#include "call_video_engine.h" #include "call_video_encoder.h" #include "call_video_decoder.h" #include "call_video_cadence.h" #include "call_camera_capture.h" #include "../../../src/call/call_video_port.h" #include "../../../src/call/call_video_playout.h" #include "../../../src/call/call_audio.h" #include "debug_config.h" #include "mem.h" #include #include #include #include #include #include namespace { int64_t nowUs() { return call_video_playout_now_us(); } } struct CallVideoEngine::Impl { std::mutex lock; std::condition_variable wake; uint64_t wakeSerial = 0; std::thread tx, rx; std::atomic running{false}; uint64_t id = 0, revision = 1; bool enabled = false; int width = 640, height = 480, codec = 0; int64_t pts = -1; int64_t clockCorrection = 0; QImage capturedImage; QByteArray cameraId; bool captureActive = false; uint64_t captured = 0; QImage shown; int64_t shownPts = -1; uint64_t publicationEpoch = 0; bool publicationPending = false; int64_t shownAt = 0; int64_t previewReportAt = 0, previewMaxUs = 0; unsigned previewFrames = 0; QString status; bool statusDirty = false; void report(const QString& text) { std::lock_guard guard(lock); status = text; statusDirty = true; DEBUG_INFO(DEBUG_CATEGORY_CALL, "desktop video: %s", qPrintable(text)); } #ifdef HAVE_STREAM_CAPTURE std::unique_ptr capture; #endif }; CallVideoEngine::CallVideoEngine(QObject* parent) : QObject(parent), m(std::make_unique()) { m_timer.setInterval(20); m_timer.setTimerType(Qt::PreciseTimer); connect(&m_timer, &QTimer::timeout, this, [this] { QString status; bool report = false; { std::lock_guard guard(m->lock); if (m->statusDirty) { status = m->status; m->statusDirty = false; report = true; } } if (report) emit statusChanged(status); }); #ifdef HAVE_STREAM_CAPTURE m->capture = std::make_unique(); connect(m->capture.get(), &CallCameraCapture::devicesChanged, this, &CallVideoEngine::camerasChanged); connect(m->capture.get(), &CallCameraCapture::failed, this, [this](const QString& error) { { std::lock_guard guard(m->lock); m->enabled = false; ++m->revision; ++m->wakeSerial; m->capturedImage = {}; m->pts = -1; } m->captureActive = false; m->wake.notify_one(); emit localImageChanged({}); m->report(error); emit statusChanged(error); }); connect(m->capture.get(), &CallCameraCapture::frameReady, this, [this](const QImage& image, qint64 pts, qint64 correction) { const int64_t now = nowUs(); { std::lock_guard guard(m->lock); if (!m->running || !m->enabled || pts <= m->pts) return; m->capturedImage = image; m->pts = pts; m->clockCorrection = correction; ++m->captured; ++m->wakeSerial; } m->wake.notify_one(); emit localImageChanged(image.scaled(320, 240, Qt::KeepAspectRatio, Qt::FastTransformation)); ++m->previewFrames; m->previewMaxUs = std::max(m->previewMaxUs, nowUs() - now); if (now - m->previewReportAt >= 1000000) { DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "video local preview call=%llu frames=%u update_max_us=%lld capture_age_us=%lld", (unsigned long long)m->id, m->previewFrames, (long long)m->previewMaxUs, (long long)(now - pts)); m->previewReportAt = now; m->previewFrames = 0; m->previewMaxUs = 0; } }); #endif } CallVideoEngine::~CallVideoEngine() { stop(); } // В GUI одновременно не больше одного уведомления; при паузе UI остаётся последний выбранный кадр. void CallVideoEngine::publishImage(const QImage& image, int64_t pts) { uint64_t epoch; { std::lock_guard guard(m->lock); m->shown = image; m->shownPts = pts; m->shownAt = nowUs(); if (m->publicationPending) return; m->publicationPending = true; epoch = m->publicationEpoch; } QMetaObject::invokeMethod(this, [this, epoch] { QImage image; int64_t pts, selected; { std::lock_guard guard(m->lock); if (epoch != m->publicationEpoch) return; image = m->shown; pts = m->shownPts; selected = m->shownAt; m->publicationPending = false; } emit imageChanged(image); if (!image.isNull() && pts >= 0 && pts % 1000000 < 34000) DEBUG_DEBUG(DEBUG_CATEGORY_DEBUG, "video timing stage=publish call=%llu pts=%lld mono_us=%lld ui_wait_us=%lld", (unsigned long long)m->id, (long long)pts, (long long)nowUs(), (long long)(nowUs() - selected)); }, Qt::QueuedConnection); } 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->captureActive) m->capture->configure(false, m->cameraId, m->width, m->height); #endif m->captureActive = false; { std::lock_guard guard(m->lock); m->capturedImage = {}; m->pts = -1; ++m->wakeSerial; } m->wake.notify_one(); emit localImageChanged({}); } 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); { std::lock_guard guard(m->lock); ++m->publicationEpoch; m->publicationPending = false; m->shown = {}; m->shownPts = -1; } m->id = 0; m->enabled = false; emit imageChanged({}); } void CallVideoEngine::setVideo(bool enabled, const QByteArray& cameraId, int width, int height, int codec) { const bool active = enabled && m->running; const bool captureChanged = active != m->captureActive || cameraId != m->cameraId || width != m->width || height != m->height; { std::lock_guard guard(m->lock); if (m->enabled == enabled && cameraId == m->cameraId && width == m->width && height == m->height && codec == m->codec && !captureChanged) return; m->enabled = enabled; m->width = width; m->height = height; m->codec = codec; ++m->revision; ++m->wakeSerial; if (captureChanged) { m->capturedImage = {}; m->pts = -1; } } m->cameraId = cameraId; m->captureActive = active; m->wake.notify_one(); if (!captureChanged) return; emit localImageChanged({}); #ifdef HAVE_STREAM_CAPTURE m->capture->configure(active, cameraId, width, height); #else if (active) m->report(tr("Video capture is unavailable in this build")); #endif } void CallVideoEngine::transmit() { call_video_caps caps = CallVideoDecoder::capabilities(); // Не связываем приём с наличием собственного encoder. 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, (uint8_t)(codec == 1 ? CALL_VIDEO_CONSTRAINED_BASELINE : CALL_VIDEO_MAIN)}; CallVideoEncoder encoder; if (encoder.open(probe, 1000000)) { call_video_limit limit{1280, 720, 30, 1280 * 720 * 30, 4000000, probe.profile, probe.level}; caps.encode[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; uint64_t userRevision = 0, rejectedGeneration = 0; unsigned excluded = 0; int64_t lastPts = -1; bool forceKey = true; int64_t reportAt = nowUs(), encodeMax = 0; uint64_t captureBase = 0; unsigned encodedCount = 0, keyCount = 0, sendBusy = 0, paced = 0; CallVideoCadence cadence; uint64_t wakeSerial = 0; auto idle = [this, &wakeSerial] { std::unique_lock guard(m->lock); m->wake.wait_for(guard, std::chrono::milliseconds(5), [&] { return !m->running || m->wakeSerial != wakeSerial; }); }; while (m->running) { struct call_video_port_status state{}; if (call_video_port_status(m->id, &state) < 0) break; if (state.error) { m->report(tr("Core rejected video format")); break; } uint64_t revision, captured; bool enabled; int width, height, codec; int64_t pts, clockCorrection; QImage image; { std::lock_guard guard(m->lock); revision = m->revision; enabled = m->enabled; width = m->width; height = m->height; codec = m->codec; pts = m->pts; captured = m->captured; clockCorrection = m->clockCorrection; image = m->capturedImage; wakeSerial = m->wakeSerial; } int64_t now = nowUs(); if (now - reportAt >= 1000000) { DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "video capture flow call=%llu interval_us=%lld captured=%llu encoded=%u keys=%u send_busy=%u paced=%u " "encode_max_us=%lld capture_age_us=%lld pts=%lld clock_correction_us=%lld", (unsigned long long)m->id, (long long)(now - reportAt), (unsigned long long)(captured - captureBase), encodedCount, keyCount, sendBusy, paced, (long long)encodeMax, (long long)(pts >= 0 ? now - pts : 0), (long long)pts, (long long)clockCorrection); reportAt = now; captureBase = captured; encodedCount = keyCount = sendBusy = paced = 0; encodeMax = 0; } if (revision != userRevision) { userRevision = revision; excluded = 0; } if (format.generation && state.decoded_generation == format.generation && !state.decoded_accepted && rejectedGeneration != format.generation) { rejectedGeneration = format.generation; excluded |= 1u << format.codec; if (!codec) { appliedRevision = 0; m->report(tr("Receiver rejected video; selecting fallback")); } else { m->report(tr("Receiver cannot decode selected codec")); std::lock_guard guard(m->lock); m->enabled = false; continue; } } if (format.generation && state.have_peer && !call_video_format_supported(&caps, &state.peer, &format)) appliedRevision = 0; if (!enabled || !state.have_peer || pts <= lastPts) { idle(); continue; } if (revision != appliedRevision) { call_video_format next{format.generation + 1, (uint16_t)width, (uint16_t)height, 30, 0, (uint8_t)codec, 0, 4000000, 0}; bool opened = false; uint32_t minimum = state.minimum_bps ? state.minimum_bps : 300000; uint32_t maximum = state.maximum_bps ? state.maximum_bps : 4000000; uint32_t initialRate = std::max(minimum, std::min(maximum, encoder.bitrate() ? encoder.bitrate() : 1000000)); for (int candidate = 2; candidate >= 1; --candidate) { if ((codec && codec != candidate) || (excluded & (1u << candidate))) continue; next.codec = candidate; next.level = next.profile = 0; if (call_video_choose(&caps, &state.peer, &next) > 0 && encoder.open(next, initialRate)) { opened = true; break; } } if (!opened) { m->report(tr("Selected video mode is unavailable to the peer")); std::lock_guard 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 if (!cadence.accept(pts)) { lastPts = pts; ++paced; idle(); continue; } if (image.isNull() || nowUs() - pts > CallCameraMaxFrameAgeUs) { DEBUG_WARN(DEBUG_CATEGORY_CALL, "video TX skips old capture age_us=%lld", (long long)(nowUs() - pts)); lastPts = pts; continue; } LiveEncodedFrame output; CallVideoRateCompletion completion; int64_t encodeStarted = nowUs(); bool encoded = encoder.encode(image, pts, forceKey, output, completion); lastPts = pts; encodeMax = std::max(encodeMax, nowUs() - encodeStarted); 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; } ++encodedCount; if (output.key) ++keyCount; 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(output.data.constData()); packet.flags = output.key ? CALL_VIDEO_KEY : 0; { std::lock_guard guard(m->lock); if (!m->running || !m->enabled || m->revision != revision || nowUs() - pts > CallCameraMaxFrameAgeUs) { DEBUG_WARN(DEBUG_CATEGORY_CALL, "video TX discards obsolete encoded frame revision=%llu/%llu enabled=%d age_us=%lld", (unsigned long long)revision, (unsigned long long)m->revision, m->enabled, (long long)(nowUs() - pts)); forceKey = true; continue; } forceKey = call_video_port_send(&packet) != 1; } // Одинаковое окно исходного PTS на всех этапах: примерно один кадр в секунду, без трассы каждого кадра. if (pts % 1000000 < 34000) DEBUG_DEBUG(DEBUG_CATEGORY_DEBUG, "video timing stage=encode call=%llu frame=%llu pts=%lld mono_us=%lld encode_us=%lld accepted=%d", (unsigned long long)m->id, (unsigned long long)frameId, (long long)pts, (long long)nowUs(), (long long)(nowUs() - encodeStarted), !forceKey); if (forceKey) ++sendBusy; #else idle(); #endif } } void CallVideoEngine::receive() { struct Picture { call_video_playout_frame timing; QImage image; }; std::deque pictures; CallVideoDecoder decoder; call_video_playout playout; call_video_playout_init(&playout, m->id); int64_t lastPictureAt = 0; uint64_t reportedGeneration = 0, rejectedGeneration = 0; while (m->running) { // Один кадр ждёт своего срока, второй позволяет пропустить устаревший показ после паузы CPU. while (pictures.size() < 2 && m->running) { call_video_frame packet{}; if (call_video_port_pull(m->id, &packet) == 1) { if (packet.format.generation == rejectedGeneration) { u_free((void*)packet.data); continue; } QImage image; const int64_t decodeStarted = nowUs(); if (decoder.decode(packet, image)) { pictures.push_back({{packet.pts_us, packet.received_us}, image}); if (packet.pts_us % 1000000 < 34000) DEBUG_DEBUG(DEBUG_CATEGORY_DEBUG, "video timing stage=decode call=%llu frame=%llu pts=%lld mono_us=%lld decode_us=%lld queue=%zu", (unsigned long long)m->id, (unsigned long long)packet.frame_id, (long long)packet.pts_us, (long long)nowUs(), (long long)(nowUs() - decodeStarted), pictures.size()); if (reportedGeneration != packet.format.generation) { reportedGeneration = packet.format.generation; call_video_port_decoded(m->id, reportedGeneration, 1); } } else { rejectedGeneration = packet.format.generation; call_video_port_decoded(m->id, rejectedGeneration, 0); } u_free((void*)packet.data); } else break; } int64_t now = nowUs(); call_clock_snapshot clock{}; call_audio_get_clock(m->id, now, &clock); auto action = call_video_playout_poll(&playout, pictures.empty() ? nullptr : &pictures.front().timing, pictures.size() > 1 ? &pictures[1].timing : nullptr, now, clock.valid, clock.pts_us); if (action != CALL_VIDEO_PLAYOUT_WAIT) { if (action == CALL_VIDEO_PLAYOUT_SHOW) { const auto& f = pictures.front(); publishImage(f.image, f.timing.pts_us); lastPictureAt = now; if (f.timing.pts_us % 1000000 < 34000) DEBUG_DEBUG(DEBUG_CATEGORY_DEBUG, "video timing stage=select call=%llu pts=%lld mono_us=%lld audio_pts=%lld", (unsigned long long)m->id, (long long)f.timing.pts_us, (long long)now, (long long)clock.pts_us); } pictures.pop_front(); continue; } // VIDEO_OFF пока отсутствует в протоколе: убираем застывшую картинку после прекращения показа. if (lastPictureAt && now - lastPictureAt >= 5000000) { publishImage({}, -1); lastPictureAt = 0; DEBUG_INFO(DEBUG_CATEGORY_CALL, "video playback hidden: no playable frame for 5s"); } std::this_thread::sleep_for(std::chrono::microseconds(std::max(100, std::min(5000, playout.deadline_us - nowUs())))); } }