c++ - 使用 C++/CLR 中的 ffmpeg 库流式传输音频，avcodec_fill_audio_frame 返回 -22 错误代码

Question

早上好。所以我正在做一个项目，我需要使用 FFmpeg 的库通过 RTMP 连接在 C++ 中流式传输实时视频和音频。据我所知，视频和音频使用两种不同的流，所以我现在尝试创建一个纯音频流。

我的问题:调用 FFmpeg 的 avcodec_fill_audio_frame() 时，我收到错误代码 -22。从 FFmpeg 库的文档来看，这告诉我什么都没有……还有很多不足之处。

我的 MainForm.h 有以下(相关)成员

private: NAudio::CoreAudioApi::MMDevice^ waveDevice_chat;
private: NAudio::Wave::WaveIn^ waveInput_chat;
private: NAudio::Wave::IWavePlayer^ waveOutput_chat;
private: NAudio::Wave::BufferedWaveProvider^ waveProvider_chat;

第一步是连接麦克风:

System::Void MainForm::btnConnectMic_Click(System::Object^  sender, System::EventArgs^  e) {
    msclr::interop::marshal_context context;
    waveEnumerator_chat = gcnew NAudio::CoreAudioApi::MMDeviceEnumerator();
    System::Collections::Generic::List<NAudio::CoreAudioApi::MMDevice^>^ wavePorts_chat = System::Linq::Enumerable::ToList(waveEnumerator_chat->EnumerateAudioEndPoints(DataFlow::Capture, DeviceState::Active));

    if (wavePorts_chat->Count > 0) {
        waveDevice_chat = (NAudio::CoreAudioApi::MMDevice^)(wavePorts_chat[0]);
        waveDevice_chat->AudioEndpointVolume->Mute = false;

        waveInput_chat = gcnew WaveIn();
        waveInput_chat->BufferMilliseconds = 50;
        waveInput_chat->DeviceNumber = 0;
        waveInput_chat->WaveFormat = gcnew NAudio::Wave::WaveFormat(44100, 1);
        waveInput_chat->DataAvailable += gcnew System::EventHandler<NAudio::Wave::WaveInEventArgs ^>(this, &MainForm::waveInput_data_available);
        waveInput_chat->StartRecording();

        waveProvider_chat = gcnew BufferedWaveProvider(gcnew NAudio::Wave::WaveFormat(44100, 1));
        waveProvider_chat->DiscardOnBufferOverflow;
    }
}

这是数据可用时由 NAudio 调用的 EventHandler 的代码

void MainForm::waveInput_data_available(System::Object^ sender, WaveInEventArgs^ e) {
    if (waveProvider_chat->BufferedBytes + e->BytesRecorded > waveProvider_chat->BufferLength)
        waveProvider_chat->ClearBuffer();
    else
        waveProvider_chat->AddSamples(e->Buffer, 0, e->BytesRecorded);
}

最后，这是应该填充我的音频帧的代码片段(这是在后台工作程序中循环运行的)

uint8_t* new_buffer;
int result
AVFrame* a_frame = av_frame_alloc();
AVStream* astrm;
AVCodec* acodec = avcodec_find_encoder(AVCodecID::AV_CODEC_ID_AAC);

/*
 *
 *
 *
 */

in the loop

    if (read_buffer->Length <= 0)
        continue;

    new_buffer = (uint8_t*)av_malloc((size_t)waveProvider_chat->BufferedBytes);
    for (int i = 0; i < waveProvider_chat->BufferedBytes; i++)
        new_buffer[i] = (uint8_t)read_buffer[i];

    AVPacket a_pkt;
    av_init_packet(&a_pkt);
    a_pkt.data = nullptr;
    a_pkt.size = 0;
    int got_a_packet = 0;

    int a_encode = avcodec_fill_audio_frame(a_frame, astrm->codec->channels, astrm->codec->sample_fmt, new_buffer, read_buffer->Length, 0);
    std::cout << "[FILL] encoded response: " << a_encode << std::endl;

Answer 1

当您查看 avcodec_fill_audio_frame 的源代码时，你看

if (buf_size < needed_size)
    return AVERROR(EINVAL);

EINVAL 恰好是 22。

所以在你的情况下，我猜，缓冲区不够大。