TimeUnit
AudioSink::GetEndTime() const
{
int64_t written;
{
MonitorAutoLock mon(mMonitor);
written = mWritten;
}
TimeUnit played = FramesToTimeUnit(written, mOutputRate) + mStartTime;
if (!played.IsValid()) {
NS_WARNING("Int overflow calculating audio end time");
return TimeUnit::Zero();
}
return played;
}
TimeUnit
AudioSink::GetEndTime() const
{
int64_t written;
{
MonitorAutoLock mon(mMonitor);
written = mWritten;
}
TimeUnit played = FramesToTimeUnit(written, mOutputRate) + mStartTime;
if (!played.IsValid()) {
NS_WARNING("Int overflow calculating audio end time");
return TimeUnit::Zero();
}
// As we may be resampling, rounding errors may occur. Ensure we never get
// past the original end time.
return std::min(mLastEndTime, played);
}
already_AddRefed<AudioData>
AudioSink::CreateAudioFromBuffer(AlignedAudioBuffer&& aBuffer,
AudioData* aReference)
{
uint32_t frames = aBuffer.Length() / mOutputChannels;
if (!frames) {
return nullptr;
}
auto duration = FramesToTimeUnit(frames, mOutputRate);
if (!duration.IsValid()) {
NS_WARNING("Int overflow in AudioSink");
mErrored = true;
return nullptr;
}
RefPtr<AudioData> data =
new AudioData(aReference->mOffset,
aReference->mTime,
duration,
frames,
Move(aBuffer),
mOutputChannels,
mOutputRate);
return data.forget();
}
RefPtr<MediaDataDecoder::DecodePromise>
VorbisDataDecoder::ProcessDecode(MediaRawData* aSample)
{
MOZ_ASSERT(mTaskQueue->IsCurrentThreadIn());
const unsigned char* aData = aSample->Data();
size_t aLength = aSample->Size();
int64_t aOffset = aSample->mOffset;
MOZ_ASSERT(mPacketCount >= 3);
if (!mLastFrameTime ||
mLastFrameTime.ref() != aSample->mTime.ToMicroseconds()) {
// We are starting a new block.
mFrames = 0;
mLastFrameTime = Some(aSample->mTime.ToMicroseconds());
}
ogg_packet pkt = InitVorbisPacket(
aData, aLength, false, aSample->mEOS,
aSample->mTimecode.ToMicroseconds(), mPacketCount++);
int err = vorbis_synthesis(&mVorbisBlock, &pkt);
if (err) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("vorbis_synthesis:%d", err)),
__func__);
}
err = vorbis_synthesis_blockin(&mVorbisDsp, &mVorbisBlock);
if (err) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("vorbis_synthesis_blockin:%d", err)),
__func__);
}
VorbisPCMValue** pcm = 0;
int32_t frames = vorbis_synthesis_pcmout(&mVorbisDsp, &pcm);
if (frames == 0) {
return DecodePromise::CreateAndResolve(DecodedData(), __func__);
}
DecodedData results;
while (frames > 0) {
uint32_t channels = mVorbisDsp.vi->channels;
uint32_t rate = mVorbisDsp.vi->rate;
AlignedAudioBuffer buffer(frames*channels);
if (!buffer) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_OUT_OF_MEMORY, __func__), __func__);
}
for (uint32_t j = 0; j < channels; ++j) {
VorbisPCMValue* channel = pcm[j];
for (uint32_t i = 0; i < uint32_t(frames); ++i) {
buffer[i*channels + j] = MOZ_CONVERT_VORBIS_SAMPLE(channel[i]);
}
}
auto duration = FramesToTimeUnit(frames, rate);
if (!duration.IsValid()) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Overflow converting audio duration")),
__func__);
}
auto total_duration = FramesToTimeUnit(mFrames, rate);
if (!total_duration.IsValid()) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Overflow converting audio total_duration")),
__func__);
}
auto time = total_duration + aSample->mTime;
if (!time.IsValid()) {
return DecodePromise::CreateAndReject(
MediaResult(
NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Overflow adding total_duration and aSample->mTime")),
__func__);
};
if (!mAudioConverter) {
AudioConfig in(
AudioConfig::ChannelLayout(channels, VorbisLayout(channels)), rate);
AudioConfig out(channels, rate);
if (!in.IsValid() || !out.IsValid()) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_DOM_MEDIA_FATAL_ERR,
RESULT_DETAIL("Invalid channel layout:%u", channels)),
__func__);
}
mAudioConverter = MakeUnique<AudioConverter>(in, out);
}
MOZ_ASSERT(mAudioConverter->CanWorkInPlace());
AudioSampleBuffer data(Move(buffer));
data = mAudioConverter->Process(Move(data));
results.AppendElement(new AudioData(aOffset, time, duration,
frames, data.Forget(), channels, rate));
mFrames += frames;
err = vorbis_synthesis_read(&mVorbisDsp, frames);
if (err) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("vorbis_synthesis_read:%d", err)),
__func__);
}
frames = vorbis_synthesis_pcmout(&mVorbisDsp, &pcm);
}
return DecodePromise::CreateAndResolve(Move(results), __func__);
}
MediaResult
FFmpegAudioDecoder<LIBAV_VER>::DoDecode(MediaRawData* aSample)
{
AVPacket packet;
mLib->av_init_packet(&packet);
packet.data = const_cast<uint8_t*>(aSample->Data());
packet.size = aSample->Size();
if (!PrepareFrame()) {
return MediaResult(
NS_ERROR_OUT_OF_MEMORY,
RESULT_DETAIL("FFmpeg audio decoder failed to allocate frame"));
}
int64_t samplePosition = aSample->mOffset;
media::TimeUnit pts = media::TimeUnit::FromMicroseconds(aSample->mTime);
while (packet.size > 0) {
int decoded;
int bytesConsumed =
mLib->avcodec_decode_audio4(mCodecContext, mFrame, &decoded, &packet);
if (bytesConsumed < 0) {
NS_WARNING("FFmpeg audio decoder error.");
return MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("FFmpeg audio error:%d", bytesConsumed));
}
if (mFrame->format != AV_SAMPLE_FMT_FLT &&
mFrame->format != AV_SAMPLE_FMT_FLTP &&
mFrame->format != AV_SAMPLE_FMT_S16 &&
mFrame->format != AV_SAMPLE_FMT_S16P &&
mFrame->format != AV_SAMPLE_FMT_S32 &&
mFrame->format != AV_SAMPLE_FMT_S32P) {
return MediaResult(
NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("FFmpeg audio decoder outputs unsupported audio format"));
}
if (decoded) {
uint32_t numChannels = mCodecContext->channels;
AudioConfig::ChannelLayout layout(numChannels);
if (!layout.IsValid()) {
return MediaResult(
NS_ERROR_DOM_MEDIA_FATAL_ERR,
RESULT_DETAIL("Unsupported channel layout:%u", numChannels));
}
uint32_t samplingRate = mCodecContext->sample_rate;
AlignedAudioBuffer audio =
CopyAndPackAudio(mFrame, numChannels, mFrame->nb_samples);
if (!audio) {
return MediaResult(NS_ERROR_OUT_OF_MEMORY, __func__);
}
media::TimeUnit duration =
FramesToTimeUnit(mFrame->nb_samples, samplingRate);
if (!duration.IsValid()) {
return MediaResult(
NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Invalid sample duration"));
}
RefPtr<AudioData> data = new AudioData(samplePosition,
pts.ToMicroseconds(),
duration.ToMicroseconds(),
mFrame->nb_samples,
Move(audio),
numChannels,
samplingRate);
mCallback->Output(data);
pts += duration;
if (!pts.IsValid()) {
return MediaResult(
NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Invalid count of accumulated audio samples"));
}
}
packet.data += bytesConsumed;
packet.size -= bytesConsumed;
samplePosition += bytesConsumed;
}
return NS_OK;
}
HRESULT
WMFAudioMFTManager::Output(int64_t aStreamOffset,
nsRefPtr<MediaData>& aOutData)
{
aOutData = nullptr;
RefPtr<IMFSample> sample;
HRESULT hr;
int typeChangeCount = 0;
while (true) {
hr = mDecoder->Output(&sample);
if (hr == MF_E_TRANSFORM_NEED_MORE_INPUT) {
return hr;
}
if (hr == MF_E_TRANSFORM_STREAM_CHANGE) {
hr = UpdateOutputType();
NS_ENSURE_TRUE(SUCCEEDED(hr), hr);
// Catch infinite loops, but some decoders perform at least 2 stream
// changes on consecutive calls, so be permissive.
// 100 is arbitrarily > 2.
NS_ENSURE_TRUE(typeChangeCount < 100, MF_E_TRANSFORM_STREAM_CHANGE);
++typeChangeCount;
continue;
}
break;
}
NS_ENSURE_TRUE(SUCCEEDED(hr), hr);
RefPtr<IMFMediaBuffer> buffer;
hr = sample->ConvertToContiguousBuffer(byRef(buffer));
NS_ENSURE_TRUE(SUCCEEDED(hr), hr);
BYTE* data = nullptr; // Note: *data will be owned by the IMFMediaBuffer, we don't need to free it.
DWORD maxLength = 0, currentLength = 0;
hr = buffer->Lock(&data, &maxLength, ¤tLength);
NS_ENSURE_TRUE(SUCCEEDED(hr), hr);
// Sometimes when starting decoding, the AAC decoder gives us samples
// with a negative timestamp. AAC does usually have preroll (or encoder
// delay) encoded into its bitstream, but the amount encoded to the stream
// is variable, and it not signalled in-bitstream. There is sometimes
// signalling in the MP4 container what the preroll amount, but it's
// inconsistent. It looks like WMF's AAC encoder may take this into
// account, so strip off samples with a negative timestamp to get us
// to a 0-timestamp start. This seems to maintain A/V sync, so we can run
// with this until someone complains...
// We calculate the timestamp and the duration based on the number of audio
// frames we've already played. We don't trust the timestamp stored on the
// IMFSample, as sometimes it's wrong, possibly due to buggy encoders?
// If this sample block comes after a discontinuity (i.e. a gap or seek)
// reset the frame counters, and capture the timestamp. Future timestamps
// will be offset from this block's timestamp.
UINT32 discontinuity = false;
sample->GetUINT32(MFSampleExtension_Discontinuity, &discontinuity);
if (mMustRecaptureAudioPosition || discontinuity) {
// Update the output type, in case this segment has a different
// rate. This also triggers on the first sample, which can have a
// different rate than is advertised in the container, and sometimes we
// don't get a MF_E_TRANSFORM_STREAM_CHANGE when the rate changes.
hr = UpdateOutputType();
NS_ENSURE_TRUE(SUCCEEDED(hr), hr);
mAudioFrameSum = 0;
LONGLONG timestampHns = 0;
hr = sample->GetSampleTime(×tampHns);
NS_ENSURE_TRUE(SUCCEEDED(hr), hr);
mAudioTimeOffset = media::TimeUnit::FromMicroseconds(timestampHns / 10);
mMustRecaptureAudioPosition = false;
}
// We can assume PCM 16 output.
int32_t numSamples = currentLength / 2;
int32_t numFrames = numSamples / mAudioChannels;
MOZ_ASSERT(numFrames >= 0);
MOZ_ASSERT(numSamples >= 0);
if (numFrames == 0) {
// All data from this chunk stripped, loop back and try to output the next
// frame, if possible.
return S_OK;
}
nsAutoArrayPtr<AudioDataValue> audioData(new AudioDataValue[numSamples]);
int16_t* pcm = (int16_t*)data;
for (int32_t i = 0; i < numSamples; ++i) {
audioData[i] = AudioSampleToFloat(pcm[i]);
}
buffer->Unlock();
media::TimeUnit timestamp =
mAudioTimeOffset + FramesToTimeUnit(mAudioFrameSum, mAudioRate);
NS_ENSURE_TRUE(timestamp.IsValid(), E_FAIL);
mAudioFrameSum += numFrames;
media::TimeUnit duration = FramesToTimeUnit(numFrames, mAudioRate);
NS_ENSURE_TRUE(duration.IsValid(), E_FAIL);
aOutData = new AudioData(aStreamOffset,
timestamp.ToMicroseconds(),
duration.ToMicroseconds(),
numFrames,
audioData.forget(),
mAudioChannels,
mAudioRate);
#ifdef LOG_SAMPLE_DECODE
LOG("Decoded audio sample! timestamp=%lld duration=%lld currentLength=%u",
timestamp.ToMicroseconds(), duration.ToMicroseconds(), currentLength);
#endif
return S_OK;
}
void
FFmpegAudioDecoder<LIBAV_VER>::DecodePacket(MediaRawData* aSample)
{
AVPacket packet;
av_init_packet(&packet);
packet.data = const_cast<uint8_t*>(aSample->Data());
packet.size = aSample->Size();
if (!PrepareFrame()) {
NS_WARNING("FFmpeg audio decoder failed to allocate frame.");
mCallback->Error();
return;
}
int64_t samplePosition = aSample->mOffset;
media::TimeUnit pts = media::TimeUnit::FromMicroseconds(aSample->mTime);
while (packet.size > 0) {
int decoded;
int bytesConsumed =
avcodec_decode_audio4(mCodecContext, mFrame, &decoded, &packet);
if (bytesConsumed < 0) {
NS_WARNING("FFmpeg audio decoder error.");
mCallback->Error();
return;
}
if (decoded) {
uint32_t numChannels = mCodecContext->channels;
uint32_t samplingRate = mCodecContext->sample_rate;
nsAutoArrayPtr<AudioDataValue> audio(
CopyAndPackAudio(mFrame, numChannels, mFrame->nb_samples));
media::TimeUnit duration =
FramesToTimeUnit(mFrame->nb_samples, samplingRate);
if (!duration.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
mCallback->Error();
return;
}
nsRefPtr<AudioData> data = new AudioData(samplePosition,
pts.ToMicroseconds(),
duration.ToMicroseconds(),
mFrame->nb_samples,
audio.forget(),
numChannels,
samplingRate);
mCallback->Output(data);
pts += duration;
if (!pts.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
mCallback->Error();
return;
}
}
packet.data += bytesConsumed;
packet.size -= bytesConsumed;
samplePosition += bytesConsumed;
}
if (mTaskQueue->IsEmpty()) {
mCallback->InputExhausted();
}
}
void
FFmpegAudioDecoder<LIBAV_VER>::DecodePacket(MediaRawData* aSample)
{
MOZ_ASSERT(mTaskQueue->IsCurrentThreadIn());
AVPacket packet;
mLib->av_init_packet(&packet);
packet.data = const_cast<uint8_t*>(aSample->Data());
packet.size = aSample->Size();
if (!PrepareFrame()) {
NS_WARNING("FFmpeg audio decoder failed to allocate frame.");
mCallback->Error();
return;
}
int64_t samplePosition = aSample->mOffset;
media::TimeUnit pts = media::TimeUnit::FromMicroseconds(aSample->mTime);
while (packet.size > 0) {
int decoded;
int bytesConsumed =
mLib->avcodec_decode_audio4(mCodecContext, mFrame, &decoded, &packet);
if (bytesConsumed < 0) {
NS_WARNING("FFmpeg audio decoder error.");
mCallback->Error();
return;
}
if (decoded) {
uint32_t numChannels = mCodecContext->channels;
AudioConfig::ChannelLayout layout(numChannels);
if (!layout.IsValid()) {
mCallback->Error();
return;
}
uint32_t samplingRate = mCodecContext->sample_rate;
AlignedAudioBuffer audio =
CopyAndPackAudio(mFrame, numChannels, mFrame->nb_samples);
media::TimeUnit duration =
FramesToTimeUnit(mFrame->nb_samples, samplingRate);
if (!audio || !duration.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
mCallback->Error();
return;
}
RefPtr<AudioData> data = new AudioData(samplePosition,
pts.ToMicroseconds(),
duration.ToMicroseconds(),
mFrame->nb_samples,
Move(audio),
numChannels,
samplingRate);
mCallback->Output(data);
pts += duration;
if (!pts.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
mCallback->Error();
return;
}
}
packet.data += bytesConsumed;
packet.size -= bytesConsumed;
samplePosition += bytesConsumed;
}
if (mTaskQueue->IsEmpty()) {
mCallback->InputExhausted();
}
}
RefPtr<MediaDataDecoder::DecodePromise>
WaveDataDecoder::ProcessDecode(MediaRawData* aSample)
{
size_t aLength = aSample->Size();
BufferReader aReader(aSample->Data(), aLength);
int64_t aOffset = aSample->mOffset;
int32_t frames = aLength * 8 / mInfo.mBitDepth / mInfo.mChannels;
AlignedAudioBuffer buffer(frames * mInfo.mChannels);
if (!buffer) {
return DecodePromise::CreateAndReject(
MediaResult(NS_ERROR_OUT_OF_MEMORY, __func__), __func__);
}
for (int i = 0; i < frames; ++i) {
for (unsigned int j = 0; j < mInfo.mChannels; ++j) {
if (mInfo.mProfile == 6) { //ALAW Data
auto res = aReader.ReadU8();
if (res.isErr()) {
return DecodePromise::CreateAndReject(
MediaResult(res.unwrapErr(), __func__), __func__);
}
int16_t decoded = DecodeALawSample(res.unwrap());
buffer[i * mInfo.mChannels + j] =
IntegerToAudioSample<AudioDataValue>(decoded);
} else if (mInfo.mProfile == 7) { //ULAW Data
auto res = aReader.ReadU8();
if (res.isErr()) {
return DecodePromise::CreateAndReject(
MediaResult(res.unwrapErr(), __func__), __func__);
}
int16_t decoded = DecodeULawSample(res.unwrap());
buffer[i * mInfo.mChannels + j] =
IntegerToAudioSample<AudioDataValue>(decoded);
} else { //PCM Data
if (mInfo.mBitDepth == 8) {
auto res = aReader.ReadU8();
if (res.isErr()) {
return DecodePromise::CreateAndReject(
MediaResult(res.unwrapErr(), __func__), __func__);
}
buffer[i * mInfo.mChannels + j] =
UInt8bitToAudioSample<AudioDataValue>(res.unwrap());
} else if (mInfo.mBitDepth == 16) {
auto res = aReader.ReadLE16();
if (res.isErr()) {
return DecodePromise::CreateAndReject(
MediaResult(res.unwrapErr(), __func__), __func__);
}
buffer[i * mInfo.mChannels + j] =
IntegerToAudioSample<AudioDataValue>(res.unwrap());
} else if (mInfo.mBitDepth == 24) {
auto res = aReader.ReadLE24();
if (res.isErr()) {
return DecodePromise::CreateAndReject(
MediaResult(res.unwrapErr(), __func__), __func__);
}
buffer[i * mInfo.mChannels + j] =
Int24bitToAudioSample<AudioDataValue>(res.unwrap());
}
}
}
}
auto duration = FramesToTimeUnit(frames, mInfo.mRate);
return DecodePromise::CreateAndResolve(
DecodedData{ new AudioData(aOffset, aSample->mTime, duration, frames,
Move(buffer), mInfo.mChannels, mInfo.mRate) },
__func__);
}
nsresult
AppleATDecoder::DecodeSample(MediaRawData* aSample)
{
// Array containing the queued decoded audio frames, about to be output.
nsTArray<AudioDataValue> outputData;
UInt32 channels = mOutputFormat.mChannelsPerFrame;
// Pick a multiple of the frame size close to a power of two
// for efficient allocation.
const uint32_t MAX_AUDIO_FRAMES = 128;
const uint32_t maxDecodedSamples = MAX_AUDIO_FRAMES * channels;
// Descriptions for _decompressed_ audio packets. ignored.
nsAutoArrayPtr<AudioStreamPacketDescription>
packets(new AudioStreamPacketDescription[MAX_AUDIO_FRAMES]);
// This API insists on having packets spoon-fed to it from a callback.
// This structure exists only to pass our state.
PassthroughUserData userData =
{ channels, (UInt32)aSample->Size(), aSample->Data() };
// Decompressed audio buffer
nsAutoArrayPtr<AudioDataValue> decoded(new AudioDataValue[maxDecodedSamples]);
do {
AudioBufferList decBuffer;
decBuffer.mNumberBuffers = 1;
decBuffer.mBuffers[0].mNumberChannels = channels;
decBuffer.mBuffers[0].mDataByteSize =
maxDecodedSamples * sizeof(AudioDataValue);
decBuffer.mBuffers[0].mData = decoded.get();
// in: the max number of packets we can handle from the decoder.
// out: the number of packets the decoder is actually returning.
UInt32 numFrames = MAX_AUDIO_FRAMES;
OSStatus rv = AudioConverterFillComplexBuffer(mConverter,
_PassthroughInputDataCallback,
&userData,
&numFrames /* in/out */,
&decBuffer,
packets.get());
if (rv && rv != kNoMoreDataErr) {
LOG("Error decoding audio stream: %d\n", rv);
return NS_ERROR_FAILURE;
}
if (numFrames) {
outputData.AppendElements(decoded.get(), numFrames * channels);
}
if (rv == kNoMoreDataErr) {
break;
}
} while (true);
if (outputData.IsEmpty()) {
return NS_OK;
}
size_t numFrames = outputData.Length() / channels;
int rate = mOutputFormat.mSampleRate;
media::TimeUnit duration = FramesToTimeUnit(numFrames, rate);
if (!duration.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
return NS_ERROR_FAILURE;
}
#ifdef LOG_SAMPLE_DECODE
LOG("pushed audio at time %lfs; duration %lfs\n",
(double)aSample->mTime / USECS_PER_S,
duration.ToSeconds());
#endif
nsAutoArrayPtr<AudioDataValue> data(new AudioDataValue[outputData.Length()]);
PodCopy(data.get(), &outputData[0], outputData.Length());
RefPtr<AudioData> audio = new AudioData(aSample->mOffset,
aSample->mTime,
duration.ToMicroseconds(),
numFrames,
data.forget(),
channels,
rate);
mCallback->Output(audio);
return NS_OK;
}
MediaResult
AppleATDecoder::DecodeSample(MediaRawData* aSample)
{
MOZ_ASSERT(mTaskQueue->IsCurrentThreadIn());
// Array containing the queued decoded audio frames, about to be output.
nsTArray<AudioDataValue> outputData;
UInt32 channels = mOutputFormat.mChannelsPerFrame;
// Pick a multiple of the frame size close to a power of two
// for efficient allocation.
const uint32_t MAX_AUDIO_FRAMES = 128;
const uint32_t maxDecodedSamples = MAX_AUDIO_FRAMES * channels;
// Descriptions for _decompressed_ audio packets. ignored.
auto packets = MakeUnique<AudioStreamPacketDescription[]>(MAX_AUDIO_FRAMES);
// This API insists on having packets spoon-fed to it from a callback.
// This structure exists only to pass our state.
PassthroughUserData userData =
{ channels, (UInt32)aSample->Size(), aSample->Data() };
// Decompressed audio buffer
AlignedAudioBuffer decoded(maxDecodedSamples);
if (!decoded) {
return NS_ERROR_OUT_OF_MEMORY;
}
do {
AudioBufferList decBuffer;
decBuffer.mNumberBuffers = 1;
decBuffer.mBuffers[0].mNumberChannels = channels;
decBuffer.mBuffers[0].mDataByteSize =
maxDecodedSamples * sizeof(AudioDataValue);
decBuffer.mBuffers[0].mData = decoded.get();
// in: the max number of packets we can handle from the decoder.
// out: the number of packets the decoder is actually returning.
UInt32 numFrames = MAX_AUDIO_FRAMES;
OSStatus rv = AudioConverterFillComplexBuffer(mConverter,
_PassthroughInputDataCallback,
&userData,
&numFrames /* in/out */,
&decBuffer,
packets.get());
if (rv && rv != kNoMoreDataErr) {
LOG("Error decoding audio sample: %d\n", rv);
return MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("Error decoding audio sample: %d @ %lld",
rv, aSample->mTime));
}
if (numFrames) {
outputData.AppendElements(decoded.get(), numFrames * channels);
}
if (rv == kNoMoreDataErr) {
break;
}
} while (true);
if (outputData.IsEmpty()) {
return NS_OK;
}
size_t numFrames = outputData.Length() / channels;
int rate = mOutputFormat.mSampleRate;
media::TimeUnit duration = FramesToTimeUnit(numFrames, rate);
if (!duration.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
return MediaResult(
NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL(
"Invalid count of accumulated audio samples: num:%llu rate:%d",
uint64_t(numFrames), rate));
}
#ifdef LOG_SAMPLE_DECODE
LOG("pushed audio at time %lfs; duration %lfs\n",
(double)aSample->mTime / USECS_PER_S,
duration.ToSeconds());
#endif
AudioSampleBuffer data(outputData.Elements(), outputData.Length());
if (!data.Data()) {
return NS_ERROR_OUT_OF_MEMORY;
}
if (mChannelLayout && !mAudioConverter) {
AudioConfig in(*mChannelLayout.get(), rate);
AudioConfig out(channels, rate);
if (!in.IsValid() || !out.IsValid()) {
return MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("Invalid audio config"));
}
mAudioConverter = MakeUnique<AudioConverter>(in, out);
}
if (mAudioConverter) {
MOZ_ASSERT(mAudioConverter->CanWorkInPlace());
data = mAudioConverter->Process(Move(data));
}
RefPtr<AudioData> audio = new AudioData(aSample->mOffset,
aSample->mTime,
duration.ToMicroseconds(),
numFrames,
data.Forget(),
channels,
rate);
mCallback->Output(audio);
return NS_OK;
}
MediaResult
FFmpegAudioDecoder<LIBAV_VER>::DoDecode(MediaRawData* aSample,
uint8_t* aData,
int aSize,
bool* aGotFrame,
DecodedData& aResults)
{
AVPacket packet;
mLib->av_init_packet(&packet);
packet.data = const_cast<uint8_t*>(aData);
packet.size = aSize;
if (aGotFrame) {
*aGotFrame = false;
}
if (!PrepareFrame()) {
return MediaResult(
NS_ERROR_OUT_OF_MEMORY,
RESULT_DETAIL("FFmpeg audio decoder failed to allocate frame"));
}
int64_t samplePosition = aSample->mOffset;
media::TimeUnit pts = aSample->mTime;
while (packet.size > 0) {
int decoded;
int bytesConsumed =
mLib->avcodec_decode_audio4(mCodecContext, mFrame, &decoded, &packet);
if (bytesConsumed < 0) {
NS_WARNING("FFmpeg audio decoder error.");
return MediaResult(NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL("FFmpeg audio error:%d", bytesConsumed));
}
if (decoded) {
if (mFrame->format != AV_SAMPLE_FMT_FLT &&
mFrame->format != AV_SAMPLE_FMT_FLTP &&
mFrame->format != AV_SAMPLE_FMT_S16 &&
mFrame->format != AV_SAMPLE_FMT_S16P &&
mFrame->format != AV_SAMPLE_FMT_S32 &&
mFrame->format != AV_SAMPLE_FMT_S32P) {
return MediaResult(
NS_ERROR_DOM_MEDIA_DECODE_ERR,
RESULT_DETAIL(
"FFmpeg audio decoder outputs unsupported audio format"));
}
uint32_t numChannels = mCodecContext->channels;
uint32_t samplingRate = mCodecContext->sample_rate;
AlignedAudioBuffer audio =
CopyAndPackAudio(mFrame, numChannels, mFrame->nb_samples);
if (!audio) {
return MediaResult(NS_ERROR_OUT_OF_MEMORY, __func__);
}
media::TimeUnit duration =
FramesToTimeUnit(mFrame->nb_samples, samplingRate);
if (!duration.IsValid()) {
return MediaResult(NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Invalid sample duration"));
}
media::TimeUnit newpts = pts + duration;
if (!newpts.IsValid()) {
return MediaResult(
NS_ERROR_DOM_MEDIA_OVERFLOW_ERR,
RESULT_DETAIL("Invalid count of accumulated audio samples"));
}
aResults.AppendElement(new AudioData(samplePosition,
pts,
duration,
mFrame->nb_samples,
Move(audio),
numChannels,
samplingRate,
mCodecContext->channel_layout));
pts = newpts;
if (aGotFrame) {
*aGotFrame = true;
}
}
packet.data += bytesConsumed;
packet.size -= bytesConsumed;
samplePosition += bytesConsumed;
}
return NS_OK;
}
