Track新增update方法，支持更新宽高采样率等信息 (#2960)

当变分辨率时候，实时更新MP4封装层的参数信息，避免出现封装层与编码（SPS）层视频宽高不一样，造成解码参数错误花屏；同时也支持更新音频采样率等信息。


---------

Co-authored-by: xia-chu <771730766@qq.com>

src/Common/MultiMediaSourceMuxer.cpp

@@ -45,6 +45,7 @@ static string getTrackInfoStr(const TrackSource *track_src){
     _StrPrinter codec_info;
     auto tracks = track_src->getTracks(true);
     for (auto &track : tracks) {
+        track->update();
         auto codec_type = track->getTrackType();
         codec_info << track->getCodecName();
         switch (codec_type) {

src/Extension/AAC.cpp

@@ -21,53 +21,56 @@ namespace mediakit{
 #ifndef ENABLE_MP4
 unsigned const samplingFrequencyTable[16] = { 96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050, 16000, 12000, 11025, 8000, 7350, 0, 0, 0 };
 
-class AdtsHeader{
+class AdtsHeader {
 public:
-    unsigned int syncword = 0; //12 bslbf 同步字The bit string ‘1111 1111 1111’，说明一个ADTS帧的开始
-    unsigned int id;        //1 bslbf   MPEG 标示符, 设置为1
-    unsigned int layer;    //2 uimsbf Indicates which layer is used. Set to ‘00’
-    unsigned int protection_absent;  //1 bslbf  表示是否误码校验
-    unsigned int profile; //2 uimsbf  表示使用哪个级别的AAC，如01 Low Complexity(LC)--- AACLC
-    unsigned int sf_index;           //4 uimsbf  表示使用的采样率下标
-    unsigned int private_bit;        //1 bslbf
-    unsigned int channel_configuration;  //3 uimsbf  表示声道数
-    unsigned int original;               //1 bslbf
-    unsigned int home;                   //1 bslbf
-    //下面的为改变的参数即每一帧都不同
-    unsigned int copyright_identification_bit;   //1 bslbf
-    unsigned int copyright_identification_start; //1 bslbf
+    unsigned int syncword = 0; // 12 bslbf 同步字The bit string ‘1111 1111 1111’，说明一个ADTS帧的开始
+    unsigned int id; // 1 bslbf   MPEG 标示符, 设置为1
+    unsigned int layer; // 2 uimsbf Indicates which layer is used. Set to ‘00’
+    unsigned int protection_absent; // 1 bslbf  表示是否误码校验
+    unsigned int profile; // 2 uimsbf  表示使用哪个级别的AAC，如01 Low Complexity(LC)--- AACLC
+    unsigned int sf_index; // 4 uimsbf  表示使用的采样率下标
+    unsigned int private_bit; // 1 bslbf
+    unsigned int channel_configuration; // 3 uimsbf  表示声道数
+    unsigned int original; // 1 bslbf
+    unsigned int home; // 1 bslbf
+    // 下面的为改变的参数即每一帧都不同
+    unsigned int copyright_identification_bit; // 1 bslbf
+    unsigned int copyright_identification_start; // 1 bslbf
     unsigned int aac_frame_length; // 13 bslbf  一个ADTS帧的长度包括ADTS头和raw data block
-    unsigned int adts_buffer_fullness;           //11 bslbf     0x7FF 说明是码率可变的码流
-    //no_raw_data_blocks_in_frame 表示ADTS帧中有number_of_raw_data_blocks_in_frame + 1个AAC原始帧.
-    //所以说number_of_raw_data_blocks_in_frame == 0
-    //表示说ADTS帧中有一个AAC数据块并不是说没有。(一个AAC原始帧包含一段时间内1024个采样及相关数据)
-    unsigned int no_raw_data_blocks_in_frame;    //2 uimsfb
+    unsigned int adts_buffer_fullness; // 11 bslbf     0x7FF 说明是码率可变的码流
+    // no_raw_data_blocks_in_frame 表示ADTS帧中有number_of_raw_data_blocks_in_frame + 1个AAC原始帧.
+    // 所以说number_of_raw_data_blocks_in_frame == 0
+    // 表示说ADTS帧中有一个AAC数据块并不是说没有。(一个AAC原始帧包含一段时间内1024个采样及相关数据)
+    unsigned int no_raw_data_blocks_in_frame; // 2 uimsfb
 };
 
 static void dumpAdtsHeader(const AdtsHeader &hed, uint8_t *out) {
-    out[0] = (hed.syncword >> 4 & 0xFF); //8bit
-    out[1] = (hed.syncword << 4 & 0xF0); //4 bit
-    out[1] |= (hed.id << 3 & 0x08); //1 bit
-    out[1] |= (hed.layer << 1 & 0x06); //2bit
-    out[1] |= (hed.protection_absent & 0x01); //1 bit
+    out[0] = (hed.syncword >> 4 & 0xFF); // 8bit
+    out[1] = (hed.syncword << 4 & 0xF0); // 4 bit
+    out[1] |= (hed.id << 3 & 0x08); // 1 bit
+    out[1] |= (hed.layer << 1 & 0x06); // 2bit
+    out[1] |= (hed.protection_absent & 0x01); // 1 bit
     out[2] = (hed.profile << 6 & 0xC0); // 2 bit
-    out[2] |= (hed.sf_index << 2 & 0x3C); //4bit
-    out[2] |= (hed.private_bit << 1 & 0x02); //1 bit
-    out[2] |= (hed.channel_configuration >> 2 & 0x03); //1 bit
+    out[2] |= (hed.sf_index << 2 & 0x3C); // 4bit
+    out[2] |= (hed.private_bit << 1 & 0x02); // 1 bit
+    out[2] |= (hed.channel_configuration >> 2 & 0x03); // 1 bit
     out[3] = (hed.channel_configuration << 6 & 0xC0); // 2 bit
-    out[3] |= (hed.original << 5 & 0x20);				//1 bit
-    out[3] |= (hed.home << 4 & 0x10);				//1 bit
-    out[3] |= (hed.copyright_identification_bit << 3 & 0x08);			//1 bit
-    out[3] |= (hed.copyright_identification_start << 2 & 0x04);		//1 bit
-    out[3] |= (hed.aac_frame_length >> 11 & 0x03);				//2 bit
-    out[4] = (hed.aac_frame_length >> 3 & 0xFF);				//8 bit
-    out[5] = (hed.aac_frame_length << 5 & 0xE0);				//3 bit
-    out[5] |= (hed.adts_buffer_fullness >> 6 & 0x1F);				//5 bit
-    out[6] = (hed.adts_buffer_fullness << 2 & 0xFC);				//6 bit
-    out[6] |= (hed.no_raw_data_blocks_in_frame & 0x03);				//2 bit
+    out[3] |= (hed.original << 5 & 0x20); // 1 bit
+    out[3] |= (hed.home << 4 & 0x10); // 1 bit
+    out[3] |= (hed.copyright_identification_bit << 3 & 0x08); // 1 bit
+    out[3] |= (hed.copyright_identification_start << 2 & 0x04); // 1 bit
+    out[3] |= (hed.aac_frame_length >> 11 & 0x03); // 2 bit
+    out[4] = (hed.aac_frame_length >> 3 & 0xFF); // 8 bit
+    out[5] = (hed.aac_frame_length << 5 & 0xE0); // 3 bit
+    out[5] |= (hed.adts_buffer_fullness >> 6 & 0x1F); // 5 bit
+    out[6] = (hed.adts_buffer_fullness << 2 & 0xFC); // 6 bit
+    out[6] |= (hed.no_raw_data_blocks_in_frame & 0x03); // 2 bit
 }
 
-static void parseAacConfig(const string &config, AdtsHeader &adts) {
+static bool parseAacConfig(const string &config, AdtsHeader &adts) {
+    if (config.size() < 2) {
+        return false;
+    }
     uint8_t cfg1 = config[0];
     uint8_t cfg2 = config[1];
 
@@ -94,6 +97,7 @@ static void parseAacConfig(const string &config, AdtsHeader &adts) {
     adts.aac_frame_length = 7;
     adts.adts_buffer_fullness = 2047;
     adts.no_raw_data_blocks_in_frame = 0;
+    return true;
 }
 #endif// ENABLE_MP4
 
@@ -168,10 +172,12 @@ int dumpAacConfig(const string &config, size_t length, uint8_t *out, size_t out_
 #endif
 }
 
-bool parseAacConfig(const string &config, int &samplerate, int &channels){
+bool parseAacConfig(const string &config, int &samplerate, int &channels) {
 #ifndef ENABLE_MP4
     AdtsHeader header;
-    parseAacConfig(config, header);
+    if (!parseAacConfig(config, header)) {
+        return false;
+    }
     samplerate = samplingFrequencyTable[header.sf_index];
     channels = header.channel_configuration;
     return true;
@@ -326,11 +332,14 @@ bool AACTrack::inputFrame_l(const Frame::Ptr &frame) {
     return false;
 }
 
+bool AACTrack::update() {
+    return parseAacConfig(_cfg, _sampleRate, _channel);
+}
+
 void AACTrack::onReady() {
-    if (_cfg.size() < 2) {
-        return;
+    if (!parseAacConfig(_cfg, _sampleRate, _channel)) {
+        _cfg.clear();
     }
-    parseAacConfig(_cfg, _sampleRate, _channel);
 }
 
 Track::Ptr AACTrack::clone() {
@@ -342,6 +351,7 @@ Sdp::Ptr AACTrack::getSdp() {
         WarnL << getCodecName() << " Track未准备好";
         return nullptr;
     }
+    update();
     return std::make_shared<AACSdp>(getConfig(), getAudioSampleRate(), getAudioChannel(), getBitRate() / 1024);
 }
 

src/Extension/AAC.h

@@ -52,6 +52,7 @@ public:
     int getAudioSampleRate() const override;
     int getAudioSampleBit() const override;
     bool inputFrame(const Frame::Ptr &frame) override;
+    bool update() override;
 
 private:
     void onReady();

src/Extension/H264.cpp

@@ -168,6 +168,10 @@ bool H264Track::inputFrame(const Frame::Ptr &frame) {
     return ret;
 }
 
+bool H264Track::update() {
+    return getAVCInfo(_sps, _width, _height, _fps);
+}
+
 void H264Track::onReady() {
     if (!getAVCInfo(_sps, _width, _height, _fps)) {
         _sps.clear();

src/Extension/H264.h

@@ -128,6 +128,7 @@ public:
     int getVideoWidth() const override;
     float getVideoFps() const override;
     bool inputFrame(const Frame::Ptr &frame) override;
+    bool update() override;
 
 private:
     void onReady();

src/Extension/H265.cpp

@@ -144,6 +144,10 @@ bool H265Track::inputFrame_l(const Frame::Ptr &frame) {
     return ret;
 }
 
+bool H265Track::update() {
+    return getHEVCInfo(_vps, _sps, _width, _height, _fps);
+}
+
 void H265Track::onReady() {
     if (!getHEVCInfo(_vps, _sps, _width, _height, _fps)) {
         _vps.clear();

src/Extension/H265.h

@@ -150,6 +150,7 @@ public:
     int getVideoHeight() const override;
     float getVideoFps() const override;
     bool inputFrame(const Frame::Ptr &frame) override;
+    bool update() override;
 
 private:
     void onReady();

src/Extension/Track.h

@@ -39,6 +39,11 @@ public:
      */
     virtual Track::Ptr clone() = 0;
 
+    /**
+     * 更新track信息，比如触发sps/pps解析
+     */
+    virtual bool update() { return false; }
+
     /**
      * 生成sdp
      * @return  sdp对象

src/Player/PlayerProxy.cpp

@@ -70,6 +70,7 @@ void PlayerProxy::setTranslationInfo()
     _transtalion_info.stream_info.clear();
     auto tracks = _muxer->getTracks();
     for (auto &track : tracks) {
+        track->update();
         _transtalion_info.stream_info.emplace_back();
         auto &back = _transtalion_info.stream_info.back();
         back.bitrate = track->getBitRate();

src/Record/MP4Muxer.cpp

@@ -198,6 +198,7 @@ bool MP4MuxerInterface::addTrack(const Track::Ptr &track) {
         return false;
     }
 
+    track->update();
     switch (track->getCodecId()) {
         case CodecG711A:
         case CodecG711U:

src/Rtmp/Rtmp.cpp

@@ -57,6 +57,7 @@ AudioMeta::AudioMeta(const AudioTrack::Ptr &audio) {
 }
 
 uint8_t getAudioRtmpFlags(const Track::Ptr &track) {
+    track->update();
     switch (track->getTrackType()) {
         case TrackAudio: {
             auto audioTrack = std::dynamic_pointer_cast<AudioTrack>(track);
@@ -115,6 +116,7 @@ uint8_t getAudioRtmpFlags(const Track::Ptr &track) {
 
 void Metadata::addTrack(AMFValue &metadata, const Track::Ptr &track) {
     Metadata::Ptr new_metadata;
+    track->update();
     switch (track->getTrackType()) {
         case TrackVideo: {
             new_metadata = std::make_shared<VideoMeta>(std::dynamic_pointer_cast<VideoTrack>(track));