Drive desktop call clock from rendered PCM and capture timing

src/call/call_audio.c

@@ -492,7 +492,9 @@ int call_audio_pull_pcm(uint64_t call_id, int16_t* out, int max_samples) {
         }
         if (call_tones_ended_pending(tones))
             n = call_tones_render(tones, out, max_samples);
-        call_jitter_silence(j, (unsigned)n);
+        memset(out + n, 0, (size_t)(max_samples - n) * sizeof(*out));
+        call_jitter_silence(j, (unsigned)max_samples);
+        n = max_samples;
     } else {
         n = call_jitter_pull(j, out, max_samples);
         int previous_gap = g_gap_samples;

tools/chatgui/src/voice_audio_io.cpp

@@ -122,7 +122,7 @@ void VoiceAudioIo::reference(const float* pcm, unsigned frames, int64_t firstTim
 }
 
 /* Выход всегда обслуживается независимо от capture/VAD. Уже потреблённый PCM становится референсом у владельца микшера. */
-void VoiceAudioIo::mix(float* stereo, unsigned frames, int64_t firstTimeUs) {
+void VoiceAudioIo::mix(float* stereo, unsigned frames, int64_t firstTimeUs, bool measured) {
     uint64_t epoch = m_playEpoch.load();
     if (!firstTimeUs) firstTimeUs = voice_time_us();
     unsigned consumed = 0;
@@ -140,9 +140,9 @@ void VoiceAudioIo::mix(float* stereo, unsigned frames, int64_t firstTimeUs) {
         ++m_playPosition;
         ++consumed;
     }
-    if (m_kind == Kind::Radio && consumed) {
+    if (consumed) {
         PlaybackProgress progress{epoch, m_playCurrent.position + m_playPosition,
-                                  firstTimeUs + (int64_t)consumed * 1000000 / 48000};
+                                  firstTimeUs + (int64_t)consumed * 1000000 / 48000, measured};
         if (!m_playProgress.push(progress)) m_progressDropped.fetch_add(1, std::memory_order_relaxed);
     }
     m_wake.notify_all();
@@ -151,11 +151,18 @@ void VoiceAudioIo::mix(float* stereo, unsigned frames, int64_t firstTimeUs) {
 void VoiceAudioIo::playbackLoop() {
     uint64_t epoch = m_playEpoch.load();
     uint64_t position = 0;
+    call_clock_snapshot clock{};
+    if (m_kind == Kind::Call && call_audio_get_clock(m_session, voice_time_us(), &clock) != 0)
+        DEBUG_WARN(DEBUG_CATEGORY_CALL, "voice-io: call clock unavailable session=%llu", (unsigned long long)m_session);
     if (m_kind == Kind::Radio) radio_audio_playback_progress(m_session, 0, 0);
     while (m_running) {
         uint64_t requested = m_playEpoch.load();
         if (requested != epoch) {
-            if (m_kind == Kind::Call) call_audio_reset_io(m_session, 1);
+            if (m_kind == Kind::Call) {
+                call_audio_reset_io(m_session, 1);
+                if (call_audio_get_clock(m_session, voice_time_us(), &clock) != 0)
+                    DEBUG_WARN(DEBUG_CATEGORY_CALL, "voice-io: reset call clock unavailable session=%llu", (unsigned long long)m_session);
+            }
             else radio_audio_reset_playback(m_session);
             position = 0;
             epoch = requested;
@@ -163,11 +170,13 @@ void VoiceAudioIo::playbackLoop() {
         PlaybackProgress progress{};
         for (unsigned count = 0; count < 32 && m_playProgress.pop(progress); ++count) {
             if (progress.epoch != epoch) continue;
-            int64_t remaining = progress.endTimeUs - voice_time_us();
+            int64_t now = voice_time_us(), remaining = progress.endTimeUs - now;
-            radio_audio_playback_progress(m_session, progress.position, remaining > 0 ? (remaining + 99) / 100 : 0);
+            if (m_kind == Kind::Call)
+                call_audio_playback_progress(m_session, clock.epoch, progress.position, progress.endTimeUs, now, progress.measured);
+            else radio_audio_playback_progress(m_session, progress.position, remaining > 0 ? (remaining + 99) / 100 : 0);
         }
         unsigned lost = m_progressDropped.exchange(0);
-        if (lost) DEBUG_WARN(DEBUG_CATEGORY_RADIO, "voice-io: playback progress overflow session=%llu dropped=%u",
+        if (lost) DEBUG_WARN(DEBUG_CATEGORY_CALL, "voice-io: playback progress overflow session=%llu dropped=%u",
                             (unsigned long long)m_session, lost);
         if (m_playback.size() < 2) {
             Playback block{}; block.epoch = epoch; block.position = position;
@@ -282,7 +291,7 @@ void VoiceAudioIo::captureLoop() {
             if (mark < markerCount) applyMarker(markers[mark].kind, markers[mark].value);
             begin = end;
         }
-        if (m_kind == Kind::Call && frames == 960) call_audio_feed_prepared_pcm(m_session, cleaned, muteMask);
+        if (m_kind == Kind::Call && frames == 960) call_audio_feed_prepared_pcm(m_session, cleaned, muteMask, frameTime);
         filled = 0; markerCount = 0;
     };
     // Отчёт работает и при отсутствии PCM, иначе длительный провал захвата невидим.
@@ -350,6 +359,7 @@ void VoiceAudioIo::captureLoop() {
             filled = 0; markerCount = 0;
             if (aec) speex_aec_reset(aec);
             if (m_kind == Kind::Radio) radio_audio_capture_discontinuity(m_session);
+            else call_audio_reset_io(m_session, 0);
         }
         expectedSequence = event.sequence + event.frames;
         captureRoute = event.route;

tools/chatgui/src/voice_audio_io.h

@@ -27,7 +27,7 @@ public:
     /* firstTimeUs — монотонное время первого физического PCM-фрейма (backend либо оценка). */
     void capture(const int16_t* pcm, unsigned frames, int64_t firstTimeUs, uint64_t route = 0, bool missing = false);
     void reference(const float* pcm, unsigned frames, int64_t firstTimeUs);
-    void mix(float* stereo, unsigned frames, int64_t firstTimeUs = 0);
+    void mix(float* stereo, unsigned frames, int64_t firstTimeUs = 0, bool measured = false);
 private:
     enum class EventKind { Pcm, Ptt, Mute, Aec, Stop };
     struct Event {
@@ -41,7 +41,7 @@ private:
         int16_t pcm[1920]{};
     };
     struct Playback { uint64_t epoch = 0, position = 0; int16_t pcm[1920]{}; };
-    struct PlaybackProgress { uint64_t epoch = 0, position = 0; int64_t endTimeUs = 0; };
+    struct PlaybackProgress { uint64_t epoch = 0, position = 0; int64_t endTimeUs = 0; bool measured = false; };
     AudioEventQueue<Event, 32> m_captureEvents, m_references;
     AudioEventQueue<Playback, 12> m_playback;
     AudioEventQueue<PlaybackProgress, 32> m_playProgress;