Browse Source

Connect desktop camera workers to negotiated call video and audio clock

master
evgeny 5 days ago
parent
commit
cabae34791
  1. 242
      tools/chatgui/src/call_video_engine.cpp
  2. 27
      tools/chatgui/src/call_video_engine.h

242
tools/chatgui/src/call_video_engine.cpp

@ -0,0 +1,242 @@
#include "call_video_engine.h"
#include "call_video_encoder.h"
#include "call_video_decoder.h"
#include "../../../src/call/call_video_port.h"
#include "../../../src/call/call_audio.h"
#include "debug_config.h"
#include "mem.h"
#include <thread>
#include <mutex>
#include <atomic>
#include <deque>
#include <chrono>
#ifdef HAVE_STREAM_CAPTURE
#include <QCamera>
#include <QCameraDevice>
#include <QMediaDevices>
#include <QMediaCaptureSession>
#include <QVideoSink>
#include <QVideoFrame>
#endif
namespace {
int64_t nowUs() { return std::chrono::duration_cast<std::chrono::microseconds>(std::chrono::steady_clock::now().time_since_epoch()).count(); }
void idle() { std::this_thread::sleep_for(std::chrono::milliseconds(5)); }
}
struct CallVideoEngine::Impl {
std::mutex lock;
std::thread tx, rx;
std::atomic<bool> running{false};
uint64_t id = 0, revision = 1;
bool enabled = false;
int width = 640, height = 480, codec = 0;
int64_t pts = -1, backend = -1, anchor = -1;
QImage shown;
uint64_t shownSerial = 0, deliveredSerial = 0;
QString status;
bool statusDirty = false;
void report(const QString& text) {
std::lock_guard<std::mutex> guard(lock); status = text; statusDirty = true;
DEBUG_INFO(DEBUG_CATEGORY_CALL, "desktop video: %s", qPrintable(text));
}
#ifdef HAVE_STREAM_CAPTURE
std::unique_ptr<QCamera> camera;
std::unique_ptr<QMediaCaptureSession> capture;
std::unique_ptr<QVideoSink> sink;
QVideoFrame raw;
#endif
};
CallVideoEngine::CallVideoEngine(QObject* parent) : QObject(parent), m(std::make_unique<Impl>()) {
m_timer.setInterval(20);
connect(&m_timer, &QTimer::timeout, this, [this] {
QImage image; QString status; bool changed = false, report = false;
{ std::lock_guard<std::mutex> guard(m->lock);
if (m->shownSerial != m->deliveredSerial) { image = m->shown; m->deliveredSerial = m->shownSerial; changed = true; }
if (m->statusDirty) { status = m->status; m->statusDirty = false; report = true; } }
if (changed) emit imageChanged(image);
if (report) emit statusChanged(status);
});
}
CallVideoEngine::~CallVideoEngine() { stop(); }
QList<QPair<QString, QByteArray>> CallVideoEngine::cameras() {
QList<QPair<QString, QByteArray>> result;
#ifdef HAVE_STREAM_CAPTURE
for (const auto& camera : QMediaDevices::videoInputs()) result.append({camera.description(), camera.id()});
#endif
return result;
}
void CallVideoEngine::start(quint64 id) {
if (m->running && m->id == id) return;
stop(); m->id = id; m->running = true; m_timer.start();
try { m->tx = std::thread(&CallVideoEngine::transmit, this); m->rx = std::thread(&CallVideoEngine::receive, this); }
catch (const std::exception& error) { DEBUG_ERROR(DEBUG_CATEGORY_CALL, "video worker start: %s", error.what()); stop(); emit statusChanged(error.what()); }
}
void CallVideoEngine::closeCamera() {
#ifdef HAVE_STREAM_CAPTURE
if (m->sink) disconnect(m->sink.get(), nullptr, this, nullptr);
if (m->camera) { disconnect(m->camera.get(), nullptr, this, nullptr); m->camera->stop(); }
m->capture.reset(); m->camera.reset(); m->sink.reset();
std::lock_guard<std::mutex> guard(m->lock); m->raw = {}; m->pts = m->backend = m->anchor = -1;
#endif
}
void CallVideoEngine::stop() {
m->running = false; m_timer.stop(); closeCamera();
if (m->tx.joinable()) m->tx.join();
if (m->rx.joinable()) m->rx.join();
if (m->id) call_video_port_close(m->id);
m->id = 0; m->enabled = false; m->shown = {}; emit imageChanged({});
}
void CallVideoEngine::setVideo(bool enabled, const QByteArray& cameraId, int width, int height, int codec) {
closeCamera();
uint64_t revision;
{ std::lock_guard<std::mutex> guard(m->lock);
m->enabled = enabled; m->width = width; m->height = height; m->codec = codec; revision = ++m->revision; }
if (!enabled || !m->running) return;
#ifdef HAVE_STREAM_CAPTURE
QCameraDevice selected;
for (const auto& camera : QMediaDevices::videoInputs()) if (camera.id() == cameraId) selected = camera;
if (selected.isNull()) { m->report(tr("Camera is unavailable")); return; }
m->capture = std::make_unique<QMediaCaptureSession>(); m->sink = std::make_unique<QVideoSink>();
m->camera = std::make_unique<QCamera>(selected); m->capture->setCamera(m->camera.get()); m->capture->setVideoSink(m->sink.get());
QCameraFormat best; int distance = INT_MAX;
for (const auto& f : selected.videoFormats()) {
if (f.maxFrameRate() < 29.9) continue;
int score = std::abs(f.resolution().width() - width) + std::abs(f.resolution().height() - height);
if (score < distance) { best = f; distance = score; }
}
if (!best.isNull()) m->camera->setCameraFormat(best);
connect(m->sink.get(), &QVideoSink::videoFrameChanged, this, [this, revision](const QVideoFrame& frame) {
if (!frame.isValid() || !m->running) return;
std::lock_guard<std::mutex> guard(m->lock);
if (revision != m->revision) return;
int64_t pts = nowUs();
if (frame.startTime() >= 0) {
if (m->backend < 0) {
m->backend = frame.startTime(); m->anchor = pts;
DEBUG_INFO(DEBUG_CATEGORY_CALL, "camera clock source=%lld monotonic=%lld calibration=arrival-estimate",
(long long)m->backend, (long long)m->anchor);
}
pts = m->anchor + frame.startTime() - m->backend;
}
if (pts > m->pts) { m->raw = frame; m->pts = pts; }
}, Qt::DirectConnection);
connect(m->camera.get(), &QCamera::errorOccurred, this, [this](QCamera::Error, const QString& error) {
DEBUG_ERROR(DEBUG_CATEGORY_CALL, "call camera failed: %s", qPrintable(error)); m->report(error);
});
m->camera->start();
#else
Q_UNUSED(cameraId); Q_UNUSED(revision); m->report(tr("Video capture is unavailable in this build"));
#endif
}
void CallVideoEngine::transmit() {
call_video_caps caps{};
// Каждый объявленный codec проверяется реальным encoder + независимым decoder при open.
for (unsigned codec = 1; codec <= 2 && m->running; ++codec) {
call_video_format probe{1, 1280, 720, 30, 0, (uint8_t)codec, (uint8_t)(codec == 1 ? 31 : 93), 4000000, 1000000};
CallVideoEncoder encoder;
if (encoder.open(probe, 1000000)) {
call_video_limit limit{1280, 720, 30, 1280 * 720 * 30, 4000000, 1000000, probe.level};
caps.encode[codec - 1] = caps.decode[codec - 1] = limit;
} else DEBUG_WARN(DEBUG_CATEGORY_CALL, "call codec probe unavailable codec=%u: %s", codec, qPrintable(encoder.error()));
}
if (!m->running) return;
if (call_video_port_open(m->id, &caps) < 0) { m->report(tr("Video transport is unavailable")); return; }
CallVideoEncoder encoder;
call_video_format format{};
uint64_t appliedRevision = 0, keySerial = 0, request = 0, frameId = 0;
int64_t lastPts = -1, lastKey = -1;
bool forceKey = true;
while (m->running) {
call_video_port_status state{};
if (call_video_port_status(m->id, &state) < 0) break;
uint64_t revision; bool enabled; int width, height, codec; int64_t pts;
#ifdef HAVE_STREAM_CAPTURE
QVideoFrame raw;
#endif
{ std::lock_guard<std::mutex> guard(m->lock);
revision = m->revision; enabled = m->enabled; width = m->width; height = m->height; codec = m->codec; pts = m->pts;
#ifdef HAVE_STREAM_CAPTURE
raw = m->raw;
#endif
}
if (!enabled || !state.have_peer || pts <= lastPts || (lastPts >= 0 && pts - lastPts < 33333)) { idle(); continue; }
if (revision != appliedRevision) {
if (lastKey >= 0 && pts - lastKey < 1000000) { idle(); continue; }
call_video_format next{format.generation + 1, (uint16_t)width, (uint16_t)height, 30, 0, (uint8_t)codec, 0, 4000000, 1000000};
if (call_video_choose(&caps, &state.peer, &next) <= 0 || !encoder.open(next, 1000000)) {
m->report(tr("Selected video mode is unavailable to the peer"));
std::lock_guard<std::mutex> guard(m->lock); if (m->revision == revision) m->enabled = false;
continue;
}
format = next; appliedRevision = revision; forceKey = true; request = 0;
if (call_video_port_format(m->id, &format, encoder.bitrate()) < 0) break;
m->report(QString("%1 · %2×%3 · %4 kbit/s").arg(format.codec == 2 ? "H.265" : "H.264").arg(width).arg(height).arg(encoder.bitrate() / 1000));
}
if (state.configured_generation != format.generation) { idle(); continue; }
if (state.key_serial != keySerial) { keySerial = state.key_serial; forceKey = true; }
if (state.rate_generation == format.generation && state.rate_request > request) {
request = state.rate_request;
if (!encoder.requestBitrate(format.generation, request, state.rate_bps))
call_video_port_applied(m->id, format.generation, request, 0);
}
#ifdef HAVE_STREAM_CAPTURE
QImage image = raw.toImage();
if (image.isNull()) { DEBUG_WARN(DEBUG_CATEGORY_CALL, "camera frame conversion failed"); lastPts = pts; continue; }
LiveEncodedFrame output; CallVideoRateCompletion completion;
bool encoded = encoder.encode(image, pts, forceKey, output, completion); lastPts = pts;
if (completion.request) {
call_video_port_applied(m->id, completion.generation, completion.request, completion.bitrate);
m->report(QString("%1 · %2×%3 · %4 kbit/s").arg(format.codec == 2 ? "H.265" : "H.264")
.arg(format.width).arg(format.height).arg(encoder.bitrate() / 1000));
}
if (!encoded) { forceKey = true; m->report(encoder.error()); continue; }
if (output.key) lastKey = pts;
call_video_frame packet{};
packet.call_id = m->id; packet.frame_id = ++frameId; packet.pts_us = pts; packet.format = format;
packet.size = packet.length = output.data.size(); packet.data = reinterpret_cast<const uint8_t*>(output.data.constData());
packet.flags = output.key ? CALL_VIDEO_KEY : 0;
forceKey = call_video_port_send(&packet) != 1;
#else
idle();
#endif
}
}
void CallVideoEngine::receive() {
struct Picture { int64_t pts; QImage image; };
std::deque<Picture> pictures;
CallVideoDecoder decoder;
uint64_t sourceEpoch = 0; int64_t lastShown = -1, report = 0; unsigned shown = 0;
while (m->running) {
int64_t now = nowUs(); call_clock_snapshot clock{};
call_audio_get_clock(m->id, now, &clock);
if (clock.valid) {
if (sourceEpoch && sourceEpoch != clock.source_epoch) {
pictures.clear(); decoder.close(); call_video_port_request_key(m->id); lastShown = -1;
DEBUG_INFO(DEBUG_CATEGORY_CALL, "video playback audio source epoch changed");
}
sourceEpoch = clock.source_epoch;
QImage selected;
while (!pictures.empty() && pictures.front().pts <= clock.pts_us) {
if (pictures.front().pts > lastShown) { selected = pictures.front().image; lastShown = pictures.front().pts; }
pictures.pop_front();
}
if (!selected.isNull()) { std::lock_guard<std::mutex> guard(m->lock); m->shown = selected; ++m->shownSerial; ++shown; }
}
if (pictures.size() < 16) {
call_video_frame packet{};
if (call_video_port_pull(m->id, &packet) == 1) {
QImage image;
if (decoder.decode(packet, image)) pictures.push_back({packet.pts_us, image});
else call_video_port_request_key(m->id);
u_free((void*)packet.data); continue;
}
}
if (now - report >= 1000000) {
DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "video playback shown=%u queue=%zu audio_valid=%d audio_pts=%lld video_pts=%lld",
shown, pictures.size(), clock.valid, (long long)clock.pts_us, (long long)lastShown); report = now;
}
idle();
}
}

27
tools/chatgui/src/call_video_engine.h

@ -0,0 +1,27 @@
#pragma once
#include <QObject>
#include <QImage>
#include <QTimer>
#include <memory>
// GUI владеет камерой, TX/RX workers — codec; отображение читает фактический audio clock.
class CallVideoEngine : public QObject {
Q_OBJECT
public:
explicit CallVideoEngine(QObject* parent = nullptr);
~CallVideoEngine() override;
void start(quint64 id);
void stop();
void setVideo(bool enabled, const QByteArray& camera, int width, int height, int codec);
static QList<QPair<QString, QByteArray>> cameras();
signals:
void imageChanged(const QImage& image);
void statusChanged(const QString& status);
private:
struct Impl;
std::unique_ptr<Impl> m;
QTimer m_timer;
void closeCamera();
void transmit();
void receive();
};
Loading…
Cancel
Save