {
 373   // TODO(strobe): ensure that the value of init_data (all PSSH headers
 374   // concatenated in arbitrary order) matches the EME spec.
 375   // See https://www.w3.org/Bugs/Public/show_bug.cgi?id=17673.
 376   size_t total_size = 0;
 377   for (size_t i = 0; i < headers.size(); i++)
 378     total_size += headers[i].raw_box.size();
 379
 380   std::vector<uint8> init_data(total_size);
 381   size_t pos = 0;
 382   for (size_t i = 0; i < headers.size(); i++) {
 383     memcpy(&init_data[pos], &headers[i].raw_box[0],
 384            headers[i].raw_box.size());
 385     pos += headers[i].raw_box.size();
 386   }
 387   encrypted_media_init_data_cb_.Run(EmeInitDataType::CENC, init_data);
 388 }
 389
 390 bool MP4StreamParser::PrepareAVCBuffer(
 391     const AVCDecoderConfigurationRecord& avc_config,
 392     std::vector<uint8>* frame_buf,
 393     std::vector<SubsampleEntry>* subsamples) const {
 394   // Convert the AVC NALU length fields to Annex B headers, as expected by
 395   // decoding libraries. Since this may enlarge the size of the buffer, we also
 396   // update the clear byte count for each subsample if encryption is used to
 397   // account for the difference in size between the length prefix and Annex B
 398   // start code.
 399   RCHECK(AVC::ConvertFrameToAnnexB(avc_config.length_size, frame_buf));
 400   if (!subsamples->empty()) {
 401     const int nalu_size_diff = 4 - avc_config.length_size;
 402     size_t expected_size = runs_->sample_size() +
 403         subsamples->size() * nalu_size_diff;
 404     RCHECK(frame_buf->size() == expected_size);
 405     for (size_t i = 0; i < subsamples->size(); i++)
 406       (*subsamples)[i].clear_bytes += nalu_size_diff;
 407   }
 408
 409   if (runs_->is_keyframe()) {
 410     // If this is a keyframe, we (re-)inject SPS and PPS headers at the start of
 411     // a frame. If subsample info is present, we also update the clear byte
 412     // count for that first subsample.
 413     RCHECK(AVC::InsertParamSetsAnnexB(avc_config, frame_buf, subsamples));
 414   }
 415
 416   DCHECK(AVC::IsValidAnnexB(*frame_buf, *subsamples));
 417   return true;
 418 }
 419
 420 bool MP4StreamParser::PrepareAACBuffer(
 421     const AAC& aac_config, std::vector<uint8>* frame_buf,
 422     std::vector<SubsampleEntry>* subsamples) const {
 423   // Append an ADTS header to every audio sample.
 424   RCHECK(aac_config.ConvertEsdsToADTS(frame_buf));
 425
 426   // As above, adjust subsample information to account for the headers. AAC is
 427   // not required to use subsample encryption, so we may need to add an entry.
 428   if (subsamples->empty()) {
 429     subsamples->push_back(SubsampleEntry(
 430         kADTSHeaderMinSize, frame_buf->size() - kADTSHeaderMinSize));
 431   } else {
 432     (*subsamples)[0].clear_bytes += kADTSHeaderMinSize;
 433   }
 434   return true;
 435 }
 436
 437 bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
 438                                     BufferQueue* video_buffers,
 439                                     bool* err) {
 440   DCHECK_EQ(state_, kEmittingSamples);
 441
 442   if (!runs_->IsRunValid()) {
 443     // Flush any buffers we've gotten in this chunk so that buffers don't
 444     // cross NewSegment() calls
 445     *err = !SendAndFlushSamples(audio_buffers, video_buffers);
 446     if (*err)
 447       return false;
 448
 449     // Remain in kEmittingSamples state, discarding data, until the end of
 450     // the current 'mdat' box has been appended to the queue.
 451     if (!queue_.Trim(mdat_tail_))
 452       return false;
 453
 454     ChangeState(kParsingBoxes);
 455     end_of_segment_cb_.Run();
 456     return true;
 457   }
 458
 459   if (!runs_->IsSampleValid()) {
 460     runs_->AdvanceRun();
 461     return true;
 462   }
 463
 464   DCHECK(!(*err));
 465
 466   const uint8* buf;
 467   int buf_size;
 468   queue_.Peek(&buf, &buf_size);
 469   if (!buf_size) return false;
 470
 471   bool audio = has_audio_ && audio_track_id_ == runs_->track_id();
 472   bool video = has_video_ && video_track_id_ == runs_->track_id();
 473
 474   // Skip this entire track if it's not one we're interested in
 475   if (!audio && !video) {
 476     runs_->AdvanceRun();
 477     return true;
 478   }
 479
 480   // Attempt to cache the auxiliary information first. Aux info is usually
 481   // placed in a contiguous block before the sample data, rather than being
 482   // interleaved. If we didn't cache it, this would require that we retain the
 483   // start of the segment buffer while reading samples. Aux info is typically
 484   // quite small compared to sample data, so this pattern is useful on
 485   // memory-constrained devices where the source buffer consumes a substantial
 486   // portion of the total system memory.
 487   if (runs_->AuxInfoNeedsToBeCached()) {
 488     queue_.PeekAt(runs_->aux_info_offset() + moof_head_, &buf, &buf_size);
 489     if (buf_size < runs_->aux_info_size()) return false;
 490     *err = !runs_->CacheAuxInfo(buf, buf_size);
 491     return !*err;
 492   }
 493
 494   queue_.PeekAt(runs_->sample_offset() + moof_head_, &buf, &buf_size);
 495   if (buf_size < runs_->sample_size()) return false;
 496
 497   scoped_ptr<DecryptConfig> decrypt_config;
 498   std::vector<SubsampleEntry> subsamples;
 499   if (runs_->is_encrypted()) {
 500     decrypt_config = runs_->GetDecryptConfig();
 501     if (!decrypt_config) {
 502       *err = true;
 503       return false;
 504     }
 505     subsamples = decrypt_config->subsamples();
 506   }
 507
 508   std::vector<uint8> frame_buf(buf, buf + runs_->sample_size());
 509   if (video) {
 510     if (!PrepareAVCBuffer(runs_->video_description().avcc,
 511                           &frame_buf, &subsamples)) {
 512       MEDIA_LOG(ERROR, media_log_) << "Failed to prepare AVC sample for decode";
 513       *err = true;
 514       return false;
 515     }
 516   }
 517
 518   if (audio) {
 519     if (ESDescriptor::IsAAC(runs_->audio_description().esds.object_type) &&
 520         !PrepareAACBuffer(runs_->audio_description().esds.aac,
 521                           &frame_buf, &subsamples)) {
 522       MEDIA_LOG(ERROR, media_log_) << "Failed to prepare AAC sample for decode";
 523       *err = true;
 524       return false;
 525     }
 526   }
 527
 528   if (decrypt_config) {
 529     if (!subsamples.empty()) {
 530     // Create a new config with the updated subsamples.
 531     decrypt_config.reset(new DecryptConfig(
 532         decrypt_config->key_id(),
 533         decrypt_config->iv(),
 534         subsamples));
 535     }
 536     // else, use the existing config.
 537   } else if ((audio && is_audio_track_encrypted_) ||
 538              (video && is_video_track_encrypted_)) {
 539     // The media pipeline requires a DecryptConfig with an empty |iv|.
 540     // TODO(ddorwin): Refactor so we do not need a fake key ID ("1");
 541     decrypt_config.reset(
 542         new DecryptConfig("1", "", std::vector<SubsampleEntry>()));
 543   }
 544
 545   StreamParserBuffer::Type buffer_type = audio ? DemuxerStream::AUDIO :
 546       DemuxerStream::VIDEO;
 547
 548   // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 549   // type and allow multiple tracks for same media type, if applicable. See
 550   // https://crbug.com/341581.
 551   //
 552   // NOTE: MPEG's "random access point" concept is equivalent to the
 553   // downstream code's "is keyframe" concept.
 554   scoped_refptr<StreamParserBuffer> stream_buf =
 555       StreamParserBuffer::CopyFrom(&frame_buf[0], frame_buf.size(),
 556                                    runs_->is_random_access_point(),
 557                                    buffer_type, 0);
 558
 559   if (decrypt_config)
 560     stream_buf->set_decrypt_config(decrypt_config.Pass());
 561
 562   stream_buf->set_duration(runs_->duration());
 563   stream_buf->set_timestamp(runs_->cts());
 564   stream_buf->SetDecodeTimestamp(runs_->dts());
 565
 566   DVLOG(3) << "Pushing frame: aud=" << audio
 567            << ", key=" << runs_->is_keyframe()
 568            << ", rap=" << runs_->is_random_access_point()
 569            << ", dur=" << runs_->duration().InMilliseconds()
 570            << ", dts=" << runs_->dts().InMilliseconds()
 571            << ", cts=" << runs_->cts().InMilliseconds()
 572            << ", size=" << runs_->sample_size();
 573
 574   if (audio) {
 575     audio_buffers->push_back(stream_buf);
 576   } else {
 577     video_buffers->push_back(stream_buf);
 578   }
 579
 580   runs_->AdvanceSample();
 581   return true;
 582 }
 583
 584 bool MP4StreamParser::SendAndFlushSamples(BufferQueue* audio_buffers,
 585                                           BufferQueue* video_buffers) {
 586   if (audio_buffers->empty() && video_buffers->empty())
 587     return true;
 588
 589   TextBufferQueueMap empty_text_map;
 590   bool success = new_buffers_cb_.Run(*audio_buffers,
 591                                      *video_buffers,
 592                                      empty_text_map);
 593   audio_buffers->clear();
 594   video_buffers->clear();
 595   return success;
 596 }
 597
 598 bool MP4StreamParser::ReadAndDiscardMDATsUntil(int64 max_clear_offset) {
 599   bool err = false;
 600   int64 upper_bound = std::min(max_clear_offset, queue_.tail());
 601   while (mdat_tail_ < upper_bound) {
 602     const uint8* buf = NULL;
 603     int size = 0;
 604     queue_.PeekAt(mdat_tail_, &buf, &size);
 605
 606     FourCC type;
 607     int box_sz;
 608     if (!BoxReader::StartTopLevelBox(buf, size, media_log_, &type, &box_sz,
 609                                      &err))
 610       break;
 611
 612     if (type != FOURCC_MDAT) {
 613       MEDIA_LOG(DEBUG, media_log_)
 614           << "Unexpected box type while parsing MDATs: "
 615           << FourCCToString(type);
 616     }
 617     mdat_tail_ += box_sz;
 618   }
 619   queue_.Trim(std::min(mdat_tail_, upper_bound));
 620   return !err;
 621 }
 622
 623 void MP4StreamParser::ChangeState(State new_state) {
 624   DVLOG(2) << "Changing state: " << new_state;
 625   state_ = new_state;
 626 }
 627
 628 bool MP4StreamParser::HaveEnoughDataToEnqueueSamples() {
 629   DCHECK_EQ(state_, kWaitingForSampleData);
 630   // For muxed content, make sure we have data up to |highest_end_offset_|
 631   // so we can ensure proper enqueuing behavior. Otherwise assume we have enough
 632   // data and allow per sample offset checks to meter sample enqueuing.
 633   // TODO(acolwell): Fix trun box handling so we don't have to special case
 634   // muxed content.
 635   return !(has_audio_ && has_video_ &&
 636            queue_.tail() < highest_end_offset_ + moof_head_);
 637 }
 638
 639 bool MP4StreamParser::ComputeHighestEndOffset(const MovieFragment& moof) {
 640   highest_end_offset_ = 0;
 641
 642   TrackRunIterator runs(moov_.get(), media_log_);
 643   RCHECK(runs.Init(moof));
 644
 645   while (runs.IsRunValid()) {
 646     int64 aux_info_end_offset = runs.aux_info_offset() + runs.aux_info_size();
 647     if (aux_info_end_offset > highest_end_offset_)
 648       highest_end_offset_ = aux_info_end_offset;
 649
 650     while (runs.IsSampleValid()) {
 651       int64 sample_end_offset = runs.sample_offset() + runs.sample_size();
 652       if (sample_end_offset > highest_end_offset_)
 653         highest_end_offset_ = sample_end_offset;
 654
 655       runs.AdvanceSample();
 656     }
 657     runs.AdvanceRun();
 658   }
 659
 660   return true;
 661 }
 662
 663 }  // namespace mp4
 664 }  // namespace media