From 6b7ffc4559db251f11927704c19fd68b760ba0c5 Mon Sep 17 00:00:00 2001 From: Kilik Kuo Date: Wed, 3 Sep 2014 11:01:37 +0800 Subject: [PATCH] Bug 1022524 - Set up audio track and video track information for Ogg files. r=rillian --- content/media/ogg/OggCodecState.cpp | 102 ++++++- content/media/ogg/OggCodecState.h | 27 ++ content/media/ogg/OggReader.cpp | 415 +++++++++++++++++++--------- content/media/ogg/OggReader.h | 7 + 4 files changed, 412 insertions(+), 139 deletions(-) diff --git a/content/media/ogg/OggCodecState.cpp b/content/media/ogg/OggCodecState.cpp index bc348a14a6b2..8456d2a07560 100644 --- a/content/media/ogg/OggCodecState.cpp +++ b/content/media/ogg/OggCodecState.cpp @@ -1104,6 +1104,9 @@ static const long SKELETON_4_0_MIN_HEADER_LEN = 80; // Minimum length in bytes of a Skeleton 4.0 index packet. static const long SKELETON_4_0_MIN_INDEX_LEN = 42; +// Minimum length in bytes of a Skeleton 3.0/4.0 Fisbone packet. +static const long SKELETON_MIN_FISBONE_LEN = 52; + // Minimum possible size of a compressed index keypoint. static const size_t MIN_KEY_POINT_SIZE = 2; @@ -1127,18 +1130,34 @@ static const size_t INDEX_FIRST_NUMER_OFFSET = 26; static const size_t INDEX_LAST_NUMER_OFFSET = 34; static const size_t INDEX_KEYPOINT_OFFSET = 42; +// Byte-offsets of the fields in the Skeleton Fisbone packet. +static const size_t FISBONE_MSG_FIELDS_OFFSET = 8; +static const size_t FISBONE_SERIALNO_OFFSET = 12; + static bool IsSkeletonBOS(ogg_packet* aPacket) { - return aPacket->bytes >= SKELETON_MIN_HEADER_LEN && + static_assert(SKELETON_MIN_HEADER_LEN >= 8, + "Minimum length of skeleton BOS header incorrect"); + return aPacket->bytes >= SKELETON_MIN_HEADER_LEN && memcmp(reinterpret_cast(aPacket->packet), "fishead", 8) == 0; } static bool IsSkeletonIndex(ogg_packet* aPacket) { + static_assert(SKELETON_4_0_MIN_INDEX_LEN >= 5, + "Minimum length of skeleton index header incorrect"); return aPacket->bytes >= SKELETON_4_0_MIN_INDEX_LEN && memcmp(reinterpret_cast(aPacket->packet), "index", 5) == 0; } +static bool IsSkeletonFisbone(ogg_packet* aPacket) +{ + static_assert(SKELETON_MIN_FISBONE_LEN >= 8, + "Minimum length of skeleton fisbone header incorrect"); + return aPacket->bytes >= SKELETON_MIN_FISBONE_LEN && + memcmp(reinterpret_cast(aPacket->packet), "fisbone", 8) == 0; +} + // Reads a variable length encoded integer at p. Will not read // past aLimit. Returns pointer to character after end of integer. static const unsigned char* ReadVariableLengthInt(const unsigned char* p, @@ -1367,6 +1386,85 @@ nsresult SkeletonState::GetDuration(const nsTArray& aTracks, return duration.isValid() ? NS_OK : NS_ERROR_FAILURE; } +bool SkeletonState::DecodeFisbone(ogg_packet* aPacket) +{ + if (aPacket->bytes < static_cast(FISBONE_MSG_FIELDS_OFFSET + 4)) { + return false; + } + uint32_t offsetMsgField = LittleEndian::readUint32(aPacket->packet + FISBONE_MSG_FIELDS_OFFSET); + + if (aPacket->bytes < static_cast(FISBONE_SERIALNO_OFFSET + 4)) { + return false; + } + uint32_t serialno = LittleEndian::readUint32(aPacket->packet + FISBONE_SERIALNO_OFFSET); + + CheckedUint32 checked_fields_pos = CheckedUint32(FISBONE_MSG_FIELDS_OFFSET) + offsetMsgField; + if (!checked_fields_pos.isValid() || + aPacket->bytes < static_cast(checked_fields_pos.value())) { + return false; + } + int64_t msgLength = aPacket->bytes - checked_fields_pos.value(); + char* msgProbe = (char*)aPacket->packet + checked_fields_pos.value(); + char* msgHead = msgProbe; + nsAutoPtr field(new MessageField()); + + const static FieldPatternType kFieldTypeMaps[] = { + {"Content-Type:", eContentType}, + {"Role:", eRole}, + {"Name:", eName}, + {"Language:", eLanguage}, + {"Title:", eTitle}, + {"Display-hint:", eDisplayHint}, + {"Altitude:", eAltitude}, + {"TrackOrder:", eTrackOrder}, + {"Track dependencies:", eTrackDependencies} + }; + + bool isContentTypeParsed = false; + while (msgLength > 1) { + if (*msgProbe == '\r' && *(msgProbe+1) == '\n') { + nsAutoCString strMsg(msgHead, msgProbe-msgHead); + for (size_t i = 0; i < ArrayLength(kFieldTypeMaps); i++) { + if (strMsg.Find(kFieldTypeMaps[i].mPatternToRecognize) != -1) { + // The content of message header fields follows [RFC2822], and the + // mandatory message field must be encoded in US-ASCII, others + // must be be encoded in UTF-8. "Content-Type" must come first + // for all of message header fields. + // See http://svn.annodex.net/standards/draft-pfeiffer-oggskeleton-current.txt. + if (i != 0 && !isContentTypeParsed) { + return false; + } + + if ((i == 0 && IsASCII(strMsg)) || (i != 0 && IsUTF8(strMsg))) { + EMsgHeaderType eHeaderType = kFieldTypeMaps[i].mMsgHeaderType; + if (!field->mValuesStore.Contains(eHeaderType)) { + uint32_t nameLen = strlen(kFieldTypeMaps[i].mPatternToRecognize); + field->mValuesStore.Put(eHeaderType, new nsCString(msgHead+nameLen, + msgProbe-msgHead-nameLen)); + } + isContentTypeParsed = i==0 ? true : isContentTypeParsed; + } + break; + } + } + msgProbe += 2; + msgLength -= 2; + msgHead = msgProbe; + continue; + } + msgLength--; + msgProbe++; + }; + + if (!mMsgFieldStore.Contains(serialno)) { + mMsgFieldStore.Put(serialno, field.forget()); + } else { + return false; + } + + return true; +} + bool SkeletonState::DecodeHeader(ogg_packet* aPacket) { nsAutoRef autoRelease(aPacket); @@ -1396,6 +1494,8 @@ bool SkeletonState::DecodeHeader(ogg_packet* aPacket) return true; } else if (IsSkeletonIndex(aPacket) && mVersion >= SKELETON_VERSION(4,0)) { return DecodeIndex(aPacket); + } else if (IsSkeletonFisbone(aPacket)) { + return DecodeFisbone(aPacket); } else if (aPacket->e_o_s) { mDoneReadingHeaders = true; return true; diff --git a/content/media/ogg/OggCodecState.h b/content/media/ogg/OggCodecState.h index 87458d9da68c..b115f31a5677 100644 --- a/content/media/ogg/OggCodecState.h +++ b/content/media/ogg/OggCodecState.h @@ -382,10 +382,35 @@ private: // version numbers. #define SKELETON_VERSION(major, minor) (((major)<<16)|(minor)) +enum EMsgHeaderType { + eContentType, + eRole, + eName, + eLanguage, + eTitle, + eDisplayHint, + eAltitude, + eTrackOrder, + eTrackDependencies +}; + +typedef struct { + const char* mPatternToRecognize; + EMsgHeaderType mMsgHeaderType; +} FieldPatternType; + +// Stores the message information for different logical bitstream. +typedef struct { + nsClassHashtable mValuesStore; +} MessageField; + class SkeletonState : public OggCodecState { public: explicit SkeletonState(ogg_page* aBosPage); ~SkeletonState(); + + nsClassHashtable mMsgFieldStore; + CodecType GetType() { return TYPE_SKELETON; } bool DecodeHeader(ogg_packet* aPacket); int64_t Time(int64_t granulepos) { return -1; } @@ -454,6 +479,8 @@ private: // Decodes an index packet. Returns false on failure. bool DecodeIndex(ogg_packet* aPacket); + // Decodes an fisbone packet. Returns false on failure. + bool DecodeFisbone(ogg_packet* aPacket); // Gets the keypoint you must seek to in order to get the keyframe required // to render the stream at time aTarget on stream with serial aSerialno. diff --git a/content/media/ogg/OggReader.cpp b/content/media/ogg/OggReader.cpp index 054c07394ea9..d7eff58e5f75 100644 --- a/content/media/ogg/OggReader.cpp +++ b/content/media/ogg/OggReader.cpp @@ -86,6 +86,51 @@ PageSync(MediaResource* aResource, // is about 4300 bytes, so we read the file in chunks larger than that. static const int PAGE_STEP = 8192; +// Return the corresponding category in aKind based on the following specs. +// (https://www.whatwg.org/specs/web-apps/current- +// work/multipage/embedded-content.html#dom-audiotrack-kind) & +// (http://wiki.xiph.org/SkeletonHeaders) +static const nsString GetKind(const nsCString& aRole) +{ + if (aRole.Find("audio/main") != -1 || aRole.Find("video/main") != -1) { + return NS_LITERAL_STRING("main"); + } else if (aRole.Find("audio/alternate") != -1 || + aRole.Find("video/alternate") != -1) { + return NS_LITERAL_STRING("alternative"); + } else if (aRole.Find("audio/audiodesc") != -1) { + return NS_LITERAL_STRING("descriptions"); + } else if (aRole.Find("audio/described") != -1) { + return NS_LITERAL_STRING("main-desc"); + } else if (aRole.Find("audio/dub") != -1) { + return NS_LITERAL_STRING("translation"); + } else if (aRole.Find("audio/commentary") != -1) { + return NS_LITERAL_STRING("commentary"); + } else if (aRole.Find("video/sign") != -1) { + return NS_LITERAL_STRING("sign"); + } else if (aRole.Find("video/captioned") != -1) { + return NS_LITERAL_STRING("captions"); + } else if (aRole.Find("video/subtitled") != -1) { + return NS_LITERAL_STRING("subtitles"); + } + return EmptyString(); +} + +static void InitTrack(MessageField* aMsgInfo, TrackInfo* aInfo, bool aEnable) +{ + MOZ_ASSERT(aMsgInfo); + MOZ_ASSERT(aInfo); + + nsCString* sName = aMsgInfo->mValuesStore.Get(eName); + nsCString* sRole = aMsgInfo->mValuesStore.Get(eRole); + nsCString* sTitle = aMsgInfo->mValuesStore.Get(eTitle); + nsCString* sLanguage = aMsgInfo->mValuesStore.Get(eLanguage); + aInfo->Init(sName? NS_ConvertUTF8toUTF16(*sName):EmptyString(), + sRole? GetKind(*sRole):EmptyString(), + sTitle? NS_ConvertUTF8toUTF16(*sTitle):EmptyString(), + sLanguage? NS_ConvertUTF8toUTF16(*sLanguage):EmptyString(), + aEnable); +} + OggReader::OggReader(AbstractMediaDecoder* aDecoder) : MediaDecoderReader(aDecoder), mMonitor("OggReader"), @@ -165,6 +210,7 @@ bool OggReader::ReadHeaders(OggCodecState* aState) void OggReader::BuildSerialList(nsTArray& aTracks) { + // Obtaining seek index information for currently active bitstreams. if (HasVideo()) { aTracks.AppendElement(mTheoraState->mSerial); } @@ -179,6 +225,159 @@ void OggReader::BuildSerialList(nsTArray& aTracks) } } +void OggReader::SetupTargetTheora(TheoraState* aTheoraState) +{ + if (mTheoraState) { + mTheoraState->Reset(); + } + nsIntRect picture = nsIntRect(aTheoraState->mInfo.pic_x, + aTheoraState->mInfo.pic_y, + aTheoraState->mInfo.pic_width, + aTheoraState->mInfo.pic_height); + + nsIntSize displaySize = nsIntSize(aTheoraState->mInfo.pic_width, + aTheoraState->mInfo.pic_height); + + // Apply the aspect ratio to produce the intrinsic display size we report + // to the element. + ScaleDisplayByAspectRatio(displaySize, aTheoraState->mPixelAspectRatio); + + nsIntSize frameSize(aTheoraState->mInfo.frame_width, + aTheoraState->mInfo.frame_height); + if (IsValidVideoRegion(frameSize, picture, displaySize)) { + // Video track's frame sizes will not overflow. Activate the video track. + mPicture = picture; + + VideoFrameContainer* container = mDecoder->GetVideoFrameContainer(); + if (container) { + container->SetCurrentFrame(gfxIntSize(displaySize.width, displaySize.height), + nullptr, + TimeStamp::Now()); + } + + // Copy Theora info data for time computations on other threads. + memcpy(&mTheoraInfo, &aTheoraState->mInfo, sizeof(mTheoraInfo)); + + mTheoraState = aTheoraState; + mTheoraSerial = aTheoraState->mSerial; + } +} + +void OggReader::SetupTargetVorbis(VorbisState* aVorbisState) +{ + if (mVorbisState) { + mVorbisState->Reset(); + } + // Copy Vorbis info data for time computations on other threads. + memcpy(&mVorbisInfo, &aVorbisState->mInfo, sizeof(mVorbisInfo)); + mVorbisInfo.codec_setup = nullptr; + mVorbisState = aVorbisState; + mVorbisSerial = aVorbisState->mSerial; +} + +void OggReader::SetupTargetOpus(OpusState* aOpusState) +{ + if (mOpusState) { + mOpusState->Reset(); + } + mOpusState = aOpusState; + mOpusSerial = aOpusState->mSerial; + mOpusPreSkip = aOpusState->mPreSkip; +} + +void OggReader::SetupTargetSkeleton(SkeletonState* aSkeletonState) +{ + // Setup skeleton related information after mVorbisState & mTheroState + // being set (if they exist). + if (aSkeletonState) { + if (!HasAudio() && !HasVideo()) { + // We have a skeleton track, but no audio or video, may as well disable + // the skeleton, we can't do anything useful with this media. + aSkeletonState->Deactivate(); + } else if (ReadHeaders(aSkeletonState) && aSkeletonState->HasIndex()) { + // Extract the duration info out of the index, so we don't need to seek to + // the end of resource to get it. + nsAutoTArray tracks; + BuildSerialList(tracks); + int64_t duration = 0; + if (NS_SUCCEEDED(aSkeletonState->GetDuration(tracks, duration))) { + ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor()); + mDecoder->SetMediaDuration(duration); + LOG(PR_LOG_DEBUG, ("Got duration from Skeleton index %lld", duration)); + } + } + } +} + +void OggReader::SetupMediaTracksInfo(const nsTArray& aSerials) +{ + // For each serial number + // 1. Retrieve a codecState from mCodecStore by this serial number. + // 2. Retrieve a message field from mMsgFieldStore by this serial number. + // 3. For now, skip if the serial number refers to a non-primary bitstream. + // 4. Setup track and other audio/video related information per different types. + for (size_t i = 0; i < aSerials.Length(); i++) { + uint32_t serial = aSerials[i]; + OggCodecState* codecState = mCodecStore.Get(serial); + + MessageField* msgInfo = nullptr; + if (mSkeletonState && mSkeletonState->mMsgFieldStore.Contains(serial)) { + mSkeletonState->mMsgFieldStore.Get(serial, &msgInfo); + } + + if (codecState->GetType() == OggCodecState::TYPE_THEORA) { + TheoraState* theoraState = static_cast(codecState); + if (!(mTheoraState && mTheoraState->mSerial == theoraState->mSerial)) { + continue; + } + + if (msgInfo) { + InitTrack(msgInfo, &mInfo.mVideo.mTrackInfo, mTheoraState == theoraState); + } + + nsIntRect picture = nsIntRect(theoraState->mInfo.pic_x, + theoraState->mInfo.pic_y, + theoraState->mInfo.pic_width, + theoraState->mInfo.pic_height); + nsIntSize displaySize = nsIntSize(theoraState->mInfo.pic_width, + theoraState->mInfo.pic_height); + nsIntSize frameSize(theoraState->mInfo.frame_width, + theoraState->mInfo.frame_height); + ScaleDisplayByAspectRatio(displaySize, theoraState->mPixelAspectRatio); + mInfo.mVideo.mDisplay = displaySize; + mInfo.mVideo.mHasVideo = IsValidVideoRegion(frameSize, picture, displaySize)? true:false; + } else if (codecState->GetType() == OggCodecState::TYPE_VORBIS) { + VorbisState* vorbisState = static_cast(codecState); + if (!(mVorbisState && mVorbisState->mSerial == vorbisState->mSerial)) { + continue; + } + + if (msgInfo) { + InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, mVorbisState == vorbisState); + } + + mInfo.mAudio.mHasAudio = true; + mInfo.mAudio.mRate = vorbisState->mInfo.rate; + mInfo.mAudio.mChannels = vorbisState->mInfo.channels; +#ifdef MOZ_OPUS + } else if (codecState->GetType() == OggCodecState::TYPE_OPUS) { + OpusState* opusState = static_cast(codecState); + if (!(mOpusState && mOpusState->mSerial == opusState->mSerial)) { + continue; + } + + if (msgInfo) { + InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, mOpusState == opusState); + } + + mInfo.mAudio.mHasAudio = true; + mInfo.mAudio.mRate = opusState->mRate; + mInfo.mAudio.mChannels = opusState->mChannels; +#endif + } + } +} + nsresult OggReader::ReadMetadata(MediaInfo* aInfo, MetadataTags** aTags) { @@ -193,6 +392,7 @@ nsresult OggReader::ReadMetadata(MediaInfo* aInfo, ogg_page page; nsAutoTArray bitstreams; + nsTArray serials; bool readAllBOS = false; while (!readAllBOS) { if (!ReadOggPage(&page)) { @@ -215,41 +415,7 @@ nsresult OggReader::ReadMetadata(MediaInfo* aInfo, codecState = OggCodecState::Create(&page); mCodecStore.Add(serial, codecState); bitstreams.AppendElement(codecState); - if (codecState && - codecState->GetType() == OggCodecState::TYPE_VORBIS && - !mVorbisState) - { - // First Vorbis bitstream, we'll play this one. Subsequent Vorbis - // bitstreams will be ignored. - mVorbisState = static_cast(codecState); - } - if (codecState && - codecState->GetType() == OggCodecState::TYPE_THEORA && - !mTheoraState) - { - // First Theora bitstream, we'll play this one. Subsequent Theora - // bitstreams will be ignored. - mTheoraState = static_cast(codecState); - } -#ifdef MOZ_OPUS - if (codecState && - codecState->GetType() == OggCodecState::TYPE_OPUS && - !mOpusState) - { - if (mOpusEnabled) { - mOpusState = static_cast(codecState); - } else { - NS_WARNING("Opus decoding disabled." - " See media.opus.enabled in about:config"); - } - } -#endif /* MOZ_OPUS */ - if (codecState && - codecState->GetType() == OggCodecState::TYPE_SKELETON && - !mSkeletonState) - { - mSkeletonState = static_cast(codecState); - } + serials.AppendElement(serial); } codecState = mCodecStore.Get(serial); @@ -261,96 +427,56 @@ nsresult OggReader::ReadMetadata(MediaInfo* aInfo, } // We've read all BOS pages, so we know the streams contained in the media. - // Now process all available header packets in the active Theora, Vorbis and - // Skeleton streams. - - // Deactivate any non-primary bitstreams. - for (uint32_t i = 0; i < bitstreams.Length(); i++) { + // 1. Process all available header packets in the Theora, Vorbis/Opus bitstreams. + // 2. Find the first encountered Theora/Vorbis/Opus bitstream, and configure + // it as the target A/V bitstream. + // 3. Deactivate the rest of bitstreams for now, until we have MediaInfo + // support multiple track infos. + for (uint32_t i = 0; i < bitstreams.Length(); ++i) { OggCodecState* s = bitstreams[i]; - if (s != mVorbisState && + if (s) { + if (s->GetType() == OggCodecState::TYPE_THEORA && ReadHeaders(s)) { + if (!mTheoraState) { + TheoraState* theoraState = static_cast(s); + SetupTargetTheora(theoraState); + } else { + s->Deactivate(); + } + } else if (s->GetType() == OggCodecState::TYPE_VORBIS && ReadHeaders(s)) { + if (!mVorbisState) { + VorbisState* vorbisState = static_cast(s); + SetupTargetVorbis(vorbisState); + *aTags = vorbisState->GetTags(); + } else { + s->Deactivate(); + } #ifdef MOZ_OPUS - s != mOpusState && -#endif /* MOZ_OPUS */ - s != mTheoraState && s != mSkeletonState) { - s->Deactivate(); - } - } - - if (mTheoraState && ReadHeaders(mTheoraState)) { - nsIntRect picture = nsIntRect(mTheoraState->mInfo.pic_x, - mTheoraState->mInfo.pic_y, - mTheoraState->mInfo.pic_width, - mTheoraState->mInfo.pic_height); - - nsIntSize displaySize = nsIntSize(mTheoraState->mInfo.pic_width, - mTheoraState->mInfo.pic_height); - - // Apply the aspect ratio to produce the intrinsic display size we report - // to the element. - ScaleDisplayByAspectRatio(displaySize, mTheoraState->mPixelAspectRatio); - - nsIntSize frameSize(mTheoraState->mInfo.frame_width, - mTheoraState->mInfo.frame_height); - if (IsValidVideoRegion(frameSize, picture, displaySize)) { - // Video track's frame sizes will not overflow. Activate the video track. - mInfo.mVideo.mHasVideo = true; - mInfo.mVideo.mDisplay = displaySize; - mPicture = picture; - - VideoFrameContainer* container = mDecoder->GetVideoFrameContainer(); - if (container) { - container->SetCurrentFrame(gfxIntSize(displaySize.width, displaySize.height), - nullptr, - TimeStamp::Now()); + } else if (s->GetType() == OggCodecState::TYPE_OPUS && ReadHeaders(s)) { + if (mOpusEnabled) { + if (!mOpusState) { + OpusState* opusState = static_cast(s); + SetupTargetOpus(opusState); + *aTags = opusState->GetTags(); + } else { + s->Deactivate(); + } + } else { + NS_WARNING("Opus decoding disabled." + " See media.opus.enabled in about:config"); + } +#endif // MOZ_OPUS + } else if (s->GetType() == OggCodecState::TYPE_SKELETON && !mSkeletonState) { + mSkeletonState = static_cast(s); + } else { + // Deactivate any non-primary bitstreams. + s->Deactivate(); } - // Copy Theora info data for time computations on other threads. - memcpy(&mTheoraInfo, &mTheoraState->mInfo, sizeof(mTheoraInfo)); - mTheoraSerial = mTheoraState->mSerial; } } - if (mVorbisState && ReadHeaders(mVorbisState)) { - mInfo.mAudio.mHasAudio = true; - mInfo.mAudio.mRate = mVorbisState->mInfo.rate; - mInfo.mAudio.mChannels = mVorbisState->mInfo.channels; - // Copy Vorbis info data for time computations on other threads. - memcpy(&mVorbisInfo, &mVorbisState->mInfo, sizeof(mVorbisInfo)); - mVorbisInfo.codec_setup = nullptr; - mVorbisSerial = mVorbisState->mSerial; - *aTags = mVorbisState->GetTags(); - } else { - memset(&mVorbisInfo, 0, sizeof(mVorbisInfo)); - } -#ifdef MOZ_OPUS - if (mOpusState && ReadHeaders(mOpusState)) { - mInfo.mAudio.mHasAudio = true; - mInfo.mAudio.mRate = mOpusState->mRate; - mInfo.mAudio.mChannels = mOpusState->mChannels; - mOpusSerial = mOpusState->mSerial; - mOpusPreSkip = mOpusState->mPreSkip; - - *aTags = mOpusState->GetTags(); - } -#endif - if (mSkeletonState) { - if (!HasAudio() && !HasVideo()) { - // We have a skeleton track, but no audio or video, may as well disable - // the skeleton, we can't do anything useful with this media. - mSkeletonState->Deactivate(); - } else if (ReadHeaders(mSkeletonState) && mSkeletonState->HasIndex()) { - // Extract the duration info out of the index, so we don't need to seek to - // the end of resource to get it. - nsAutoTArray tracks; - BuildSerialList(tracks); - int64_t duration = 0; - if (NS_SUCCEEDED(mSkeletonState->GetDuration(tracks, duration))) { - ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor()); - mDecoder->SetMediaDuration(duration); - LOG(PR_LOG_DEBUG, ("Got duration from Skeleton index %lld", duration)); - } - } - } + SetupTargetSkeleton(mSkeletonState); + SetupMediaTracksInfo(serials); if (HasAudio() || HasVideo()) { ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor()); @@ -673,42 +799,55 @@ bool OggReader::ReadOggChain() return false; } - nsAutoPtr info(new MediaInfo()); + MessageField* msgInfo = nullptr; + if (mSkeletonState && mSkeletonState->mMsgFieldStore.Contains(serial)) { + mSkeletonState->mMsgFieldStore.Get(serial, &msgInfo); + } + if ((newVorbisState && ReadHeaders(newVorbisState)) && (mVorbisState->mInfo.rate == newVorbisState->mInfo.rate) && (mVorbisState->mInfo.channels == newVorbisState->mInfo.channels)) { - mVorbisState->Reset(); - mVorbisState = newVorbisState; - mVorbisSerial = mVorbisState->mSerial; + + SetupTargetVorbis(newVorbisState); LOG(PR_LOG_DEBUG, ("New vorbis ogg link, serial=%d\n", mVorbisSerial)); + + if (msgInfo) { + InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, true); + } + mInfo.mAudio.mRate = newVorbisState->mInfo.rate; + mInfo.mAudio.mChannels = newVorbisState->mInfo.channels; + chained = true; - info->mAudio.mRate = mVorbisState->mInfo.rate; - info->mAudio.mChannels = mVorbisState->mInfo.channels; - tags = mVorbisState->GetTags(); + tags = newVorbisState->GetTags(); } #ifdef MOZ_OPUS if ((newOpusState && ReadHeaders(newOpusState)) && (mOpusState->mRate == newOpusState->mRate) && (mOpusState->mChannels == newOpusState->mChannels)) { - mOpusState->Reset(); - mOpusState = newOpusState; - mOpusSerial = mOpusState->mSerial; + + SetupTargetOpus(newOpusState); + + if (msgInfo) { + InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, true); + } + mInfo.mAudio.mRate = newOpusState->mRate; + mInfo.mAudio.mChannels = newOpusState->mChannels; + chained = true; - info->mAudio.mRate = mOpusState->mRate; - info->mAudio.mChannels = mOpusState->mChannels; - tags = mOpusState->GetTags(); + tags = newOpusState->GetTags(); } #endif if (chained) { SetChained(true); { - info->mAudio.mHasAudio = HasAudio(); - info->mVideo.mHasVideo = HasVideo(); - int rate = info->mAudio.mRate; + mInfo.mAudio.mHasAudio = HasAudio(); + mInfo.mVideo.mHasVideo = HasVideo(); + nsAutoPtr info(new MediaInfo()); + *info = mInfo; ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor()); - mDecoder->QueueMetadata((mDecodedAudioFrames * USECS_PER_S) / rate, + mDecoder->QueueMetadata((mDecodedAudioFrames * USECS_PER_S) / mInfo.mAudio.mRate, info.forget(), tags); } return true; @@ -1024,7 +1163,6 @@ int64_t OggReader::RangeEndTime(int64_t aStartOffset, endTime = -1; break; } - continue; } @@ -1054,7 +1192,6 @@ int64_t OggReader::RangeEndTime(int64_t aStartOffset, OggCodecState* codecState = nullptr; codecState = mCodecStore.Get(serial); - if (!codecState) { // This page is from a bitstream which we haven't encountered yet. // It's probably from a new "link" in a "chained" ogg. Don't @@ -1143,7 +1280,9 @@ OggReader::SelectSeekRange(const nsTArray& ranges, OggReader::IndexedSeekResult OggReader::RollbackIndexedSeek(int64_t aOffset) { - mSkeletonState->Deactivate(); + if (mSkeletonState) { + mSkeletonState->Deactivate(); + } MediaResource* resource = mDecoder->GetResource(); NS_ENSURE_TRUE(resource != nullptr, SEEK_FATAL_ERROR); nsresult res = resource->Seek(nsISeekableStream::NS_SEEK_SET, aOffset); diff --git a/content/media/ogg/OggReader.h b/content/media/ogg/OggReader.h index 82006b4e1de4..12a581b56a48 100644 --- a/content/media/ogg/OggReader.h +++ b/content/media/ogg/OggReader.h @@ -268,6 +268,13 @@ private: // various SkeletonState functions. void BuildSerialList(nsTArray& aTracks); + // Setup target bitstreams for decoding. + void SetupTargetTheora(TheoraState* aTheoraState); + void SetupTargetVorbis(VorbisState* aVorbisState); + void SetupTargetOpus(OpusState* aOpusState); + void SetupTargetSkeleton(SkeletonState* aSkeletonState); + void SetupMediaTracksInfo(const nsTArray& aSerials); + OggCodecStore mCodecStore; // Decode state of the Theora bitstream we're decoding, if we have video.