utun2/tools/chatgui/tests/test_voice_audio_io.cpp

/* Настоящие workers/AEC; граница сервисного PCM заменена наблюдателями без сети/устройства. */
#include "../src/voice_audio_io.h"
extern "C" {
#include "call/call_audio.h"
#include "radio/radio_audio.h"
#include "debug_config.h"
}
#include <vector>
#include <array>
#include <chrono>
#include <thread>
#include <mutex>
#include <cmath>
#include <cstdio>
#include <cstdlib>
#include <algorithm>
#define REQUIRE(x) do { if (!(x)) { fprintf(stderr, "voice-io:%d %s\n", __LINE__, #x); abort(); } } while (0)
static std::mutex mutex;
static std::vector<int16_t> capture, transmitted;
static std::vector<uint8_t> masks;
static bool talking;
static std::atomic<int> pulls{0}, blocks{0};
static std::atomic<bool> stall{false}, stalled{false};
extern "C" int call_audio_pull_pcm(uint64_t, int16_t* out, int count) { ++pulls; std::fill_n(out, count, 1000); return count; }
extern "C" int radio_audio_pull_pcm(uint64_t, int16_t* out, int count) {
    ++pulls; for (int i = 0; i < count; ++i) out[i] = i % 2 ? 2000 : 1000; return count;
}
extern "C" void call_audio_reset_io(uint64_t, int) {}
extern "C" void radio_audio_reset_playback(uint64_t) {}
extern "C" void radio_audio_capture_discontinuity(uint64_t) {}
extern "C" void radio_audio_talk_begin(uint64_t) { std::lock_guard lock(mutex); talking = true; }
extern "C" void radio_audio_talk_end(uint64_t) { std::lock_guard lock(mutex); talking = false; }
extern "C" int radio_audio_feed_pcm(uint64_t, const int16_t* pcm, int count) {
    if (stall) { stalled = true; while (stall) std::this_thread::sleep_for(std::chrono::milliseconds(1)); }
    std::lock_guard lock(mutex);
    capture.insert(capture.end(), pcm, pcm + count);
    if (talking) transmitted.insert(transmitted.end(), pcm, pcm + count);
    ++blocks; return 0;
}
extern "C" int call_audio_feed_prepared_pcm(uint64_t, const int16_t* pcm, const uint8_t* mask) {
    std::lock_guard lock(mutex);
    capture.insert(capture.end(), pcm, pcm + 960); masks.insert(masks.end(), mask, mask + 960);
    ++blocks; return 0;
}
static int64_t now_us() {
    return std::chrono::duration_cast<std::chrono::microseconds>(std::chrono::steady_clock::now().time_since_epoch()).count();
}
template<class F> static void await(F ready) {
    for (int i = 0; i < 2000 && !ready(); ++i) std::this_thread::sleep_for(std::chrono::milliseconds(1));
    REQUIRE(ready());
}
static void reset() { std::lock_guard lock(mutex); capture.clear(); transmitted.clear(); masks.clear(); talking = false; blocks = 0; }
int main() {
    debug_config_init(); debug_set_level(DEBUG_LEVEL_WARN);
    VoiceAudioIo voice;
    std::array<int16_t, 960> pcm{}; pcm.fill(1234);
    voice.start(VoiceAudioIo::Kind::Radio, 7); REQUIRE(voice.startCapture(1, false));
    voice.capture(pcm.data(), 240, 0); voice.setPtt(true);
    voice.capture(pcm.data(), 360, 0); voice.setPtt(false); voice.capture(pcm.data(), 360, 0);
    await([] { return blocks.load() >= 3; }); voice.stopCapture();
    REQUIRE(capture.size() == 960 && transmitted.size() == 360 && !talking);
    reset(); REQUIRE(voice.startCapture(1, false));
    voice.setPtt(true); voice.capture(pcm.data(), 120, 0); voice.setPtt(false); voice.stopCapture();
    REQUIRE(capture.size() == 120 && transmitted.size() == 120 && !talking);
    reset(); REQUIRE(voice.startCapture(1, false)); stall = true;
    voice.capture(pcm.data(), 960, 0); await([] { return stalled.load(); });
    int before = pulls;
    for (int i = 0; i < 10; ++i) {
        float output[1920]{}; voice.mix(output, 960);
        std::this_thread::sleep_for(std::chrono::milliseconds(3));
    }
    REQUIRE(pulls > before + 3); stall = false; voice.stopCapture(); voice.stop();
    reset(); voice.start(VoiceAudioIo::Kind::Call, 42); REQUIRE(voice.startCapture(1, false));
    voice.capture(pcm.data(), 480, 0); voice.setMuted(true); voice.capture(pcm.data(), 480, 0);
    await([] { return blocks.load() == 1; }); voice.stopCapture();
    REQUIRE(masks.size() == 960);
    for (int i = 0; i < 960; ++i) REQUIRE(masks[i] == (i >= 480) && capture[i] == 1234);
    voice.stop(); reset();
    voice.start(VoiceAudioIo::Kind::Call, 42); REQUIRE(voice.startCapture(1, false));
    voice.capture(pcm.data(), 960, now_us() - 300000); // Backend latency не является возрастом worker-очереди.
    await([] { return blocks.load() == 1; }); voice.stopCapture(); voice.stop();
    REQUIRE(capture.size() == 960); reset();
    // Референс противофазный: mono downmix дал бы ноль и не смог подавить эхо.
    voice.start(VoiceAudioIo::Kind::Call, 43); REQUIRE(voice.startCapture(1, true));
    uint32_t rng = 1234567;
    double echoPower = 0, outputPower = 0;
    int64_t time = now_us();
    for (int frame = 0; frame < 240; ++frame) {
        float reference[1920];
        for (int i = 0; i < 960; ++i) {
            rng = rng * 1664525u + 1013904223u;
            int16_t sample = (int16_t)((rng >> 18) - 8192);
            reference[2*i] = sample / 32768.0f; reference[2*i+1] = -reference[2*i];
            pcm[i] = sample / 2;
            if (frame >= 180) echoPower += (double)pcm[i] * pcm[i];
        }
        voice.reference(reference, 960, time + frame * 20000);
        voice.capture(pcm.data(), 960, time + frame * 20000);
        await([&] { return blocks.load() == frame + 1; });
    }
    voice.stopCapture(); voice.stop();
    for (size_t i = 180 * 960; i < capture.size(); ++i) outputPower += (double)capture[i] * capture[i];
    printf("stereo reference residual ratio=%.6f\n", outputPower / echoPower);
    REQUIRE(outputPower < echoPower * 0.1);
    puts("voice workers: exact PTT/mute boundaries, short tail, independent RX, stereo AEC PASS");
}