kotatogram/kotatogram-desktop
kotatogram/kotatogram-desktop
1
0
Fork 0
Code Issues Releases 45 Packages Wiki Activity
kotatogram-desktop/Telegram/SourceFiles/audio.cpp
John Preston 1b11a7feae voice message recording / sending done
2015-05-29 21:52:43 +03:00

1556 lines
No EOL
48 KiB
C++
Raw Blame History

/*
This file is part of Telegram Desktop,
the official desktop version of Telegram messaging app, see https://telegram.org
Telegram Desktop is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
It is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
Full license: https://github.com/telegramdesktop/tdesktop/blob/master/LICENSE
Copyright (c) 2014 John Preston, https://desktop.telegram.org
*/
#include "stdafx.h"
#include "audio.h"
#include <AL/al.h>
#include <AL/alc.h>
extern "C" {
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
#include <libavutil/opt.h>
#include <libswresample/swresample.h>
}
namespace {
ALCdevice *audioDevice = 0;
ALCcontext *audioContext = 0;
ALuint notifySource = 0;
ALuint notifyBuffer = 0;
QMutex playerMutex;
AudioPlayer *player = 0;
AudioCapture *capture = 0;
}
bool _checkALCError() {
ALenum errCode;
if ((errCode = alcGetError(audioDevice)) != ALC_NO_ERROR) {
LOG(("Audio Error: (alc) %1, %2").arg(errCode).arg((const char *)alcGetString(audioDevice, errCode)));
return false;
}
return true;
}
bool _checkCaptureError(ALCdevice *device) {
ALenum errCode;
if ((errCode = alcGetError(device)) != ALC_NO_ERROR) {
LOG(("Audio Error: (capture) %1, %2").arg(errCode).arg((const char *)alcGetString(audioDevice, errCode)));
return false;
}
return true;
}
bool _checkALError() {
ALenum errCode;
if ((errCode = alGetError()) != AL_NO_ERROR) {
LOG(("Audio Error: (al) %1, %2").arg(errCode).arg((const char *)alGetString(errCode)));
return false;
}
return true;
}
void audioInit() {
if (!capture) {
capture = new AudioCapture();
}
uint64 ms = getms();
if (audioDevice) return;
audioDevice = alcOpenDevice(0);
if (!audioDevice) {
LOG(("Audio Error: default sound device not present."));
return;
}
ALCint attributes[] = { ALC_STEREO_SOURCES, 8, 0 };
audioContext = alcCreateContext(audioDevice, attributes);
alcMakeContextCurrent(audioContext);
if (!_checkALCError()) return audioFinish();
ALfloat v[] = { 0.f, 0.f, -1.f, 0.f, 1.f, 0.f };
alListener3f(AL_POSITION, 0.f, 0.f, 0.f);
alListener3f(AL_VELOCITY, 0.f, 0.f, 0.f);
alListenerfv(AL_ORIENTATION, v);
alDistanceModel(AL_NONE);
alGenSources(1, &notifySource);
alSourcef(notifySource, AL_PITCH, 1.f);
alSourcef(notifySource, AL_GAIN, 1.f);
alSource3f(notifySource, AL_POSITION, 0, 0, 0);
alSource3f(notifySource, AL_VELOCITY, 0, 0, 0);
alSourcei(notifySource, AL_LOOPING, 0);
alGenBuffers(1, &notifyBuffer);
if (!_checkALError()) return audioFinish();
QFile notify(st::newMsgSound);
if (!notify.open(QIODevice::ReadOnly)) return audioFinish();
QByteArray blob = notify.readAll();
const char *data = blob.constData();
if (blob.size() < 44) return audioFinish();
if (*((const uint32*)(data + 0)) != 0x46464952) return audioFinish(); // ChunkID - "RIFF"
if (*((const uint32*)(data + 4)) != uint32(blob.size() - 8)) return audioFinish(); // ChunkSize
if (*((const uint32*)(data + 8)) != 0x45564157) return audioFinish(); // Format - "WAVE"
if (*((const uint32*)(data + 12)) != 0x20746d66) return audioFinish(); // Subchunk1ID - "fmt "
uint32 subchunk1Size = *((const uint32*)(data + 16)), extra = subchunk1Size - 16;
if (subchunk1Size < 16 || (extra && extra < 2)) return audioFinish();
if (*((const uint16*)(data + 20)) != 1) return audioFinish(); // AudioFormat - PCM (1)
uint16 numChannels = *((const uint16*)(data + 22));
if (numChannels != 1 && numChannels != 2) return audioFinish();
uint32 sampleRate = *((const uint32*)(data + 24));
uint32 byteRate = *((const uint32*)(data + 28));
uint16 blockAlign = *((const uint16*)(data + 32));
uint16 bitsPerSample = *((const uint16*)(data + 34));
if (bitsPerSample % 8) return audioFinish();
uint16 bytesPerSample = bitsPerSample / 8;
if (bytesPerSample != 1 && bytesPerSample != 2) return audioFinish();
if (blockAlign != numChannels * bytesPerSample) return audioFinish();
if (byteRate != sampleRate * blockAlign) return audioFinish();
if (extra) {
uint16 extraSize = *((const uint16*)(data + 36));
if (uint32(extraSize + 2) != extra) return audioFinish();
if (uint32(blob.size()) < 44 + extra) return audioFinish();
}
if (*((const uint32*)(data + extra + 36)) != 0x61746164) return audioFinish(); // Subchunk2ID - "data"
uint32 subchunk2Size = *((const uint32*)(data + extra + 40));
if (subchunk2Size % (numChannels * bytesPerSample)) return audioFinish();
uint32 numSamples = subchunk2Size / (numChannels * bytesPerSample);
if (uint32(blob.size()) < 44 + extra + subchunk2Size) return audioFinish();
data += 44 + extra;
ALenum format = 0;
switch (bytesPerSample) {
case 1:
switch (numChannels) {
case 1: format = AL_FORMAT_MONO8; break;
case 2: format = AL_FORMAT_STEREO8; break;
}
break;
case 2:
switch (numChannels) {
case 1: format = AL_FORMAT_MONO16; break;
case 2: format = AL_FORMAT_STEREO16; break;
}
break;
}
if (!format) return audioFinish();
alBufferData(notifyBuffer, format, data, subchunk2Size, sampleRate);
alSourcei(notifySource, AL_BUFFER, notifyBuffer);
if (!_checkALError()) return audioFinish();
player = new AudioPlayer();
alcSuspendContext(audioContext);
av_register_all();
avcodec_register_all();
LOG(("Audio init time: %1").arg(getms() - ms));
}
void audioPlayNotify() {
if (!audioPlayer()) return;
audioPlayer()->processContext();
alSourcePlay(notifySource);
emit audioPlayer()->faderOnTimer();
}
void audioFinish() {
if (player) {
delete player;
}
if (capture) {
delete capture;
}
alSourceStop(notifySource);
if (alIsBuffer(notifyBuffer)) {
alDeleteBuffers(1, &notifyBuffer);
notifyBuffer = 0;
}
if (alIsSource(notifySource)) {
alDeleteSources(1, &notifySource);
notifySource = 0;
}
if (audioContext) {
alcMakeContextCurrent(NULL);
alcDestroyContext(audioContext);
audioContext = 0;
}
if (audioDevice) {
alcCloseDevice(audioDevice);
audioDevice = 0;
}
}
AudioPlayer::AudioPlayer() : _current(0),
_fader(new AudioPlayerFader(&_faderThread)),
_loader(new AudioPlayerLoaders(&_loaderThread)) {
connect(this, SIGNAL(faderOnTimer()), _fader, SLOT(onTimer()));
connect(this, SIGNAL(loaderOnStart(AudioData*)), _loader, SLOT(onStart(AudioData*)));
connect(this, SIGNAL(loaderOnCancel(AudioData*)), _loader, SLOT(onCancel(AudioData*)));
connect(&_faderThread, SIGNAL(started()), _fader, SLOT(onInit()));
connect(&_loaderThread, SIGNAL(started()), _loader, SLOT(onInit()));
connect(&_faderThread, SIGNAL(finished()), _fader, SLOT(deleteLater()));
connect(&_loaderThread, SIGNAL(finished()), _loader, SLOT(deleteLater()));
connect(_loader, SIGNAL(needToCheck()), _fader, SLOT(onTimer()));
connect(_loader, SIGNAL(error(AudioData*)), this, SLOT(onError(AudioData*)));
connect(_fader, SIGNAL(needToPreload(AudioData*)), _loader, SLOT(onLoad(AudioData*)));
connect(_fader, SIGNAL(playPositionUpdated(AudioData*)), this, SIGNAL(updated(AudioData*)));
connect(_fader, SIGNAL(audioStopped(AudioData*)), this, SIGNAL(stopped(AudioData*)));
connect(_fader, SIGNAL(error(AudioData*)), this, SLOT(onError(AudioData*)));
_loaderThread.start();
_faderThread.start();
}
AudioPlayer::~AudioPlayer() {
{
QMutexLocker lock(&playerMutex);
player = 0;
}
for (int32 i = 0; i < AudioVoiceMsgSimultaneously; ++i) {
alSourceStop(_data[i].source);
if (alIsBuffer(_data[i].buffers[0])) {
alDeleteBuffers(3, _data[i].buffers);
for (int32 j = 0; j < 3; ++j) {
_data[i].buffers[j] = _data[i].samplesCount[j] = 0;
}
}
if (alIsSource(_data[i].source)) {
alDeleteSources(1, &_data[i].source);
_data[i].source = 0;
}
}
_faderThread.quit();
_loaderThread.quit();
_faderThread.wait();
_loaderThread.wait();
}
void AudioPlayer::onError(AudioData *audio) {
emit stopped(audio);
}
bool AudioPlayer::updateCurrentStarted(int32 pos) {
if (pos < 0) {
if (alIsSource(_data[_current].source)) {
alGetSourcei(_data[_current].source, AL_SAMPLE_OFFSET, &pos);
} else {
pos = 0;
}
}
if (!_checkALError()) {
_data[_current].state = AudioPlayerStopped;
onError(_data[_current].audio);
return false;
}
_data[_current].started = _data[_current].position = pos + _data[_current].skipStart;
return true;
}
void AudioPlayer::play(AudioData *audio) {
QMutexLocker lock(&playerMutex);
bool startNow = true;
if (_data[_current].audio != audio) {
switch (_data[_current].state) {
case AudioPlayerStarting:
case AudioPlayerResuming:
case AudioPlayerPlaying:
_data[_current].state = AudioPlayerFinishing;
updateCurrentStarted();
startNow = false;
break;
case AudioPlayerPausing: _data[_current].state = AudioPlayerFinishing; startNow = false; break;
case AudioPlayerPaused: _data[_current].state = AudioPlayerStopped; break;
}
if (_data[_current].audio) {
emit loaderOnCancel(_data[_current].audio);
emit faderOnTimer();
}
}
int32 index = 0;
for (; index < AudioVoiceMsgSimultaneously; ++index) {
if (_data[index].audio == audio) {
_current = index;
break;
}
}
if (index == AudioVoiceMsgSimultaneously && ++_current >= AudioVoiceMsgSimultaneously) {
_current -= AudioVoiceMsgSimultaneously;
}
_data[_current].audio = audio;
_data[_current].fname = audio->already(true);
_data[_current].data = audio->data;
if (_data[_current].fname.isEmpty() && _data[_current].data.isEmpty()) {
_data[_current].state = AudioPlayerStopped;
onError(audio);
} else if (updateCurrentStarted(0)) {
_data[_current].state = startNow ? AudioPlayerPlaying : AudioPlayerStarting;
_data[_current].loading = true;
emit loaderOnStart(audio);
}
}
void AudioPlayer::pauseresume() {
QMutexLocker lock(&playerMutex);
switch (_data[_current].state) {
case AudioPlayerPausing:
case AudioPlayerPaused:
if (_data[_current].state == AudioPlayerPaused) {
updateCurrentStarted();
}
_data[_current].state = AudioPlayerResuming;
processContext();
alSourcePlay(_data[_current].source);
break;
case AudioPlayerStarting:
case AudioPlayerResuming:
case AudioPlayerPlaying:
_data[_current].state = AudioPlayerPausing;
updateCurrentStarted();
break;
case AudioPlayerFinishing: _data[_current].state = AudioPlayerPausing; break;
}
emit faderOnTimer();
}
void AudioPlayer::currentState(AudioData **audio, AudioPlayerState *state, int64 *position, int64 *duration, int32 *frequency) {
QMutexLocker lock(&playerMutex);
if (audio) *audio = _data[_current].audio;
if (state) *state = _data[_current].state;
if (position) *position = _data[_current].position;
if (duration) *duration = _data[_current].duration;
if (frequency) *frequency = _data[_current].frequency;
}
void AudioPlayer::clearStoppedAtStart(AudioData *audio) {
QMutexLocker lock(&playerMutex);
if (_data[_current].audio == audio && _data[_current].state == AudioPlayerStoppedAtStart) {
_data[_current].state = AudioPlayerStopped;
}
}
void AudioPlayer::processContext() {
_fader->processContext();
}
AudioCapture::AudioCapture() : _capture(new AudioCaptureInner(&_captureThread)) {
connect(this, SIGNAL(captureOnStart()), _capture, SLOT(onStart()));
connect(this, SIGNAL(captureOnStop(bool)), _capture, SLOT(onStop(bool)));
connect(_capture, SIGNAL(done(QByteArray,qint32)), this, SIGNAL(onDone(QByteArray,qint32)));
connect(_capture, SIGNAL(update(qint16,qint32)), this, SIGNAL(onUpdate(qint16,qint32)));
connect(_capture, SIGNAL(error()), this, SIGNAL(onError()));
connect(&_captureThread, SIGNAL(started()), _capture, SLOT(onInit()));
connect(&_captureThread, SIGNAL(finished()), _capture, SLOT(deleteLater()));
_captureThread.start();
}
void AudioCapture::start() {
emit captureOnStart();
}
void AudioCapture::stop(bool needResult) {
emit captureOnStop(needResult);
}
AudioCapture::~AudioCapture() {
capture = 0;
_captureThread.quit();
_captureThread.wait();
}
AudioPlayer *audioPlayer() {
return player;
}
AudioCapture *audioCapture() {
return capture;
}
AudioPlayerFader::AudioPlayerFader(QThread *thread) : _timer(this), _suspendFlag(false) {
moveToThread(thread);
_timer.moveToThread(thread);
_suspendTimer.moveToThread(thread);
_timer.setSingleShot(true);
connect(&_timer, SIGNAL(timeout()), this, SLOT(onTimer()));
_suspendTimer.setSingleShot(true);
connect(&_suspendTimer, SIGNAL(timeout()), this, SLOT(onSuspendTimer()));
connect(this, SIGNAL(stopSuspend()), this, SLOT(onSuspendTimerStop()), Qt::QueuedConnection);
}
void AudioPlayerFader::onInit() {
}
void AudioPlayerFader::onTimer() {
bool hasFading = false, hasPlaying = false;
QMutexLocker lock(&playerMutex);
AudioPlayer *voice = audioPlayer();
if (!voice) return;
for (int32 i = 0; i < AudioVoiceMsgSimultaneously; ++i) {
AudioPlayer::Msg &m(voice->_data[i]);
if (m.state == AudioPlayerStopped || m.state == AudioPlayerStoppedAtStart || m.state == AudioPlayerPaused || !m.source) continue;
bool playing = false, fading = false;
ALint pos = 0;
ALint state = AL_INITIAL;
alGetSourcei(m.source, AL_SAMPLE_OFFSET, &pos);
alGetSourcei(m.source, AL_SOURCE_STATE, &state);
if (!_checkALError()) {
m.state = AudioPlayerStopped;
emit error(m.audio);
} else {
switch (m.state) {
case AudioPlayerFinishing:
case AudioPlayerPausing:
case AudioPlayerStarting:
case AudioPlayerResuming:
fading = true;
break;
case AudioPlayerPlaying:
playing = true;
break;
}
if (fading && (state == AL_PLAYING || !m.loading)) {
if (state != AL_PLAYING) {
fading = false;
if (m.source) {
alSourcef(m.source, AL_GAIN, 1);
alSourceStop(m.source);
}
m.state = AudioPlayerStopped;
emit audioStopped(m.audio);
} else if (1000 * (pos + m.skipStart - m.started) >= AudioFadeDuration * m.frequency) {
fading = false;
alSourcef(m.source, AL_GAIN, 1);
switch (m.state) {
case AudioPlayerFinishing: alSourceStop(m.source); m.state = AudioPlayerStopped; break;
case AudioPlayerPausing: alSourcePause(m.source); m.state = AudioPlayerPaused; break;
case AudioPlayerStarting:
case AudioPlayerResuming:
m.state = AudioPlayerPlaying;
playing = true;
break;
}
} else {
float64 newGain = 1000. * (pos + m.skipStart - m.started) / (AudioFadeDuration * m.frequency);
if (m.state == AudioPlayerPausing || m.state == AudioPlayerFinishing) {
newGain = 1. - newGain;
}
alSourcef(m.source, AL_GAIN, newGain);
}
} else if (playing && (state == AL_PLAYING || !m.loading)) {
if (state != AL_PLAYING) {
playing = false;
if (m.source) {
alSourceStop(m.source);
alSourcef(m.source, AL_GAIN, 1);
}
m.state = AudioPlayerStopped;
emit audioStopped(m.audio);
}
}
if (pos + m.skipStart - m.position >= AudioCheckPositionDelta) {
m.position = pos + m.skipStart;
emit playPositionUpdated(m.audio);
}
if (!m.loading && m.skipEnd > 0 && m.position + AudioPreloadSamples + m.skipEnd > m.duration) {
m.loading = true;
emit needToPreload(m.audio);
}
if (playing) hasPlaying = true;
if (fading) hasFading = true;
}
}
if (!hasPlaying) {
ALint state = AL_INITIAL;
alGetSourcei(notifySource, AL_SOURCE_STATE, &state);
if (_checkALError() && state == AL_PLAYING) {
hasPlaying = true;
}
}
if (hasFading) {
_timer.start(AudioFadeTimeout);
processContext();
} else if (hasPlaying) {
_timer.start(AudioCheckPositionTimeout);
processContext();
} else {
QMutexLocker lock(&_suspendMutex);
_suspendFlag = true;
_suspendTimer.start(AudioSuspendTimeout);
}
}
void AudioPlayerFader::onSuspendTimer() {
QMutexLocker lock(&_suspendMutex);
if (_suspendFlag) {
alcSuspendContext(audioContext);
}
}
void AudioPlayerFader::onSuspendTimerStop() {
if (_suspendTimer.isActive()) _suspendTimer.stop();
}
void AudioPlayerFader::processContext() {
QMutexLocker lock(&_suspendMutex);
_suspendFlag = false;
emit stopSuspend();
alcProcessContext(audioContext);
}
class AudioPlayerLoader {
public:
AudioPlayerLoader(const QString &fname, const QByteArray &data) : fname(fname), data(data), dataPos(0) {
}
virtual ~AudioPlayerLoader() {
}
bool check(const QString &fname, const QByteArray &data) {
return this->fname == fname && this->data.size() == data.size();
}
virtual bool open() = 0;
virtual int64 duration() = 0;
virtual int32 frequency() = 0;
virtual int32 format() = 0;
virtual void started() = 0;
virtual bool readMore(QByteArray &result, int64 &samplesAdded) = 0;
protected:
QString fname;
QByteArray data;
QFile f;
int32 dataPos;
bool openFile() {
if (data.isEmpty()) {
if (f.isOpen()) f.close();
f.setFileName(fname);
if (!f.open(QIODevice::ReadOnly)) {
LOG(("Audio Error: could not open file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(f.error()).arg(f.errorString()));
return false;
}
}
dataPos = 0;
return true;
}
};
static const uint32 AVBlockSize = 4096; // 4Kb
static const AVSampleFormat _toFormat = AV_SAMPLE_FMT_S16;
static const int64_t _toChannelLayout = AV_CH_LAYOUT_STEREO;
static const int32 _toChannels = 2;
class FFMpegLoader : public AudioPlayerLoader {
public:
FFMpegLoader(const QString &fname, const QByteArray &data) : AudioPlayerLoader(fname, data),
freq(AudioVoiceMsgFrequency), fmt(AL_FORMAT_STEREO16),
sampleSize(2 * sizeof(short)), srcRate(AudioVoiceMsgFrequency), dstRate(AudioVoiceMsgFrequency),
maxResampleSamples(1024), dstSamplesData(0), len(0),
ioBuffer(0), ioContext(0), fmtContext(0), codec(0), codecContext(0), streamId(0), frame(0), swrContext(0),
_opened(false) {
frame = av_frame_alloc();
}
bool open() {
if (!AudioPlayerLoader::openFile()) {
return false;
}
ioBuffer = (uchar*)av_malloc(AVBlockSize);
if (data.isEmpty()) {
ioContext = avio_alloc_context(ioBuffer, AVBlockSize, 0, static_cast<void*>(this), &FFMpegLoader::_read_file, 0, &FFMpegLoader::_seek_file);
} else {
ioContext = avio_alloc_context(ioBuffer, AVBlockSize, 0, static_cast<void*>(this), &FFMpegLoader::_read_data, 0, &FFMpegLoader::_seek_data);
}
fmtContext = avformat_alloc_context();
if (!fmtContext) {
LOG(("Audio Error: Unable to avformat_alloc_context for file '%1', data size '%2'").arg(fname).arg(data.size()));
return false;
}
fmtContext->pb = ioContext;
int res = 0;
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
if ((res = avformat_open_input(&fmtContext, 0, 0, 0)) < 0) {
LOG(("Audio Error: Unable to avformat_open_input for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
_opened = true;
if ((res = avformat_find_stream_info(fmtContext, 0)) < 0) {
LOG(("Audio Error: Unable to avformat_find_stream_info for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
streamId = av_find_best_stream(fmtContext, AVMEDIA_TYPE_AUDIO, -1, -1, &codec, 0);
if (streamId < 0) {
LOG(("Audio Error: Unable to av_find_best_stream for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(streamId).arg(av_make_error_string(err, sizeof(err), streamId)));
return false;
}
// Get a pointer to the codec context for the audio stream
codecContext = fmtContext->streams[streamId]->codec;
av_opt_set_int(codecContext, "refcounted_frames", 1, 0);
if ((res = avcodec_open2(codecContext, codec, 0)) < 0) {
LOG(("Audio Error: Unable to avcodec_open2 for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
freq = fmtContext->streams[streamId]->codec->sample_rate;
len = (fmtContext->streams[streamId]->duration * freq) / fmtContext->streams[streamId]->time_base.den;
uint64_t layout = fmtContext->streams[streamId]->codec->channel_layout;
inputFormat = fmtContext->streams[streamId]->codec->sample_fmt;
switch (layout) {
case AV_CH_LAYOUT_MONO:
switch (inputFormat) {
case AV_SAMPLE_FMT_U8:
case AV_SAMPLE_FMT_U8P: fmt = AL_FORMAT_MONO8; sampleSize = 1; break;
case AV_SAMPLE_FMT_S16:
case AV_SAMPLE_FMT_S16P: fmt = AL_FORMAT_MONO16; sampleSize = 2; break;
default:
sampleSize = -1; // convert needed
break;
}
break;
case AV_CH_LAYOUT_STEREO:
switch (inputFormat) {
case AV_SAMPLE_FMT_U8: fmt = AL_FORMAT_STEREO8; sampleSize = sizeof(short); break;
case AV_SAMPLE_FMT_S16: fmt = AL_FORMAT_STEREO16; sampleSize = 2 * sizeof(short); break;
default:
sampleSize = -1; // convert needed
break;
}
break;
default:
sampleSize = -1; // convert needed
break;
}
if (freq != 44100 && freq != 48000) {
sampleSize = -1; // convert needed
}
if (sampleSize < 0) {
swrContext = swr_alloc();
if (!swrContext) {
LOG(("Audio Error: Unable to swr_alloc for file '%1', data size '%2'").arg(fname).arg(data.size()));
return false;
}
int64_t src_ch_layout = layout, dst_ch_layout = _toChannelLayout;
srcRate = freq;
AVSampleFormat src_sample_fmt = inputFormat, dst_sample_fmt = _toFormat;
dstRate = (freq != 44100 && freq != 48000) ? AudioVoiceMsgFrequency : freq;
av_opt_set_int(swrContext, "in_channel_layout", src_ch_layout, 0);
av_opt_set_int(swrContext, "in_sample_rate", srcRate, 0);
av_opt_set_sample_fmt(swrContext, "in_sample_fmt", src_sample_fmt, 0);
av_opt_set_int(swrContext, "out_channel_layout", dst_ch_layout, 0);
av_opt_set_int(swrContext, "out_sample_rate", dstRate, 0);
av_opt_set_sample_fmt(swrContext, "out_sample_fmt", dst_sample_fmt, 0);
if ((res = swr_init(swrContext)) < 0) {
LOG(("Audio Error: Unable to swr_init for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
sampleSize = _toChannels * sizeof(short);
freq = dstRate;
len = av_rescale_rnd(len, dstRate, srcRate, AV_ROUND_UP);
fmt = AL_FORMAT_STEREO16;
maxResampleSamples = av_rescale_rnd(AVBlockSize / sampleSize, dstRate, srcRate, AV_ROUND_UP);
if ((res = av_samples_alloc_array_and_samples(&dstSamplesData, 0, _toChannels, maxResampleSamples, _toFormat, 0)) < 0) {
LOG(("Audio Error: Unable to av_samples_alloc for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
}
return true;
}
int64 duration() {
return len;
}
int32 frequency() {
return freq;
}
int32 format() {
return fmt;
}
void started() {
}
bool readMore(QByteArray &result, int64 &samplesAdded) {
int res;
if ((res = av_read_frame(fmtContext, &avpkt)) < 0) {
if (res != AVERROR_EOF) {
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
LOG(("Audio Error: Unable to av_read_frame() file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
}
return false;
}
if (avpkt.stream_index == streamId) {
avcodec_get_frame_defaults(frame);
int got_frame = 0;
if ((res = avcodec_decode_audio4(codecContext, frame, &got_frame, &avpkt)) < 0) {
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
LOG(("Audio Error: Unable to avcodec_decode_audio4() file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
if (got_frame) {
if (dstSamplesData) { // convert needed
int64_t dstSamples = av_rescale_rnd(swr_get_delay(swrContext, srcRate) + frame->nb_samples, dstRate, srcRate, AV_ROUND_UP);
if (dstSamples > maxResampleSamples) {
maxResampleSamples = dstSamples;
av_free(dstSamplesData[0]);
if ((res = av_samples_alloc(dstSamplesData, 0, _toChannels, maxResampleSamples, _toFormat, 1)) < 0) {
dstSamplesData[0] = 0;
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
LOG(("Audio Error: Unable to av_samples_alloc for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
}
if ((res = swr_convert(swrContext, dstSamplesData, dstSamples, (const uint8_t**)frame->extended_data, frame->nb_samples)) < 0) {
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
LOG(("Audio Error: Unable to swr_convert for file '%1', data size '%2', error %3, %4").arg(fname).arg(data.size()).arg(res).arg(av_make_error_string(err, sizeof(err), res)));
return false;
}
int32 resultLen = av_samples_get_buffer_size(0, _toChannels, res, _toFormat, 1);
result.append((const char*)dstSamplesData[0], resultLen);
samplesAdded += resultLen / sampleSize;
} else {
result.append((const char*)frame->extended_data[0], frame->nb_samples * sampleSize);
samplesAdded += frame->nb_samples;
}
}
}
av_free_packet(&avpkt);
return true;
}
~FFMpegLoader() {
if (ioContext) av_free(ioContext);
if (codecContext) avcodec_close(codecContext);
if (swrContext) swr_free(&swrContext);
if (dstSamplesData) {
if (dstSamplesData[0]) {
av_freep(&dstSamplesData[0]);
}
av_freep(&dstSamplesData);
}
if (_opened) {
avformat_close_input(&fmtContext);
} else if (ioBuffer) {
av_free(ioBuffer);
}
if (fmtContext) avformat_free_context(fmtContext);
av_frame_free(&frame);
}
private:
int32 freq, fmt, channels;
int32 sampleSize, srcRate, dstRate, maxResampleSamples;
uint8_t **dstSamplesData;
int64 len;
uchar *ioBuffer;
AVIOContext *ioContext;
AVFormatContext *fmtContext;
AVCodec *codec;
AVCodecContext *codecContext;
AVPacket avpkt;
int32 streamId;
AVSampleFormat inputFormat;
AVFrame *frame;
SwrContext *swrContext;
bool _opened;
static int _read_data(void *opaque, uint8_t *buf, int buf_size) {
FFMpegLoader *l = reinterpret_cast<FFMpegLoader*>(opaque);
int32 nbytes = qMin(l->data.size() - l->dataPos, int32(buf_size));
if (nbytes <= 0) {
return 0;
}
memcpy(buf, l->data.constData() + l->dataPos, nbytes);
l->dataPos += nbytes;
return nbytes;
}
static int64_t _seek_data(void *opaque, int64_t offset, int whence) {
FFMpegLoader *l = reinterpret_cast<FFMpegLoader*>(opaque);
int32 newPos = -1;
switch (whence) {
case SEEK_SET: newPos = offset; break;
case SEEK_CUR: newPos = l->dataPos + offset; break;
case SEEK_END: newPos = l->data.size() + offset; break;
}
if (newPos < 0 || newPos > l->data.size()) {
return -1;
}
l->dataPos = newPos;
return l->dataPos;
}
static int _read_file(void *opaque, uint8_t *buf, int buf_size) {
FFMpegLoader *l = reinterpret_cast<FFMpegLoader*>(opaque);
return int(l->f.read((char*)(buf), buf_size));
}
static int64_t _seek_file(void *opaque, int64_t offset, int whence) {
FFMpegLoader *l = reinterpret_cast<FFMpegLoader*>(opaque);
switch (whence) {
case SEEK_SET: return l->f.seek(offset) ? l->f.pos() : -1;
case SEEK_CUR: return l->f.seek(l->f.pos() + offset) ? l->f.pos() : -1;
case SEEK_END: return l->f.seek(l->f.size() + offset) ? l->f.pos() : -1;
}
return -1;
}
};
AudioPlayerLoaders::AudioPlayerLoaders(QThread *thread) {
moveToThread(thread);
}
AudioPlayerLoaders::~AudioPlayerLoaders() {
for (Loaders::iterator i = _loaders.begin(), e = _loaders.end(); i != e; ++i) {
delete i.value();
}
_loaders.clear();
}
void AudioPlayerLoaders::onInit() {
}
void AudioPlayerLoaders::onStart(AudioData *audio) {
Loaders::iterator i = _loaders.find(audio);
if (i != _loaders.end()) {
delete (*i);
_loaders.erase(i);
}
onLoad(audio);
}
void AudioPlayerLoaders::loadError(Loaders::iterator i) {
emit error(i.key());
delete (*i);
_loaders.erase(i);
}
void AudioPlayerLoaders::onLoad(AudioData *audio) {
bool started = false;
int32 audioindex = -1;
AudioPlayerLoader *l = 0;
Loaders::iterator j = _loaders.end();
{
QMutexLocker lock(&playerMutex);
AudioPlayer *voice = audioPlayer();
if (!voice) return;
for (int32 i = 0; i < AudioVoiceMsgSimultaneously; ++i) {
AudioPlayer::Msg &m(voice->_data[i]);
if (m.audio != audio || !m.loading) continue;
audioindex = i;
j = _loaders.find(audio);
if (j != _loaders.end() && !j.value()->check(m.fname, m.data)) {
delete j.value();
_loaders.erase(j);
j = _loaders.end();
}
if (j == _loaders.end()) {
QByteArray header = m.data.mid(0, 8);
if (header.isEmpty()) {
QFile f(m.fname);
if (!f.open(QIODevice::ReadOnly)) {
LOG(("Audio Error: could not open file '%1'").arg(m.fname));
m.state = AudioPlayerStoppedAtStart;
emit error(audio);
return;
}
header = f.read(8);
}
if (header.size() < 8) {
LOG(("Audio Error: could not read header from file '%1', data size %2").arg(m.fname).arg(m.data.isEmpty() ? QFileInfo(m.fname).size() : m.data.size()));
m.state = AudioPlayerStoppedAtStart;
emit error(audio);
return;
}
l = (j = _loaders.insert(audio, new FFMpegLoader(m.fname, m.data))).value();
int ret;
if (!l->open()) {
m.state = AudioPlayerStoppedAtStart;
return loadError(j);
}
int64 duration = l->duration();
if (duration <= 0) {
m.state = AudioPlayerStoppedAtStart;
return loadError(j);
}
m.duration = duration;
m.frequency = l->frequency();
if (!m.frequency) m.frequency = AudioVoiceMsgFrequency;
m.skipStart = 0;
m.skipEnd = duration;
m.position = 0;
m.started = 0;
started = true;
} else {
if (!m.skipEnd) continue;
l = j.value();
}
break;
}
}
if (j == _loaders.end()) {
LOG(("Audio Error: trying to load part of audio, that is not playing at the moment"));
emit error(audio);
return;
}
if (started) {
l->started();
}
bool finished = false;
QByteArray result;
int64 samplesAdded = 0, frequency = l->frequency(), format = l->format();
while (result.size() < AudioVoiceMsgBufferSize) {
if (!l->readMore(result, samplesAdded)) {
finished = true;
break;
}
{
QMutexLocker lock(&playerMutex);
AudioPlayer *voice = audioPlayer();
if (!voice) return;
AudioPlayer::Msg &m(voice->_data[audioindex]);
if (m.audio != audio || !m.loading || !l->check(m.fname, m.data)) {
LOG(("Audio Error: playing changed while loading"));
m.state = AudioPlayerStopped;
return loadError(j);
}
}
}
QMutexLocker lock(&playerMutex);
AudioPlayer *voice = audioPlayer();
if (!voice) return;
AudioPlayer::Msg &m(voice->_data[audioindex]);
if (m.audio != audio || !m.loading || !l->check(m.fname, m.data)) {
LOG(("Audio Error: playing changed while loading"));
m.state = AudioPlayerStopped;
return loadError(j);
}
if (started) {
if (m.source) {
alSourceStop(m.source);
for (int32 i = 0; i < 3; ++i) {
if (m.samplesCount[i]) {
alSourceUnqueueBuffers(m.source, 1, m.buffers + i);
m.samplesCount[i] = 0;
}
}
m.nextBuffer = 0;
}
}
if (samplesAdded) {
if (!m.source) {
alGenSources(1, &m.source);
alSourcef(m.source, AL_PITCH, 1.f);
alSourcef(m.source, AL_GAIN, 1.f);
alSource3f(m.source, AL_POSITION, 0, 0, 0);
alSource3f(m.source, AL_VELOCITY, 0, 0, 0);
alSourcei(m.source, AL_LOOPING, 0);
}
if (!m.buffers[m.nextBuffer]) alGenBuffers(3, m.buffers);
if (!_checkALError()) {
m.state = AudioPlayerStopped;
return loadError(j);
}
if (m.samplesCount[m.nextBuffer]) {
alSourceUnqueueBuffers(m.source, 1, m.buffers + m.nextBuffer);
m.skipStart += m.samplesCount[m.nextBuffer];
}
m.samplesCount[m.nextBuffer] = samplesAdded;
alBufferData(m.buffers[m.nextBuffer], format, result.constData(), result.size(), frequency);
alSourceQueueBuffers(m.source, 1, m.buffers + m.nextBuffer);
m.skipEnd -= samplesAdded;
m.nextBuffer = (m.nextBuffer + 1) % 3;
if (!_checkALError()) {
m.state = AudioPlayerStopped;
return loadError(j);
}
} else {
finished = true;
}
if (finished) {
m.skipEnd = 0;
m.duration = m.skipStart + m.samplesCount[0] + m.samplesCount[1] + m.samplesCount[2];
}
m.loading = false;
if (m.state == AudioPlayerResuming || m.state == AudioPlayerPlaying || m.state == AudioPlayerStarting) {
ALint state = AL_INITIAL;
alGetSourcei(m.source, AL_SOURCE_STATE, &state);
if (_checkALError()) {
if (state != AL_PLAYING) {
voice->processContext();
alSourcePlay(m.source);
emit needToCheck();
}
}
}
}
void AudioPlayerLoaders::onCancel(AudioData *audio) {
Loaders::iterator i = _loaders.find(audio);
if (i != _loaders.end()) {
delete (*i);
_loaders.erase(i);
}
QMutexLocker lock(&playerMutex);
AudioPlayer *voice = audioPlayer();
if (!voice) return;
for (int32 i = 0; i < AudioVoiceMsgSimultaneously; ++i) {
AudioPlayer::Msg &m(voice->_data[i]);
if (m.audio == audio) {
m.loading = false;
}
}
}
struct AudioCapturePrivate {
AudioCapturePrivate() :
device(0), fmt(0), ioBuffer(0), ioContext(0), fmtContext(0), stream(0), codec(0), codecContext(0), opened(false),
srcSamples(0), dstSamples(0), maxDstSamples(0), dstSamplesSize(0), fullSamples(0), srcSamplesData(0), dstSamplesData(0),
swrContext(0), lastUpdate(0), level(0), dataPos(0) {
}
ALCdevice *device;
AVOutputFormat *fmt;
uchar *ioBuffer;
AVIOContext *ioContext;
AVFormatContext *fmtContext;
AVStream *stream;
AVCodec *codec;
AVCodecContext *codecContext;
bool opened;
int32 srcSamples, dstSamples, maxDstSamples, dstSamplesSize, fullSamples;
uint8_t **srcSamplesData, **dstSamplesData;
SwrContext *swrContext;
int32 lastUpdate;
int64 level;
QByteArray data;
int32 dataPos;
static int _read_data(void *opaque, uint8_t *buf, int buf_size) {
AudioCapturePrivate *l = reinterpret_cast<AudioCapturePrivate*>(opaque);
int32 nbytes = qMin(l->data.size() - l->dataPos, int32(buf_size));
if (nbytes <= 0) {
return 0;
}
memcpy(buf, l->data.constData() + l->dataPos, nbytes);
l->dataPos += nbytes;
return nbytes;
}
static int _write_data(void *opaque, uint8_t *buf, int buf_size) {
AudioCapturePrivate *l = reinterpret_cast<AudioCapturePrivate*>(opaque);
if (buf_size <= 0) return 0;
if (l->dataPos + buf_size > l->data.size()) l->data.resize(l->dataPos + buf_size);
memcpy(l->data.data() + l->dataPos, buf, buf_size);
l->dataPos += buf_size;
return buf_size;
}
static int64_t _seek_data(void *opaque, int64_t offset, int whence) {
AudioCapturePrivate *l = reinterpret_cast<AudioCapturePrivate*>(opaque);
int32 newPos = -1;
switch (whence) {
case SEEK_SET: newPos = offset; break;
case SEEK_CUR: newPos = l->dataPos + offset; break;
case SEEK_END: newPos = l->data.size() + offset; break;
}
if (newPos < 0) {
return -1;
}
l->dataPos = newPos;
return l->dataPos;
}
};
AudioCaptureInner::AudioCaptureInner(QThread *thread) : d(new AudioCapturePrivate()) {
moveToThread(thread);
_timer.moveToThread(thread);
connect(&_timer, SIGNAL(timeout()), this, SLOT(onTimeout()));
}
AudioCaptureInner::~AudioCaptureInner() {
onStop(false);
delete d;
}
void AudioCaptureInner::onInit() {
}
void AudioCaptureInner::onStart() {
// Start OpenAL Capture
d->device = alcCaptureOpenDevice(0, AudioVoiceMsgFrequency, AL_FORMAT_MONO16, AudioVoiceMsgBufferSize);
if (!d->device) {
LOG(("Audio Error: capture device not present!"));
emit error();
return;
}
alcCaptureStart(d->device);
if (!_checkCaptureError(d->device)) {
alcCaptureCloseDevice(d->device);
d->device = 0;
emit error();
return;
}
// Create encoding context
d->ioBuffer = (uchar*)av_malloc(AVBlockSize);
d->ioContext = avio_alloc_context(d->ioBuffer, AVBlockSize, 1, static_cast<void*>(d), &AudioCapturePrivate::_read_data, &AudioCapturePrivate::_write_data, &AudioCapturePrivate::_seek_data);
int res = 0;
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
AVOutputFormat *fmt = 0;
while ((fmt = av_oformat_next(fmt))) {
if (fmt->name == QLatin1String("opus")) {
break;
}
}
if (!fmt) {
LOG(("Audio Error: Unable to find opus AVOutputFormat for capture"));
onStop(false);
emit error();
return;
}
if ((res = avformat_alloc_output_context2(&d->fmtContext, fmt, 0, 0)) < 0) {
LOG(("Audio Error: Unable to avformat_alloc_output_context2 for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
d->fmtContext->pb = d->ioContext;
d->fmtContext->flags |= AVFMT_FLAG_CUSTOM_IO;
d->opened = true;
// Add audio stream
d->codec = avcodec_find_encoder(fmt->audio_codec);
if (!d->codec) {
LOG(("Audio Error: Unable to avcodec_find_encoder for capture"));
onStop(false);
emit error();
return;
}
d->stream = avformat_new_stream(d->fmtContext, d->codec);
if (!d->stream) {
LOG(("Audio Error: Unable to avformat_new_stream for capture"));
onStop(false);
emit error();
return;
}
d->stream->id = d->fmtContext->nb_streams - 1;
d->codecContext = d->stream->codec;
av_opt_set_int(d->codecContext, "refcounted_frames", 1, 0);
d->codecContext->sample_fmt = AV_SAMPLE_FMT_FLTP;
d->codecContext->bit_rate = 0;// 64000;
d->codecContext->sample_rate = AudioVoiceMsgFrequency;
d->codecContext->channels = 1;
if (d->fmtContext->oformat->flags & AVFMT_GLOBALHEADER) {
d->codecContext->flags |= CODEC_FLAG_GLOBAL_HEADER;
}
// Open audio stream
if ((res = avcodec_open2(d->codecContext, d->codec, NULL)) < 0) {
LOG(("Audio Error: Unable to avcodec_open2 for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
// Alloc source samples
d->srcSamples = (d->codecContext->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE) ? 10000 : d->codecContext->frame_size;
//if ((res = av_samples_alloc_array_and_samples(&d->srcSamplesData, 0, d->codecContext->channels, d->srcSamples, d->codecContext->sample_fmt, 0)) < 0) {
// LOG(("Audio Error: Unable to av_samples_alloc_array_and_samples for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
// onStop(false);
// emit error();
// return;
//}
// Using _captured directly
// Prepare resampling
d->swrContext = swr_alloc();
if (!d->swrContext) {
fprintf(stderr, "Could not allocate resampler context\n");
exit(1);
}
av_opt_set_int(d->swrContext, "in_channel_count", d->codecContext->channels, 0);
av_opt_set_int(d->swrContext, "in_sample_rate", d->codecContext->sample_rate, 0);
av_opt_set_sample_fmt(d->swrContext, "in_sample_fmt", AV_SAMPLE_FMT_S16, 0);
av_opt_set_int(d->swrContext, "out_channel_count", d->codecContext->channels, 0);
av_opt_set_int(d->swrContext, "out_sample_rate", d->codecContext->sample_rate, 0);
av_opt_set_sample_fmt(d->swrContext, "out_sample_fmt", d->codecContext->sample_fmt, 0);
if ((res = swr_init(d->swrContext)) < 0) {
LOG(("Audio Error: Unable to swr_init for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
d->maxDstSamples = d->srcSamples;
if ((res = av_samples_alloc_array_and_samples(&d->dstSamplesData, 0, d->codecContext->channels, d->maxDstSamples, d->codecContext->sample_fmt, 0)) < 0) {
LOG(("Audio Error: Unable to av_samples_alloc_array_and_samples for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
d->dstSamplesSize = av_samples_get_buffer_size(0, d->codecContext->channels, d->maxDstSamples, d->codecContext->sample_fmt, 0);
// Write file header
if ((res = avformat_write_header(d->fmtContext, 0)) < 0) {
LOG(("Audio Error: Unable to avformat_write_header for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
_timer.start(50);
_captured.clear();
_captured.reserve(AudioVoiceMsgBufferSize);
DEBUG_LOG(("Audio Capture: started!"));
}
void AudioCaptureInner::onStop(bool needResult) {
if (!_timer.isActive()) return; // in onStop() already
_timer.stop();
if (needResult) {
onTimeout(); // get last data
}
// Write what is left
if (!_captured.isEmpty()) {
int32 fadeSamples = AudioVoiceMsgFade * AudioVoiceMsgFrequency / 1000, capturedSamples = _captured.size() / sizeof(short);
if ((_captured.size() % sizeof(short)) || (d->fullSamples + capturedSamples < AudioVoiceMsgFrequency) || (capturedSamples < fadeSamples)) {
d->fullSamples = 0;
d->dataPos = 0;
d->data.clear();
} else {
float64 coef = 1. / fadeSamples, fadedFrom = 0;
for (short *ptr = ((short*)_captured.data()) + capturedSamples, *end = ptr - fadeSamples; ptr != end; ++fadedFrom) {
--ptr;
*ptr = qRound(fadedFrom * coef * *ptr);
}
if (capturedSamples % d->srcSamples) {
int32 s = _captured.size();
_captured.resize(s + (d->srcSamples - (capturedSamples % d->srcSamples)) * sizeof(short));
memset(_captured.data() + s, 0, _captured.size() - s);
}
int32 framesize = d->srcSamples * d->codecContext->channels * sizeof(short), encoded = 0;
while (_captured.size() >= encoded + framesize) {
writeFrame(encoded, framesize);
encoded += framesize;
}
if (encoded != _captured.size()) {
d->fullSamples = 0;
d->dataPos = 0;
d->data.clear();
}
}
}
DEBUG_LOG(("Audio Capture: stopping (need result: %1), size: %2, samples: %3").arg(logBool(needResult)).arg(d->data.size()).arg(d->fullSamples));
_captured = QByteArray();
// Finish stream
if (d->device) {
av_write_trailer(d->fmtContext);
}
QByteArray result = d->fullSamples ? d->data : QByteArray();
qint32 samples = d->fullSamples;
if (d->device) {
alcCaptureStop(d->device);
alcCaptureCloseDevice(d->device);
d->device = 0;
if (d->ioContext) {
av_free(d->ioContext);
d->ioContext = 0;
}
if (d->codecContext) {
avcodec_close(d->codecContext);
d->codecContext = 0;
}
if (d->srcSamplesData) {
if (d->srcSamplesData[0]) {
av_freep(&d->srcSamplesData[0]);
}
av_freep(&d->srcSamplesData);
}
if (d->dstSamplesData) {
if (d->dstSamplesData[0]) {
av_freep(&d->dstSamplesData[0]);
}
av_freep(&d->dstSamplesData);
}
d->fullSamples = 0;
if (d->swrContext) {
swr_free(&d->swrContext);
d->swrContext = 0;
}
if (d->opened) {
avformat_close_input(&d->fmtContext);
d->opened = false;
d->ioBuffer = 0;
} else if (d->ioBuffer) {
av_free(d->ioBuffer);
d->ioBuffer = 0;
}
if (d->fmtContext) {
avformat_free_context(d->fmtContext);
d->fmtContext = 0;
}
d->fmt = 0;
d->stream = 0;
d->codec = 0;
d->lastUpdate = 0;
d->level = 0;
d->dataPos = 0;
d->data.clear();
}
if (needResult) emit done(result, samples);
}
void AudioCaptureInner::onTimeout() {
if (!d->device) {
_timer.stop();
return;
}
ALint samples;
alcGetIntegerv(d->device, ALC_CAPTURE_SAMPLES, sizeof(samples), &samples);
if (!_checkCaptureError(d->device)) {
onStop(false);
emit error();
return;
}
if (samples > 0) {
// Get samples from OpenAL
int32 s = _captured.size(), news = s + samples * sizeof(short);
if (news / AudioVoiceMsgBufferSize > s / AudioVoiceMsgBufferSize) {
_captured.reserve(((news / AudioVoiceMsgBufferSize) + 1) * AudioVoiceMsgBufferSize);
}
_captured.resize(news);
alcCaptureSamples(d->device, (ALCvoid *)(_captured.data() + s), samples);
if (!_checkCaptureError(d->device)) {
onStop(false);
emit error();
return;
}
// Count new recording level and update view
int32 skipSamples = AudioVoiceMsgSkip * AudioVoiceMsgFrequency / 1000, fadeSamples = AudioVoiceMsgFade * AudioVoiceMsgFrequency / 1000;
int32 levelindex = d->fullSamples + (s / sizeof(short));
for (const short *ptr = (const short*)(_captured.constData() + s), *end = (const short*)(_captured.constData() + news); ptr < end; ++ptr, ++levelindex) {
if (levelindex > skipSamples) {
if (levelindex < skipSamples + fadeSamples) {
d->level += qRound(qAbs(*ptr) * float64(levelindex - skipSamples) / fadeSamples);
} else {
d->level += qAbs(*ptr);
}
}
}
qint32 samplesFull = d->fullSamples + _captured.size() / sizeof(short), samplesSinceUpdate = samplesFull - d->lastUpdate;
if (samplesSinceUpdate > AudioVoiceMsgUpdateView * AudioVoiceMsgFrequency / 1000) {
emit update(d->level / samplesSinceUpdate, samplesFull);
d->lastUpdate = samplesFull;
d->level = 0;
}
// Write frames
int32 framesize = d->srcSamples * d->codecContext->channels * sizeof(short), encoded = 0;
while (_captured.size() >= encoded + framesize + fadeSamples * sizeof(short)) {
writeFrame(encoded, framesize);
encoded += framesize;
}
// Collapse the buffer
if (encoded > 0) {
int32 goodSize = _captured.size() - encoded;
memmove(_captured.data(), _captured.constData() + encoded, goodSize);
_captured.resize(goodSize);
}
} else {
DEBUG_LOG(("Audio Capture: no samples to capture."));
}
}
void AudioCaptureInner::writeFrame(int32 offset, int32 framesize) {
// Prepare audio frame
if (framesize % sizeof(short)) { // in the middle of a sample
LOG(("Audio Error: Bad framesize in writeFrame() for capture, framesize %1, %2").arg(framesize));
onStop(false);
emit error();
return;
}
int32 samplesCnt = framesize / sizeof(short);
int res = 0;
char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
short *srcSamplesDataChannel = (short*)(_captured.data() + offset), **srcSamplesData = &srcSamplesDataChannel;
// memcpy(d->srcSamplesData[0], _captured.constData() + offset, framesize);
int32 skipSamples = AudioVoiceMsgSkip * AudioVoiceMsgFrequency / 1000, fadeSamples = AudioVoiceMsgFade * AudioVoiceMsgFrequency / 1000;
if (d->fullSamples < skipSamples + fadeSamples) {
int32 fadedCnt = qMin(samplesCnt, skipSamples + fadeSamples - d->fullSamples);
float64 coef = 1. / fadeSamples, fadedFrom = d->fullSamples - skipSamples;
short *ptr = (short*)srcSamplesData[0], *zeroEnd = ptr + qMin(samplesCnt, qMax(0, skipSamples - d->fullSamples)), *end = ptr + fadedCnt;
for (; ptr != zeroEnd; ++ptr, ++fadedFrom) {
*ptr = 0;
}
for (; ptr != end; ++ptr, ++fadedFrom) {
*ptr = qRound(fadedFrom * coef * *ptr);
}
}
// Convert to final format
d->dstSamples = av_rescale_rnd(swr_get_delay(d->swrContext, d->codecContext->sample_rate) + d->srcSamples, d->codecContext->sample_rate, d->codecContext->sample_rate, AV_ROUND_UP);
if (d->dstSamples > d->maxDstSamples) {
d->maxDstSamples = d->dstSamples;
av_free(d->dstSamplesData[0]);
if ((res = av_samples_alloc(d->dstSamplesData, 0, d->codecContext->channels, d->dstSamples, d->codecContext->sample_fmt, 0)) < 0) {
LOG(("Audio Error: Unable to av_samples_alloc for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
d->dstSamplesSize = av_samples_get_buffer_size(0, d->codecContext->channels, d->maxDstSamples, d->codecContext->sample_fmt, 0);
}
if ((res = swr_convert(d->swrContext, d->dstSamplesData, d->dstSamples, (const uint8_t **)srcSamplesData, d->srcSamples)) < 0) {
LOG(("Audio Error: Unable to swr_convert for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
// Write audio frame
AVPacket pkt = { 0 }; // data and size must be 0;
AVFrame *frame = avcodec_alloc_frame();
int gotPacket;
av_init_packet(&pkt);
frame->nb_samples = d->dstSamples;
avcodec_fill_audio_frame(frame, d->codecContext->channels, d->codecContext->sample_fmt, d->dstSamplesData[0], d->dstSamplesSize, 0);
if ((res = avcodec_encode_audio2(d->codecContext, &pkt, frame, &gotPacket)) < 0) {
LOG(("Audio Error: Unable to avcodec_encode_audio2 for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
if (gotPacket) {
pkt.stream_index = d->stream->index;
if ((res = av_interleaved_write_frame(d->fmtContext, &pkt)) < 0) {
LOG(("Audio Error: Unable to av_interleaved_write_frame for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
onStop(false);
emit error();
return;
}
}
d->fullSamples += samplesCnt;
avcodec_free_frame(&frame);
}
View git blame Copy permalink