Implement ISO BMFF support in Media Source.

media/mp4/box_definitions.cc

Index: media/mp4/box_definitions.cc
diff --git a/media/mp4/box_definitions.cc b/media/mp4/box_definitions.cc
new file mode 100644
index 0000000000000000000000000000000000000000..4cb8e5522b379adf09920cd0e9997154f8a4c441
--- /dev/null
+++ b/media/mp4/box_definitions.cc
@@ -0,0 +1,550 @@
+// Copyright (c) 2012 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.
+
+#include "media/mp4/box_definitions.h"
+
+#include "base/logging.h"
+#include "media/mp4/box_reader.h"
+#include "media/mp4/rcheck.h"
+#include "media/mp4/fourccs.h"
+
+namespace media {
+namespace mp4 {
+
+bool Parse(BoxReader* r, FileType* type) {
+  RCHECK(r->Read(&type->major_brand) && r->Read(&type->minor_version));
+  size_t num_brands = (r->size() - r->pos()) / sizeof(FourCC);
+  return r->SkipBytes(sizeof(FourCC) * num_brands);  // compatible_brands
+}
+
+ProtectionSystemSpecificHeader::ProtectionSystemSpecificHeader() {}
+ProtectionSystemSpecificHeader::~ProtectionSystemSpecificHeader() {}
+
+bool Parse(BoxReader* r, ProtectionSystemSpecificHeader* pssh) {
+  uint32 size;
+  return r->SkipBytes(4) &&
+         r->ReadVec(&pssh->system_id, 16) &&
+         r->Read(&size) &&
+         r->ReadVec(&pssh->data, size);
+}
+
+SampleAuxiliaryInformationOffset::SampleAuxiliaryInformationOffset() {}
+SampleAuxiliaryInformationOffset::~SampleAuxiliaryInformationOffset() {}
+
+bool Parse(BoxReader* r, SampleAuxiliaryInformationOffset* saio) {
+  RCHECK(r->ReadFullBoxHeader());
+  if (r->flags() & 1)
+    RCHECK(r->SkipBytes(8));
+
+  uint32 count;
+  RCHECK(r->Read(&count) &&
+         r->HasBytes(count * (r->version() == 1 ? 8 : 4)));
+  saio->offsets.resize(count);
+
+  for (uint32 i = 0; i < count; i++) {
+    if (r->version() == 1) {
+      RCHECK(r->Read(&saio->offsets[i]));
+    } else {
+      RCHECK(r->Read4(&saio->offsets[i]));
+    }
+  }
+  return true;
+}
+
+SampleAuxiliaryInformationSize::SampleAuxiliaryInformationSize()
+  : default_sample_info_size(0), sample_count(0) {
+}
+
+SampleAuxiliaryInformationSize::~SampleAuxiliaryInformationSize() {}
+
+bool Parse(BoxReader* r, SampleAuxiliaryInformationSize* saiz) {
+  RCHECK(r->ReadFullBoxHeader());
+  if (r->flags() & 1)
+    RCHECK(r->SkipBytes(8));
+
+  RCHECK(r->Read(&saiz->default_sample_info_size) &&
+         r->Read(&saiz->sample_count));
+  if (saiz->default_sample_info_size == 0)
+    return r->ReadVec(&saiz->sample_info_sizes, saiz->sample_count);
+  return true;
+}
+
+bool Parse(BoxReader* r, OriginalFormat* format) {
+  return r->Read(&format->format);
+}
+
+bool Parse(BoxReader* r, SchemeType* schm) {
+  RCHECK(r->SkipBytes(4) && r->Read(&schm->type) && r->Read(&schm->version));
+  RCHECK(schm->type == FOURCC_CENC);
+  return true;
+}
+
+TrackEncryption::TrackEncryption()
+  : is_encrypted(false), default_iv_size(0) {
+}
+
+TrackEncryption::~TrackEncryption() {}
+
+bool Parse(BoxReader* r, TrackEncryption* tenc) {
+  uint8 flag;
+  RCHECK(r->SkipBytes(2) &&
+         r->Read(&flag) &&
+         r->Read(&tenc->default_iv_size) &&
+         r->ReadVec(&tenc->default_kid, 16));
+  tenc->is_encrypted = (flag != 0);
+  if (tenc->is_encrypted) {
+    RCHECK(tenc->default_iv_size == 8 || tenc->default_iv_size == 16);
+  } else {
+    RCHECK(tenc->default_iv_size == 0);
+  }
+  return true;
+}
+
+bool Parse(BoxReader* r, SchemeInfo* schi) {
+  return r->ScanChildren() && r->ReadChild(&schi->track_encryption);
+}
+
+bool Parse(BoxReader* r, ProtectionSchemeInfo* sinf) {
+  return r->ScanChildren() &&
+         r->ReadChild(&sinf->type) &&
+         r->ReadChild(&sinf->info);
+}
+
+bool Parse(BoxReader* r, MovieHeader* mvhd) {
+  RCHECK(r->ReadFullBoxHeader());
+
+  if (r->version() == 1) {
+    RCHECK(r->Read(&mvhd->creation_time) &&
+           r->Read(&mvhd->modification_time) &&
+           r->Read(&mvhd->timescale) &&
+           r->Read(&mvhd->duration));
+  } else {
+    RCHECK(r->Read4(&mvhd->creation_time) &&
+           r->Read4(&mvhd->modification_time) &&
+           r->Read(&mvhd->timescale) &&
+           r->Read4(&mvhd->duration));
+  }
+
+  RCHECK(r->Read(&mvhd->rate) &&
+         r->Read(&mvhd->volume) &&
+         r->SkipBytes(10) &&  // reserved
+         r->SkipBytes(36) &&  // matrix
+         r->SkipBytes(24) &&  // predefined zero
+         r->Read(&mvhd->next_track_id));
+  return true;
+}
+
+TrackHeader::TrackHeader() {}
+TrackHeader::~TrackHeader() {}
+
+
+bool Parse(BoxReader* r, TrackHeader* tkhd) {
+  RCHECK(r->ReadFullBoxHeader());
+  if (r->version() == 1) {
+    RCHECK(r->Read(&tkhd->creation_time) &&
+           r->Read(&tkhd->modification_time) &&
+           r->Read(&tkhd->track_id) &&
+           r->SkipBytes(4) &&    // reserved
+           r->Read(&tkhd->duration));
+  } else {
+    RCHECK(r->Read4(&tkhd->creation_time) &&
+           r->Read4(&tkhd->modification_time) &&
+           r->Read(&tkhd->track_id) &&
+           r->SkipBytes(4) &&   // reserved
+           r->Read4(&tkhd->duration));
+  }
+
+  RCHECK(r->SkipBytes(8) &&  // reserved
+         r->Read(&tkhd->layer) &&
+         r->Read(&tkhd->alternate_group) &&
+         r->Read(&tkhd->volume) &&
+         r->SkipBytes(2) &&  // reserved
+         r->SkipBytes(36) &&  // matrix
+         r->Read(&tkhd->width) &&
+         r->Read(&tkhd->height));
+  tkhd->width >>= 16;
+  tkhd->height >>= 16;
+  return true;
+}
+
+SampleDescription::SampleDescription() {}
+SampleDescription::~SampleDescription() {}
+
+bool Parse(BoxReader* r, SampleDescription* desc) {
+  uint32 count;
+  RCHECK(r->SkipBytes(4) && r->Read(&count) && r->ScanChildren());
+  desc->video_entries.clear();
+  desc->audio_entries.clear();
+
+  // Note: this value is preset before scanning begins. See comments in the
+  // Parse(Media*) function.
+  if (desc->type == kVideo) {
+    RCHECK(r->ReadAllChildren(&desc->video_entries));
+  } else if (desc->type == kAudio) {
+    RCHECK(r->ReadAllChildren(&desc->audio_entries));
+  }
+  return true;
+}
+
+SampleTable::SampleTable() {}
+SampleTable::~SampleTable() {}
+
+bool Parse(BoxReader* r, SampleTable* table) {
+  return r->ScanChildren() &&
+         r->ReadChild(&table->description);
+}
+
+EditList::EditList() {}
+EditList::~EditList() {}
+
+bool Parse(BoxReader* r, EditList* list) {
+  uint32 count;
+  RCHECK(r->ReadFullBoxHeader() && r->Read(&count));
+
+  if (r->version() == 1) {
+    RCHECK(r->HasBytes(count * 20));
+  } else {
+    RCHECK(r->HasBytes(count * 12));
+  }
+  list->edits.resize(count);
+
+  for (auto edit = list->edits.begin(); edit != list->edits.end(); ++edit) {
+    if (r->version() == 1) {
+      RCHECK(r->Read(&edit->segment_duration) &&
+             r->Read(&edit->media_time));
+    } else {
+      RCHECK(r->Read4(&edit->segment_duration) &&
+             r->Read4(&edit->media_time));
+    }
+    RCHECK(r->Read(&edit->media_rate_integer) &&
+           r->Read(&edit->media_rate_fraction));
+  }
+  return true;
+}
+
+Edit::Edit() {}
+Edit::~Edit() {}
+
+bool Parse(BoxReader* r, Edit* box) {
+  return r->ScanChildren() && r->ReadChild(&box->list);
+}
+
+bool Parse(BoxReader* r, HandlerReference* reference) {
+  FourCC type;
+  RCHECK(r->SkipBytes(8) && r->Read(&type));
+  // Note: remaining fields in box ignored
+  if (type == FOURCC_VIDE) {
+
+    reference->type = kVideo;
+  } else if (type == FOURCC_SOUN) {
+    reference->type = kAudio;
+  } else {
+    reference->type = kInvalid;
+  }
+  return true;
+}
+
+SampleEntry::SampleEntry() {}
+SampleEntry::~SampleEntry() {}
+
+bool ParseSampleEntryHead(BoxReader* r, SampleEntry* entry) {
+  entry->format = r->type();
+  return r->SkipBytes(6) && r->Read(&entry->data_reference_index);
+}
+
+AudioSampleEntry::AudioSampleEntry() {}
+AudioSampleEntry::~AudioSampleEntry() {}
+
+bool Parse(BoxReader* r, AudioSampleEntry* entry) {
+  RCHECK(ParseSampleEntryHead(r, entry) &&
+         r->SkipBytes(8) &&
+         r->Read(&entry->channelcount) &&
+         r->Read(&entry->samplesize) &&
+         r->SkipBytes(4) &&
+         r->Read(&entry->samplerate));
+  // Convert from 16.16 fixed point to integer
+  entry->samplerate >>= 16;
+
+  RCHECK(r->ScanChildren());
+  if (entry->format == FOURCC_ENCA) {
+    RCHECK(r->ReadChild(&entry->sinf));
+  }
+  return true;
+}
+
+VideoSampleEntry::VideoSampleEntry() {}
+VideoSampleEntry::~VideoSampleEntry() {}
+
+bool Parse(BoxReader* r, VideoSampleEntry* entry) {
+  RCHECK(ParseSampleEntryHead(r, entry) &&
+         r->SkipBytes(16) &&
+         r->Read(&entry->width) &&
+         r->Read(&entry->height) &&
+         r->SkipBytes(50));
+
+  RCHECK(r->ScanChildren());
+  if (entry->format == FOURCC_ENCV) {
+    RCHECK(r->ReadChild(&entry->sinf));
+  }
+
+  // TODO(strobe): Widevine test files do not follow spec here; check disabled
+  //if (entry->format == FOURCC_AVC1 ||
+  //    (entry->format == FOURCC_ENCV &&
+  //     entry->sinf.format.format == FOURCC_AVC1)) {
+    RCHECK(r->ReadChild(&entry->avcc));
+  //}
+  return true;
+}
+
+MediaHeader::MediaHeader() {}
+MediaHeader::~MediaHeader() {}
+
+bool Parse(BoxReader* r, MediaHeader* header) {
+  RCHECK(r->ReadFullBoxHeader());
+
+  if (r->version() == 1) {
+    RCHECK(r->Read(&header->creation_time) &&
+           r->Read(&header->modification_time) &&
+           r->Read(&header->timescale) &&
+           r->Read(&header->duration));
+  } else {
+    RCHECK(r->Read4(&header->creation_time) &&
+           r->Read4(&header->modification_time) &&
+           r->Read(&header->timescale) &&
+           r->Read4(&header->duration));
+  }
+  // Skip language information
+  return r->SkipBytes(4);
+}
+
+bool Parse(BoxReader* r, MediaInformation* information) {
+  return r->ScanChildren() &&
+         r->ReadChild(&information->sample_table);
+}
+
+bool Parse(BoxReader* r, Media* media) {
+  RCHECK(r->ScanChildren() &&
+         r->ReadChild(&media->header) &&
+         r->ReadChild(&media->handler));
+
+  // Maddeningly, the HandlerReference box specifies how to parse the
+  // SampleDescription box, making the latter the only box (of those that we
+  // support) which cannot be parsed correctly on its own (or even with
+  // information from its strict ancestor tree). We thus copy the handler type
+  // to the sample description box *before* parsing it to provide this
+  // information while parsing.
+  media->information.sample_table.description.type = media->handler.type;
+  RCHECK(r->ReadChild(&media->information));
+  return true;
+}
+
+bool Parse(BoxReader* r, Track* track) {
+  RCHECK(r->ScanChildren() &&
+         r->ReadChild(&track->header) &&
+         r->ReadChild(&track->media) &&
+         r->MaybeReadChild(&track->edit));
+  return true;
+}
+
+bool Parse(BoxReader* r, MovieExtendsHeader* header) {
+  RCHECK(r->Read(&header->fragment_duration));
+  return true;
+}
+
+bool Parse(BoxReader* r, TrackExtends* extends) {
+  RCHECK(r->ReadFullBoxHeader() &&
+         r->Read(&extends->track_id) &&
+         r->Read(&extends->default_sample_description_index) &&
+         r->Read(&extends->default_sample_duration) &&
+         r->Read(&extends->default_sample_size) &&
+         r->Read(&extends->default_sample_flags));
+  return true;
+}
+
+MovieExtends::MovieExtends() {};
+MovieExtends::~MovieExtends() {};
+
+bool Parse(BoxReader* r, MovieExtends* extends) {
+  extends->header.fragment_duration = 0;
+  return r->ScanChildren() &&
+         r->MaybeReadChild(&extends->header) &&
+         // TODO(strobe): does not detect correspondence rule (trex must match
+         // one-to-one with trak)
+         r->ReadChildren(&extends->tracks);
+}
+
+Movie::Movie() {};
+Movie::~Movie() {};
+
+bool Parse(BoxReader* r, Movie* moov) {
+  return r->ScanChildren() &&
+         r->ReadChild(&moov->header) &&
+         r->ReadChildren(&moov->tracks) &&
+         // Media Source specific: 'mvex' required
+         r->ReadChild(&moov->extends) &&
+         r->MaybeReadChildren(&moov->pssh);
+}
+
+bool Parse(BoxReader* r, TrackFragmentDecodeTime* tfdt) {
+  RCHECK(r->ReadFullBoxHeader());
+  if (r->version() == 1)
+    return r->Read(&tfdt->decode_time);
+  else
+    return r->Read4(&tfdt->decode_time);
+}
+
+bool Parse(BoxReader* r, MovieFragmentHeader* mfhd) {
+  return r->SkipBytes(4) && r->Read(&mfhd->sequence_number);
+}
+
+bool Parse(BoxReader* r, TrackFragmentHeader* tfhd) {
+  memset(tfhd, 0, sizeof(TrackFragmentHeader));
+
+  RCHECK(r->ReadFullBoxHeader() && r->Read(&tfhd->track_id));
+
+  // Media Source specific: we require that 'default-base-is-moof' (14496-12
+  // Amendment 2) be set, and reject tracks that set 'base-data-offset-present'.
+  //
+  // TODO(strobe): Test files do not have 'default-base-is-moof' set explicitly
+  // (although they are structured as if it is set). Because this flag is hidden
+  // in Amendment 2, it might not be set by all implementors. Should we reject
+  // such files, or try them anyway?
+  //
+  // RCHECK((flags & 0x020000) && !(flags & 0x1));
+  RCHECK(!(r->flags() & 0x1));
+
+  if (r->flags() & 0x2)
+    RCHECK(r->SkipBytes(4));  // sample_description_index
+
+  if (r->flags() & 0x8)
+    RCHECK(r->Read(&tfhd->default_sample_duration));
+
+  if (r->flags() & 0x10)
+    RCHECK(r->Read(&tfhd->default_sample_size));
+
+  if (r->flags() & 0x20)
+    RCHECK(r->Read(&tfhd->default_sample_flags));
+
+  return true;
+}
+
+TrackFragmentRun::TrackFragmentRun() {}
+TrackFragmentRun::~TrackFragmentRun() {}
+
+bool Parse(BoxReader* r, TrackFragmentRun* trun) {
+  RCHECK(r->ReadFullBoxHeader() &&
+         r->Read(&trun->sample_count));
+  const uint32 flags = r->flags();
+
+  bool data_offset_present = flags & 0x1;
+  bool first_sample_flags_present = flags & 0x4;
+  bool sample_duration_present = flags & 0x100;
+  bool sample_size_present = flags & 0x200;
+  bool sample_flags_present = flags & 0x400;
+  bool sample_composition_time_offsets_present = flags & 0x800;
+
+  if (data_offset_present) {
+    RCHECK(r->Read(&trun->data_offset));
+  } else {
+    trun->data_offset = 0;
+  }
+
+  uint32 first_sample_flags;
+  if (first_sample_flags_present)
+    RCHECK(r->Read(&first_sample_flags));
+
+  int fields = sample_duration_present + sample_size_present +
+      sample_flags_present + sample_composition_time_offsets_present;
+  RCHECK(r->HasBytes(fields * trun->sample_count));
+
+  if (sample_duration_present)
+    trun->sample_durations.resize(trun->sample_count);
+  if (sample_size_present)
+    trun->sample_sizes.resize(trun->sample_count);
+  if (sample_flags_present)
+    trun->sample_flags.resize(trun->sample_count);
+  if (sample_composition_time_offsets_present)
+    trun->sample_composition_time_offsets.resize(trun->sample_count);
+
+  for (uint32 i = 0; i < trun->sample_count; ++i) {
+    if (sample_duration_present)
+      RCHECK(r->Read(&trun->sample_durations[i]));
+    if (sample_size_present)
+      RCHECK(r->Read(&trun->sample_sizes[i]));
+    if (sample_flags_present)
+      RCHECK(r->Read(&trun->sample_flags[i]));
+    if (sample_composition_time_offsets_present)
+      RCHECK(r->Read(&trun->sample_composition_time_offsets[i]));
+  }
+
+  if (first_sample_flags_present) {
+    if (trun->sample_flags.size() == 0) {
+      trun->sample_flags.push_back(first_sample_flags);
+    } else {
+      trun->sample_flags[0] = first_sample_flags;
+    }
+  }
+  return true;
+}
+
+TrackFragment::TrackFragment() {}
+TrackFragment::~TrackFragment() {}
+
+bool Parse(BoxReader* r, TrackFragment* traf) {
+  return r->ScanChildren() &&
+         r->ReadChild(&traf->header) &&
+         // Media Source specific: 'tfdt' required
+         r->ReadChild(&traf->decode_time) &&
+         r->MaybeReadChildren(&traf->runs) &&
+         r->MaybeReadChild(&traf->auxiliary_offset) &&
+         r->MaybeReadChild(&traf->auxiliary_size);
+}
+
+MovieFragment::MovieFragment() {}
+MovieFragment::~MovieFragment() {}
+
+bool Parse(BoxReader* r, MovieFragment* moof) {
+  RCHECK(r->ScanChildren() &&
+         r->ReadChild(&moof->header) &&
+         r->ReadChildren(&moof->tracks) &&
+         r->MaybeReadChildren(&moof->pssh));
+  return true;
+}
+
+bool Parse(BoxReader* r, SegmentIndex* sidx) {
+  RCHECK(r->ReadFullBoxHeader() &&
+         r->Read(&sidx->reference_id) &&
+         r->Read(&sidx->timescale));
+
+  if (r->version() == 0) {
+    RCHECK(r->Read4(&sidx->earliest_presentation_time) &&
+           r->Read4(&sidx->first_offset));
+  } else {
+    RCHECK(r->Read(&sidx->earliest_presentation_time) &&
+           r->Read(&sidx->first_offset));
+  }
+
+  uint32 reference_count;
+  RCHECK(r->SkipBytes(2) &&  // reserved
+         r->Read(&reference_count) &&
+         r->HasBytes(12 * reference_count));
+
+  for (uint32 i = 0; i < reference_count; ++i) {
+    uint32 flags;
+    RCHECK(r->Read(&sidx->sizes[i]) &&
+           r->Read(&sidx->durations[i]) &&
+           r->Read(&flags));
+
+    uint32 type = flags / 0x10000000 % 8;
+
+    RCHECK(!(sidx->sizes[i] & 0x8000000) != 0 ||
+            (flags & 0x80000000) != 0x80000000 ||
+            (type != 0 && type != 1) ||
+            (flags % 0x10000000 != 0));
+  }
+  return true;
+}
+
+}  // namespace mp4
+}  // namespace media