Prevent delayed camera startup from shifting video into the future

tools/chatgui/src/call_video_capture_clock.h

@@ -0,0 +1,19 @@
+#pragma once
+#include <cstdint>
+
+// Первый callback может опоздать при запуске камеры. Его задержка не должна навсегда сдвигать PTS в будущее.
+struct CallVideoCaptureClock {
+    int64_t source = -1, offset = 0, correction = 0;
+    int64_t map(int64_t next, int64_t arrival) {
+        if (next < 0) { source = -1; return arrival; }
+        if (source < 0 || next < source) offset = arrival - next;
+        source = next;
+        int64_t pts = next + offset;
+        if (pts > arrival) {
+            correction += pts - arrival;
+            offset -= pts - arrival;
+            pts = arrival;
+        }
+        return pts;
+    }
+};

tools/chatgui/src/call_video_engine.cpp

@@ -2,6 +2,7 @@
 #include "call_video_encoder.h"
 #include "call_video_decoder.h"
 #include "call_video_cadence.h"
+#include "call_video_capture_clock.h"
 #include "call_camera_rate.h"
 #include "../../../src/call/call_video_port.h"
 #include "../../../src/call/call_audio.h"
@@ -32,7 +33,8 @@ struct CallVideoEngine::Impl {
     uint64_t id = 0, revision = 1;
     bool enabled = false;
     int width = 640, height = 480, codec = 0;
-    int64_t pts = -1, backend = -1, anchor = -1;
+    int64_t pts = -1;
+    CallVideoCaptureClock captureClock;
     uint64_t captured = 0;
     QImage shown;
     uint64_t shownSerial = 0, deliveredSerial = 0;
@@ -109,7 +111,7 @@ void CallVideoEngine::closeCamera() {
     if (m->camera) { disconnect(m->camera.get(), nullptr, this, nullptr); m->camera->stop(); }
     m->cameraRate.restore();
     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; }
+    { std::lock_guard<std::mutex> guard(m->lock); m->raw = {}; m->pts = -1; m->captureClock = {}; }
 #endif
     m->previewPts = -1; m->previewVisible = false; m->previewReportAt = nowUs(); m->previewFrames = 0; m->previewMaxUs = 0;
     emit localImageChanged({});
@@ -152,15 +154,12 @@ void CallVideoEngine::setVideo(bool enabled, const QByteArray& cameraId, int wid
         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;
-        }
+        const int64_t arrival = nowUs(), source = frame.startTime();
+        if (m->pts < 0) DEBUG_INFO(DEBUG_CATEGORY_CALL, "camera clock source=%lld monotonic=%lld calibration=arrival-upper-bound",
+            (long long)source, (long long)arrival);
+        if (m->captureClock.source >= 0 && source < m->captureClock.source)
+            DEBUG_WARN(DEBUG_CATEGORY_CALL, "camera clock reset source=%lld previous=%lld", (long long)source, (long long)m->captureClock.source);
+        const int64_t pts = m->captureClock.map(source, arrival);
         if (pts > m->pts) { m->raw = frame; m->pts = pts; ++m->captured; }
     }, Qt::DirectConnection);
     connect(m->camera.get(), &QCamera::errorOccurred, this, [this](QCamera::Error, const QString& error) {
@@ -201,13 +200,14 @@ void CallVideoEngine::transmit() {
         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;
+        uint64_t revision, captured; bool enabled; int width, height, codec; int64_t pts, clockCorrection;
 #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;
           captured = m->captured;
+          clockCorrection = m->captureClock.correction;
 #ifdef HAVE_STREAM_CAPTURE
           raw = m->raw;
 #endif
@@ -216,10 +216,10 @@ void CallVideoEngine::transmit() {
         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",
+                "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)(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; }