audio_demo/rtmp_push.cpp

#include <iostream>
#include <unistd.h>
#include <cmath>
#include "timing.h"
#include "log/logger.h"
#include "common.h"

#include <modules/audio_processing/include/audio_processing.h>
#include <modules/audio_processing/include/config.h>
#include "alsa_dev.h"
#include "rnnoise_plugin.h"

using namespace toolkit;

#define MIX_AUDIO_CHANNELS 1
#define MIX_AUDIO_RATE 32000
#define MIX_OUTPUT_RATE 48000
#define MIX_AUDIO_SAMPLES (10 * MIX_AUDIO_RATE / 1000)


struct audio_buf_t
{
    uint8_t* data;
    int index;
    int size;
};

struct RtmpConfig {
    char url[1024];
    AVFormatContext *formatCtx;
    AVStream *stream;
    AVCodecContext *codecCtx;
    SwrContext *swrCtx;
};

struct CallContext {
    RtmpConfig rtmp;
    std::thread *rtmp_thread;
    std::thread *alsa_thread;
    std::mutex *mutex;
    std::vector<audio_buf_t> *list;
    webrtc::AudioProcessing *apm;
    webrtc::AudioProcessing *apm2;
    webrtc::StreamConfig *rtc_stream_config;
    alsa::AlsaDev *alsa;
    alsa::Config alsa_config;
    // rnnoise
    RnNoiseCommonPlugin *rnnoise;
    bool rnnoise_enable;
    float vadThreshold;// (0, 1)
    uint32_t vadGracePeriodBlocks;// (0, 20)
    uint32_t retroactiveVADGraceBlocks;// 0
    //
    bool running;
};

webrtc::AudioProcessing::Config webtcConfigInit()
{
    webrtc::AudioProcessing::Config apmConfig;
    apmConfig.pipeline.maximum_internal_processing_rate = MIX_OUTPUT_RATE;
    apmConfig.pipeline.multi_channel_capture = MIX_AUDIO_CHANNELS > 1 ? true : false;
    apmConfig.pipeline.multi_channel_render = MIX_AUDIO_CHANNELS > 1 ? true : false;
    //PreAmplifier
    apmConfig.pre_amplifier.enabled = false;
    apmConfig.pre_amplifier.fixed_gain_factor = 0.7f;
    //HighPassFilter
    apmConfig.high_pass_filter.enabled = true;
    apmConfig.high_pass_filter.apply_in_full_band = false;
    //EchoCanceller
    apmConfig.echo_canceller.enabled = false;
    apmConfig.echo_canceller.mobile_mode = false;
    apmConfig.echo_canceller.export_linear_aec_output = false;
    apmConfig.echo_canceller.enforce_high_pass_filtering = false;
    //NoiseSuppression
    apmConfig.noise_suppression.enabled = false;
    apmConfig.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kVeryHigh;
    apmConfig.noise_suppression.analyze_linear_aec_output_when_available = false;
    //TransientSuppression
    apmConfig.transient_suppression.enabled = false;
    //VoiceDetection
    apmConfig.voice_detection.enabled = true;
    //GainController1
    apmConfig.gain_controller1.enabled = false;
    // kAdaptiveAnalog 自适应模拟模式
    // kAdaptiveDigital 自适应数字增益模式
    // kFixedDigital 固定数字增益模式
    apmConfig.gain_controller1.mode = webrtc::AudioProcessing::Config::GainController1::kFixedDigital;
    apmConfig.gain_controller1.target_level_dbfs = 6;   // 目标音量
    apmConfig.gain_controller1.compression_gain_db = 60; // 增益能力
    apmConfig.gain_controller1.enable_limiter = true;   // 压限器开关
    apmConfig.gain_controller1.analog_level_minimum = 0;
    apmConfig.gain_controller1.analog_level_maximum = 255;
    apmConfig.gain_controller1.analog_gain_controller.enabled = true;
    // apmConfig.gain_controller1.analog_gain_controller.startup_min_volume = webrtc::kAgcStartupMinVolume;
    apmConfig.gain_controller1.analog_gain_controller.startup_min_volume = 0;
    apmConfig.gain_controller1.analog_gain_controller.clipped_level_min = 0;
    apmConfig.gain_controller1.analog_gain_controller.enable_agc2_level_estimator = false;
    apmConfig.gain_controller1.analog_gain_controller.enable_digital_adaptive = true;
    //GainController2
    apmConfig.gain_controller2.enabled = true;
    apmConfig.gain_controller2.fixed_digital.gain_db = 20.2f;
    apmConfig.gain_controller2.adaptive_digital.enabled = true;
    apmConfig.gain_controller2.adaptive_digital.vad_probability_attack = 1.f;
    apmConfig.gain_controller2.adaptive_digital.level_estimator = webrtc::AudioProcessing::Config::GainController2::kRms;
    apmConfig.gain_controller2.adaptive_digital.level_estimator_adjacent_speech_frames_threshold = 1;
    apmConfig.gain_controller2.adaptive_digital.use_saturation_protector = true;
    apmConfig.gain_controller2.adaptive_digital.initial_saturation_margin_db = 20.f;
    apmConfig.gain_controller2.adaptive_digital.extra_saturation_margin_db = 2.f;
    apmConfig.gain_controller2.adaptive_digital.gain_applier_adjacent_speech_frames_threshold = 1;
    apmConfig.gain_controller2.adaptive_digital.max_gain_change_db_per_second = 3.f;
    apmConfig.gain_controller2.adaptive_digital.max_output_noise_level_dbfs = -50.f;
    //ResidualEchoDetector
    apmConfig.residual_echo_detector.enabled = false;
    //LevelEstimation
    apmConfig.level_estimation.enabled = false;

    return apmConfig;
}

static int64_t t_analyze = 0;
static int64_t t_render = 0;
static int64_t t_capture = 0;
static int64_t t_process = 0;


int pushInit(RtmpConfig *config);
void pushDestory(RtmpConfig *config);
void push_thread(CallContext *ctx);
void capture_thread(CallContext *ctx);

int main()
{
    std::string push_url = "rtmp://192.168.15.248:1935/live/1";

    //初始化日志系统
    Logger::Instance().add(std::make_shared<ConsoleChannel> ());
    Logger::Instance().add(std::make_shared<FileChannel>());
    Logger::Instance().setWriter(std::make_shared<AsyncLogWriter>());

    // webrtc初始化
    webrtc::AudioProcessing *apm = webrtc::AudioProcessingBuilder().Create();
    if (!apm) {
        PrintI("create apm failed.\n");
        return -1;
    }
    // apm 增益
    webrtc::AudioProcessing::Config config = webtcConfigInit();
    apm->ApplyConfig(config);
    apm->Initialize();

    // apm2 降噪
    config.gain_controller1.enabled = false;
    webrtc::AudioProcessing *apm2 = webrtc::AudioProcessingBuilder().Create();
    apm2->ApplyConfig(config);
    apm2->Initialize();

    webrtc::StreamConfig streamConfig;
    streamConfig.set_has_keyboard(false);
    streamConfig.set_num_channels(MIX_AUDIO_CHANNELS);
    streamConfig.set_sample_rate_hz(MIX_OUTPUT_RATE);
    PrintI("webrtc params: {\n%s\n}\n", config.ToString().c_str());

    // rnnoise
    float vad_threshold = 0.85;

    // alsa设备参数
    alsa::Config alsaConfig;
    sprintf(alsaConfig.device, "default");
    alsaConfig.period_time = MIX_AUDIO_SAMPLES * 1000000 / MIX_AUDIO_RATE;
    alsaConfig.buffer_time = 5 * alsaConfig.period_time;
    alsaConfig.channels    = MIX_AUDIO_CHANNELS;
    alsaConfig.format      = SND_PCM_FORMAT_S16_LE;
    alsaConfig.rate        = MIX_AUDIO_RATE;

    // 上下文
    CallContext pushCtx;
    memset(&pushCtx, 0, sizeof(pushCtx));
    strcpy(pushCtx.rtmp.url, push_url.data());
    pushCtx.mutex = new std::mutex;
    pushCtx.list  = new std::vector<audio_buf_t>();
    pushCtx.apm   = apm;
    pushCtx.apm2  = apm2;
    pushCtx.alsa_config = alsaConfig;
    pushCtx.rtc_stream_config = &streamConfig;
    pushCtx.vadThreshold = vad_threshold;
    pushCtx.vadGracePeriodBlocks = 0;
    pushCtx.retroactiveVADGraceBlocks = 0;

    char c;
    bool quit = false;
    /*
    while ((c = getchar()) != EOF && !quit)
    {
        switch (c)
        {
            case 'q': {
                InfoL << "app quit";
                quit = true;
                pushCtx.running = false;
                if (pushCtx.rtmp_thread && pushCtx.rtmp_thread->joinable())
                    pushCtx.rtmp_thread->join();
                if (pushCtx.alsa_thread && pushCtx.alsa_thread->joinable())
                    pushCtx.alsa_thread->join();
                break;
            }
            case 's': {
                InfoL << "start push: " << pushCtx.rtmp.url;
                pushCtx.running = true;
                pushCtx.alsa_thread = new std::thread(capture_thread, &pushCtx);
                pushCtx.rtmp_thread = new std::thread(push_thread, &pushCtx);
                break;
            }
        }
    }
    */
    std::string input_str;
    
    while (getline(std::cin, input_str)) {
        if (input_str == std::string("quit") || 
            input_str == std::string("q")) {
            InfoL << "app quit";
            quit = true;
            pushCtx.running = false;
            if (pushCtx.rtmp_thread && pushCtx.rtmp_thread->joinable())
                pushCtx.rtmp_thread->join();
            if (pushCtx.alsa_thread && pushCtx.alsa_thread->joinable())
                pushCtx.alsa_thread->join();
            break;
        }
        else if (input_str == std::string("s")) {
            InfoL << "start push: " << pushCtx.rtmp.url;
            pushCtx.running = true;
            pushCtx.alsa_thread = new std::thread(capture_thread, &pushCtx);
            pushCtx.rtmp_thread = new std::thread(push_thread, &pushCtx);
        }
        else if (input_str.find("-e") == 0) {
            std::string enable_str = input_str.substr(2, input_str.size() - 2);
            trim(enable_str);
            pushCtx.rnnoise_enable = atoi(enable_str.c_str()) > 0 ? true : false;
            InfoL << "rnnoise enable: " << pushCtx.rnnoise_enable;
        }
        else if (input_str.find("-t") == 0) {
            std::string th_str = input_str.substr(2, input_str.size() - 2);
            trim(th_str);
            int threshold = atoi(th_str.c_str());
            pushCtx.vadThreshold = std::max(std::min(threshold / 100.f, 0.99f), 0.f);
            InfoL << "VAD Threshold(%): " << pushCtx.vadThreshold;
        }
        else if (input_str.find("-p") == 0) {
            std::string period_str = input_str.substr(2, input_str.size() - 2);
            trim(period_str);
            int period = atoi(period_str.c_str());
            pushCtx.vadGracePeriodBlocks = std::max(std::min(period, 20), 0);
            InfoL << "VAD Grace Period (ms): " << pushCtx.vadGracePeriodBlocks;
        }
    }
    InfoL << "push end";
    if (apm) {
        apm->Initialize();
        delete apm;
    }
    for (auto buf: *pushCtx.list) {
        free(buf.data);
    }
    pushCtx.list->clear();
    delete pushCtx.list;
    if (pushCtx.rtmp_thread) delete pushCtx.rtmp_thread;
    if (pushCtx.alsa_thread) delete pushCtx.alsa_thread;
    delete pushCtx.mutex;
    return 0;
}

#include "rnnoise/rnnoise.h"

void capture_thread(CallContext *ctx)
{
    // 声卡初始化
    alsa::AlsaDev usbCaptureDev;
    if (usbCaptureDev.applyConfig(ctx->alsa_config) < 0) {
        PrintE("alsa config failed.\n");
        return ;
    }
    if (usbCaptureDev.init(SND_PCM_STREAM_CAPTURE) < 0) {
        PrintE("alsa init failed.\n");
        return ;
    }
    PrintI("alsa init: %s\n", usbCaptureDev.configToString());
    ctx->alsa = &usbCaptureDev;

    uint8_t *capData = nullptr;
    int buffer_size = usbCaptureDev.getFrames() * usbCaptureDev.getFrameSize();
    capData = (uint8_t *)malloc(buffer_size);
    assert(capData);

    // rnnoise
    ctx->rnnoise = new RnNoiseCommonPlugin(MIX_AUDIO_CHANNELS);
    ctx->rnnoise->init();
    std::vector<float *> in;
    std::vector<float *> out;
    for (int ch = 0; ch < MIX_AUDIO_CHANNELS; ++ch) {
        in.push_back(new float[10 * MIX_OUTPUT_RATE / 1000]);
        out.push_back(new float[10 * MIX_OUTPUT_RATE / 1000]);
    }
    auto rnn = rnnoise_create(NULL);

    // 重采样
    AVFrame *inputFrame = av_frame_alloc();
    {
        inputFrame->sample_rate      = MIX_AUDIO_RATE;
        inputFrame->format           = AV_SAMPLE_FMT_S16;
        inputFrame->channels         = MIX_AUDIO_CHANNELS;
        inputFrame->nb_samples       = MIX_AUDIO_SAMPLES;
        inputFrame->channel_layout   = av_get_default_channel_layout(MIX_AUDIO_CHANNELS);

        int size = av_samples_get_buffer_size(nullptr,
                                          inputFrame->channels, inputFrame->nb_samples, (AVSampleFormat)inputFrame->format, 1);
        uint8_t *buffer = (uint8_t *)av_malloc(size);
        avcodec_fill_audio_frame(inputFrame, inputFrame->channels, (AVSampleFormat)inputFrame->format,
                                (const uint8_t*)buffer, size, 1);
        InfoL << "input frame samples: " << inputFrame->nb_samples
        << ", buffer_size: " << size;
    }
    AVFrame *outputFrame = av_frame_alloc();
    {
        outputFrame->format                 = AV_SAMPLE_FMT_S16;
        outputFrame->channels               = MIX_AUDIO_CHANNELS;
        outputFrame->channel_layout         = av_get_default_channel_layout(MIX_AUDIO_CHANNELS);
        outputFrame->sample_rate            = MIX_OUTPUT_RATE;
        outputFrame->nb_samples             = 10 * MIX_OUTPUT_RATE / 1000;

        int output_bz = av_samples_get_buffer_size(NULL, outputFrame->channels, outputFrame->nb_samples, (AVSampleFormat)outputFrame->format, 0);
        uint8_t *samples_data = (uint8_t *)av_malloc(output_bz);
        avcodec_fill_audio_frame(outputFrame, outputFrame->channels, (AVSampleFormat)outputFrame->format, samples_data, output_bz, 0);
        InfoL << "output frame samples: " << outputFrame->nb_samples
        << ", buffer_size: " << output_bz;
    }
    SwrContext *swrCtx = swr_alloc_set_opts(nullptr, 
                                            outputFrame->channel_layout,
                                            (AVSampleFormat)outputFrame->format,
                                            outputFrame->sample_rate,
                                            inputFrame->channel_layout,
                                            (AVSampleFormat)inputFrame->format,
                                            inputFrame->sample_rate,
                                            0, nullptr);
    if (!swrCtx) {
        PrintE("swr_alloc_set_opts failed.\n");
        return ;
    }
    swr_init(swrCtx);

    int output_size = av_samples_get_buffer_size(NULL, outputFrame->channels, outputFrame->nb_samples, (AVSampleFormat)outputFrame->format, 0);
    uint8_t *output_buf = outputFrame->data[0];


    FILE *input_fp = fopen("/root/rtmp_push_in.pcm", "wb");
    FILE *output_fp = fopen("/root/rtmp_push_out.pcm", "wb");
    while (ctx->running) {
        // 采集
        t_capture = gettimeofday();
        size_t read_size = usbCaptureDev.read(capData, buffer_size);
        // PrintI("alsa read %d\n", read_size);
        if (read_size <= 0) {
            msleep(1);
            continue;
        }

        // 重采样
        memcpy(inputFrame->data[0], capData, buffer_size);
        // fwrite(inputFrame->data[0], 1, buffer_size, input_fp);
        {
            const uint8_t** in = (const uint8_t**)inputFrame->data;
            uint8_t **out = outputFrame->data;

            int len2, out_data_size;

            len2 = swr_convert(swrCtx, out, outputFrame->nb_samples, in, inputFrame->nb_samples);
            if (len2 < 0) {
                printf("swr_convert failed. \n");
                break;
            }
            int out_size = len2;
            while (len2 > 0) {
                len2 = swr_convert(swrCtx, out, outputFrame->nb_samples, nullptr, 0);
                out_size += len2;
            }
            // InfoL << "swr convert output: " << out_size;
        }

        // 降噪
        {
            t_process = gettimeofday();
            ctx->apm->ProcessStream((int16_t *)output_buf, *ctx->rtc_stream_config, *ctx->rtc_stream_config, (int16_t *)output_buf);
            // ctx->apm2->ProcessStream((int16_t *)capData, *ctx->rtc_stream_config, *ctx->rtc_stream_config, (int16_t *)capData);

        }

        // rnnoise
        if (ctx->rnnoise_enable) {
            int16_t *ptr = (int16_t *)output_buf;
            int nb_samples = outputFrame->nb_samples;
            for(int i = 0; i < nb_samples; ++i) {
                for (int ch = 0; ch < MIX_AUDIO_CHANNELS; ++ch) {
                    in[ch][i] = ptr[i * MIX_AUDIO_CHANNELS + ch];
                }
            }
            const float *input[] = {in[0]};
            float *output[] = {out[0]};
            
            for (int i = 0; i < nb_samples; ++i)
                for (int ch = 0; ch < MIX_AUDIO_CHANNELS; ++ch) 
                    output[ch][i] = input[ch][i];

            ctx->rnnoise->process(input, output, nb_samples, 
                ctx->vadThreshold, ctx->vadGracePeriodBlocks, ctx->retroactiveVADGraceBlocks);
            // for (int ch = 0; ch < MIX_AUDIO_CHANNELS; ++ch)
            //     rnnoise_process_frame(rnn, output[ch], input[ch]);
            

            // PrintI("rnnoise process: vadThreshold=%lf, vadGracePeriodBlocks=%d, retroactiveVADGraceBlocks=%d\n",
            //     ctx->vadThreshold, ctx->vadGracePeriodBlocks, ctx->retroactiveVADGraceBlocks);
            for (int i = 0; i < nb_samples; ++i)
                for (int ch = 0; ch < MIX_AUDIO_CHANNELS; ++ch)
                    ptr[i * MIX_AUDIO_CHANNELS + ch] = output[ch][i];
        }
        // fwrite(output_buf, 1, output_size, output_fp);
        // 音频缓存到队列
        {
            // uint8_t *buffer = (uint8_t *)malloc(buffer_size);
            // memcpy(buffer, capData, buffer_size);
            uint8_t *buffer = (uint8_t *)malloc(output_size);
            memcpy(buffer, output_buf, output_size);
            std::unique_lock<std::mutex> lck(*ctx->mutex);
            audio_buf_t out;
            out.data = buffer;
            out.index = 0;
            out.size = output_size;
            ctx->list->emplace_back(out);
        }
    }
    InfoL << "capture thread end";
    usbCaptureDev.destory();
    if (capData) free(capData);
    ctx->alsa = nullptr;
}

void push_thread(CallContext *ctx)
{
    RtmpConfig rtmp;
    if (pushInit(&ctx->rtmp) < 0) {
        return ;
    }
    memcpy(&rtmp, &ctx->rtmp, sizeof(rtmp));
    AVRational av;
    int64_t pts = 0;
    AVPacket *pkt = av_packet_alloc();
    int ret;
    av.den = rtmp.codecCtx->sample_rate;
    av.num = 1;

    AVFrame *inputFrame = av_frame_alloc();
    {
        inputFrame->sample_rate      = MIX_OUTPUT_RATE;
        inputFrame->format           = AV_SAMPLE_FMT_S16;
        inputFrame->channels         = MIX_AUDIO_CHANNELS;
        inputFrame->nb_samples       = 1024 * MIX_OUTPUT_RATE / 44100;
        inputFrame->channel_layout   = av_get_default_channel_layout(MIX_AUDIO_CHANNELS);

        int size = av_samples_get_buffer_size(nullptr, inputFrame->channels, inputFrame->nb_samples, (AVSampleFormat)inputFrame->format, 1);
        uint8_t *buffer = (uint8_t *)av_malloc(size);
        avcodec_fill_audio_frame(inputFrame, inputFrame->channels, (AVSampleFormat)inputFrame->format,
                                (const uint8_t*)buffer, size, 1);
    }
    AVFrame *outputFrame = av_frame_alloc();
    {
        outputFrame->format         = rtmp.codecCtx->sample_fmt;
        outputFrame->channel_layout = rtmp.codecCtx->channel_layout;
        outputFrame->sample_rate    = rtmp.codecCtx->sample_rate;
        outputFrame->nb_samples     = rtmp.codecCtx->frame_size;
        outputFrame->channels       = rtmp.codecCtx->channels;

        int output_bz = av_samples_get_buffer_size(NULL, outputFrame->channels, outputFrame->nb_samples, (AVSampleFormat)outputFrame->format, 0);
        uint8_t *samples_data = (uint8_t *)av_malloc(output_bz);
        avcodec_fill_audio_frame(outputFrame, outputFrame->channels, (AVSampleFormat)outputFrame->format, samples_data, output_bz, 0);
    }
    // 写入帧头
    ret = avformat_write_header(rtmp.formatCtx, nullptr);
    if (ret < 0) {
        PrintE("avformat_write_header failed.\n");
        return ;
    }
    int frames = 0;
    while (ctx->running) {
        if (frames <= 0) frames = inputFrame->nb_samples;
        while (frames > 0 && ctx->list->size() > 0) 
        {
            std::unique_lock<std::mutex> lck(*ctx->mutex);
            auto nsData = ctx->list->begin();


            int needSize = frames * sizeof(int16_t) * inputFrame->channels;
            int readSize = (nsData->size - nsData->index) >= needSize ? needSize : (nsData->size - nsData->index);

            memcpy(inputFrame->data[0] + (inputFrame->nb_samples - frames)*sizeof(int16_t)*inputFrame->channels, nsData->data + nsData->index, readSize);

            frames -= readSize/(sizeof(int16_t) * inputFrame->channels);
            nsData->index += readSize;

            if (nsData->index >= nsData->size) {
                free(nsData->data);
                ctx->list->erase(ctx->list->begin());
            }
        }
        if (frames > 0) continue;
        // 重采样
        {
            const uint8_t** in = (const uint8_t**)inputFrame->data;
            uint8_t **out = outputFrame->data;

            int len2, out_data_size;

            len2 = swr_convert(rtmp.swrCtx, out, outputFrame->nb_samples, in, inputFrame->nb_samples);
            if (len2 < 0) {
                printf("swr_convert failed. \n");
                break;
            }

            // out_data_size = len2 * rtmp.codecCtx->channels * av_get_bytes_per_sample(rtmp.codecCtx->sample_fmt);
            // if (ns_fp) fwrite(outputFrame->data[0], 1, out_data_size, ns_fp);
        }


        // 推流到远端
        if (pts > INT64_MAX) pts = 0;
        outputFrame->pts = pts;
        pts += av_rescale_q(outputFrame->nb_samples, av, rtmp.codecCtx->time_base);

        ret = avcodec_send_frame(rtmp.codecCtx, outputFrame);
        if (ret < 0) {
            PrintE("avcodec_send_frame failed: %d\n", ret);
            break;
        }
        while (ret >= 0) {
            ret = avcodec_receive_packet(rtmp.codecCtx, pkt);
            if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) {
                break;
            } else if (ret < 0) {
                fprintf(stderr, "Error during encoding\n");
                break;
            }

            // 将数据包时间戳从编码器时间基转换到流时间基
            pkt->stream_index = rtmp.stream->index;
            av_packet_rescale_ts(pkt, rtmp.codecCtx->time_base, rtmp.stream->time_base);
            pkt->duration = av_rescale_q(pkt->duration, rtmp.codecCtx->time_base, rtmp.stream->time_base);

            // 写入数据包到输出媒体文件
            ret = av_interleaved_write_frame(rtmp.formatCtx, pkt);
            if (ret < 0) {
                fprintf(stderr, "Error while writing audio frame\n");
                break;
            }

            // 释放数据包
            av_packet_unref(pkt);
        }
    }
    InfoL << "push thread end";
    if (ctx->running)
        ctx->running = false;
    // 写入帧尾
    av_write_trailer(rtmp.formatCtx);
    // 释放线程资源
    av_packet_free(&pkt);
    av_frame_free(&inputFrame);
    av_frame_free(&outputFrame);
    pushDestory(&ctx->rtmp);
    memset(&rtmp, 0, sizeof(rtmp));
}

int pushInit(RtmpConfig *config)
{
    if (nullptr == strstr(config->url, "rtmp://")) {
        PrintE("url error, url: %s\n", config->url);
        return -1;
    }
    AVCodec *codec = nullptr;
    AVCodecContext *codecCtx = nullptr;
    AVFormatContext *afctx = nullptr;
    AVCodecParameters *codecPar = nullptr;
    SwrContext *swrCtx = nullptr;
    AVStream *audio_st = nullptr;
    AVDictionary *opts = nullptr;
    int ret;

    // 打开输出流
    ret = avformat_alloc_output_context2(&afctx, nullptr, "flv", config->url);
    if (ret < 0) {
        PrintE("open output failed.\n");
        goto fail;
    }
    if ( !(afctx->oformat->flags & AVFMT_NOFILE) ) {
        ret = avio_open(&afctx->pb, config->url, AVIO_FLAG_WRITE);
        if (ret < 0) {
            PrintE("avio_open failed.\n");
            goto fail;
        }
    }

    // 创建音频流
    audio_st = avformat_new_stream(afctx, codec);
    if (!audio_st) {
        PrintE("alloc new audio stream failed.\n");
        goto fail;
    }
    // 设置编码参数
    codecPar = afctx->streams[audio_st->index]->codecpar;
    codecPar->codec_id          = AV_CODEC_ID_AAC;
    codecPar->codec_type        = AVMEDIA_TYPE_AUDIO;
    codecPar->codec_tag         = 0;
    codecPar->bit_rate          = 128 * 1024;
    codecPar->sample_rate       = 44100;
    codecPar->channel_layout    = av_get_default_channel_layout(MIX_AUDIO_CHANNELS);
    codecPar->channels          = av_get_channel_layout_nb_channels(codecPar->channel_layout);
    codecPar->format            = AV_SAMPLE_FMT_FLTP;

    // 编码器初始化
    codec = avcodec_find_encoder(codecPar->codec_id);
    if (!codec) {
        PrintE("find codec aac failed.\n");
        return -1;
    }

    codecCtx = avcodec_alloc_context3(codec);
    if (!codecCtx) {
        PrintE("alloc codec context failed.\n");
        goto fail;
    }

    ret = avcodec_parameters_to_context(codecCtx, codecPar);
    if (ret < 0) {
        PrintE("copt codec params failed.\n");
        goto fail;
    }
    // 禁用缓冲
    av_dict_set(&opts, "fflags", "nobuffer", AV_DICT_MATCH_CASE);
//    av_dict_set(&opts, "rtmp_live", "1", AV_DICT_MATCH_CASE);
    // 打开编码器
    ret = avcodec_open2(codecCtx, codec, &opts);
    if (ret < 0) {
        PrintE("open codec {} failed.\n", codec->id);
        goto fail;
    }
    audio_st->codecpar->codec_tag = 0;
    // 释放字典资源
    av_dict_free(&opts);

    // 打印输出流信息
    av_dump_format(afctx, 0, config->url, 1);

    // 重采样初始化
    swrCtx = swr_alloc_set_opts(nullptr,
                                // output
                                codecCtx->channel_layout,
                                codecCtx->sample_fmt,
                                codecCtx->sample_rate,
                                // input
                                av_get_default_channel_layout(MIX_AUDIO_CHANNELS),
                                AV_SAMPLE_FMT_S16,
                                MIX_OUTPUT_RATE,
                                0, nullptr);
    if (!swrCtx) {
        PrintE("swr_alloc_set_opts failed.\n");
        goto fail;
    }
    swr_init(swrCtx);

    config->codecCtx = codecCtx;
    config->formatCtx = afctx;
    config->stream = audio_st;
    config->swrCtx = swrCtx;
    PrintI("rtmp push init ok.\n");
    return 0;
fail:
    if (afctx) {
        if (afctx->pb)
            avio_close(afctx->pb);
        avformat_free_context(afctx);
    }
    if (codecCtx) {
        avcodec_close(codecCtx);
        avcodec_free_context(&codecCtx);
    }
    if (swrCtx) {
        swr_close(swrCtx);
        swr_free(&swrCtx);
    }
    return -1;
}
void pushDestory(RtmpConfig *config) {
    if (config->formatCtx) {
        if (config->formatCtx->pb)
            avio_close(config->formatCtx->pb);
        avformat_free_context(config->formatCtx);
    }
    if (config->codecCtx) {
        avcodec_close(config->codecCtx);
        avcodec_free_context(&config->codecCtx);
    }
    if (config->swrCtx) {
        swr_close(config->swrCtx);
        swr_free(&config->swrCtx);
    }
}