fix(audio): simplify uac sink and prebuild sync harness

Cargo.lock

@@ -1642,7 +1642,7 @@ checksum = "09edd9e8b54e49e587e4f6295a7d29c3ea94d469cb40ab8ca70b288248a81db2"
 
 [[package]]
 name = "lesavka_client"
-version = "0.13.12"
+version = "0.13.13"
 dependencies = [
  "anyhow",
  "async-stream",
@@ -1676,7 +1676,7 @@ dependencies = [
 
 [[package]]
 name = "lesavka_common"
-version = "0.13.12"
+version = "0.13.13"
 dependencies = [
  "anyhow",
  "base64",
@@ -1688,7 +1688,7 @@ dependencies = [
 
 [[package]]
 name = "lesavka_server"
-version = "0.13.12"
+version = "0.13.13"
 dependencies = [
  "anyhow",
  "base64",

client/Cargo.toml

@@ -4,7 +4,7 @@ path                    = "src/main.rs"
 
 [package]
 name                    = "lesavka_client"
-version                 = "0.13.12"
+version                 = "0.13.13"
 edition                 = "2024"
 
 [dependencies]

common/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name                    = "lesavka_common"
-version                 = "0.13.12"
+version                 = "0.13.13"
 edition                 = "2024"
 build                   = "build.rs"
 

scripts/ci/quality_gate_baseline.json

@@ -278,7 +278,7 @@
     },
     "server/src/audio/voice_input.rs": {
       "line_percent": 100.0,
-      "loc": 426
+      "loc": 469
     },
     "server/src/bin/lesavka_uvc/control_payloads.rs": {
       "line_percent": 100.0,

scripts/manual/run_upstream_av_sync.sh

@@ -22,8 +22,12 @@ VIDEO_SIZE=${VIDEO_SIZE:-1280x720}
 VIDEO_FPS=${VIDEO_FPS:-30}
 VIDEO_FORMAT=${VIDEO_FORMAT:-mjpeg}
 REMOTE_AUDIO_SOURCE=${REMOTE_AUDIO_SOURCE:-auto}
+REMOTE_AUDIO_QUIESCE_USER_AUDIO=${REMOTE_AUDIO_QUIESCE_USER_AUDIO:-auto}
 SSH_OPTS=${SSH_OPTS:-"-o BatchMode=yes -o ConnectTimeout=5"}
 LOCAL_AUDIO_SANITY=${LOCAL_AUDIO_SANITY:-1}
+PROBE_PREBUILD=${PROBE_PREBUILD:-1}
+PROBE_BIN=${PROBE_BIN:-"${REPO_ROOT}/target/debug/lesavka-sync-probe"}
+ANALYZE_BIN=${ANALYZE_BIN:-"${REPO_ROOT}/target/debug/lesavka-sync-analyze"}
 
 mkdir -p "${LOCAL_OUTPUT_DIR}"
 STAMP="$(date +%Y%m%d-%H%M%S)"
@@ -34,6 +38,23 @@ if [[ "${LOCAL_AUDIO_SANITY}" != "0" ]]; then
   "${SCRIPT_DIR}/run_local_audio_sanity.sh"
 fi
 
+if [[ "${PROBE_PREBUILD}" != "0" ]]; then
+  echo "==> prebuilding sync probe/analyzer before opening the capture window"
+  (
+    cd "${REPO_ROOT}"
+    cargo build -p lesavka_client --bin lesavka-sync-probe --bin lesavka-sync-analyze
+  )
+fi
+
+if [[ ! -x "${PROBE_BIN}" ]]; then
+  echo "sync probe binary not found at ${PROBE_BIN}" >&2
+  exit 1
+fi
+if [[ ! -x "${ANALYZE_BIN}" ]]; then
+  echo "sync analyzer binary not found at ${ANALYZE_BIN}" >&2
+  exit 1
+fi
+
 echo "==> starting Tethys capture on ${TETHYS_HOST}"
 ssh ${SSH_OPTS} "${TETHYS_HOST}" bash -s -- \
   "${REMOTE_CAPTURE}" \
@@ -41,7 +62,8 @@ ssh ${SSH_OPTS} "${TETHYS_HOST}" bash -s -- \
   "${VIDEO_SIZE}" \
   "${VIDEO_FPS}" \
   "${VIDEO_FORMAT}" \
-  "${REMOTE_AUDIO_SOURCE}" <<'REMOTE_CAPTURE_SCRIPT' &
+  "${REMOTE_AUDIO_SOURCE}" \
+  "${REMOTE_AUDIO_QUIESCE_USER_AUDIO}" <<'REMOTE_CAPTURE_SCRIPT' &
 set -euo pipefail
 remote_capture=$1
 capture_seconds=$2
@@ -49,6 +71,7 @@ video_size=$3
 video_fps=$4
 video_format=$5
 remote_audio_source=$6
+remote_audio_quiesce_user_audio=$7
 
 rm -f "${remote_capture}"
 video_args=(-f video4linux2 -framerate "${video_fps}" -video_size "${video_size}")
@@ -56,6 +79,18 @@ if [[ -n "${video_format}" ]]; then
   video_args+=(-input_format "${video_format}")
 fi
 
+restore_user_audio() {
+  systemctl --user start pipewire.socket pipewire-pulse.socket wireplumber.service >/dev/null 2>&1 || true
+  sleep 1
+  systemctl --user start pipewire.service pipewire-pulse.service >/dev/null 2>&1 || true
+}
+
+quiesce_user_audio() {
+  systemctl --user stop pipewire-pulse.service pipewire.service wireplumber.service \
+    pipewire-pulse.socket pipewire.socket >/dev/null 2>&1 || true
+  sleep 1
+}
+
 resolve_pulse_source() {
   if ! command -v pactl >/dev/null 2>&1; then
     return 1
@@ -91,6 +126,31 @@ else
   exit 64
 fi
 
+quiesce_for_alsa=0
+case "${remote_audio_quiesce_user_audio}" in
+  1|true|yes)
+    quiesce_for_alsa=1
+    ;;
+  auto)
+    if [[ "${audio_mode}" == "alsa" ]]; then
+      quiesce_for_alsa=1
+    fi
+    ;;
+  0|false|no)
+    quiesce_for_alsa=0
+    ;;
+  *)
+    printf 'unsupported REMOTE_AUDIO_QUIESCE_USER_AUDIO=%s\n' "${remote_audio_quiesce_user_audio}" >&2
+    exit 64
+    ;;
+esac
+
+if [[ "${quiesce_for_alsa}" == "1" ]]; then
+  printf 'quiescing Tethys user audio before raw ALSA capture\n' >&2
+  quiesce_user_audio
+  trap restore_user_audio EXIT
+fi
+
 if [[ "${audio_mode}" == "pulse" ]]; then
   printf 'using Pulse source: %s\n' "${pulse_source}" >&2
   ffmpeg -hide_banner -loglevel error -y \
@@ -126,7 +186,7 @@ echo "==> running local Lesavka sync probe against ${LESAVKA_SERVER_ADDR}"
 probe_status=0
 (
   cd "${REPO_ROOT}"
-  cargo run -p lesavka_client --bin lesavka-sync-probe -- \
+  "${PROBE_BIN}" \
     --server "${LESAVKA_SERVER_ADDR}" \
     --duration-seconds "${PROBE_DURATION_SECONDS}" \
     --warmup-seconds "${PROBE_WARMUP_SECONDS}"
@@ -158,7 +218,7 @@ fi
 echo "==> analyzing capture"
 (
   cd "${REPO_ROOT}"
-  cargo run -p lesavka_client --bin lesavka-sync-analyze -- "${LOCAL_CAPTURE}"
+  "${ANALYZE_BIN}" "${LOCAL_CAPTURE}"
 )
 
 echo "==> done"

server/Cargo.toml

@@ -10,7 +10,7 @@ bench                   = false
 
 [package]
 name                    = "lesavka_server"
-version                 = "0.13.12"
+version                 = "0.13.13"
 edition                 = "2024"
 autobins                = false
 

server/src/audio/voice_input.rs

@@ -119,6 +119,10 @@ fn voice_sink_session_clock_align_enabled() -> bool {
         .unwrap_or(true)
 }
 
+fn voice_sink_delay_queue_enabled(compensation_us: i64) -> bool {
+    compensation_us > 0
+}
+
 impl Voice {
     #[cfg(coverage)]
     pub async fn new(_alsa_dev: &str) -> anyhow::Result<Self> {
@@ -210,24 +214,31 @@ impl Voice {
         let latency_time_us = voice_sink_latency_time_us();
         let compensation_us = voice_sink_compensation_us();
         let clock_align_enabled = voice_sink_session_clock_align_enabled();
+        let compensation_ns = (compensation_us.max(0) as u64).saturating_mul(1_000);
+        let delay_queue_enabled = voice_sink_delay_queue_enabled(compensation_us);
 
         alsa_sink.set_property("device", alsa_dev);
         alsa_sink.set_property("sync", clock_align_enabled);
-        alsa_sink.set_property("async", clock_align_enabled);
+        // The UAC gadget is a live sink, so async preroll buys us little and
+        // has repeatedly stranded post-queue buffers behind a sink that never
+        // fully arms. Keep timestamp sync, but let the sink start immediately.
+        alsa_sink.set_property("async", false);
         alsa_sink.set_property("enable-last-sample", false);
         alsa_sink.set_property("provide-clock", false);
         alsa_sink.set_property("buffer-time", buffer_time_us);
         alsa_sink.set_property("latency-time", latency_time_us);
-        let compensation_ns = (compensation_us.max(0) as u64).saturating_mul(1_000);
+        if delay_queue_enabled {
-        delay_queue.set_property("max-size-buffers", 0u32);
+            delay_queue.set_property("max-size-buffers", 0u32);
-        delay_queue.set_property("max-size-bytes", 0u32);
+            delay_queue.set_property("max-size-bytes", 0u32);
-        delay_queue.set_property("max-size-time", compensation_ns);
+            delay_queue.set_property("max-size-time", compensation_ns);
-        delay_queue.set_property("min-threshold-time", compensation_ns);
+            delay_queue.set_property("min-threshold-time", compensation_ns);
+        }
         tracing::info!(
             %alsa_dev,
             buffer_time_us,
             latency_time_us,
             compensation_us,
+            delay_queue_enabled,
             clock_align_enabled,
             "🎤 UAC sink low-latency timing armed"
         );
@@ -235,27 +246,51 @@ impl Voice {
             crate::media_timing::prepare_pipeline_clock_sync(&pipeline);
         }
 
-        pipeline.add_many([
+        if delay_queue_enabled {
-            appsrc.upcast_ref(),
+            pipeline.add_many([
-            &decodebin,
+                appsrc.upcast_ref(),
-            &convert,
+                &decodebin,
-            &resample,
+                &convert,
-            &capsfilter,
+                &resample,
-            &level,
+                &capsfilter,
-            &delay_queue,
+                &level,
-            &post_level,
+                &delay_queue,
-            &alsa_sink,
+                &post_level,
-        ])?;
+                &alsa_sink,
+            ])?;
+        } else {
+            pipeline.add_many([
+                appsrc.upcast_ref(),
+                &decodebin,
+                &convert,
+                &resample,
+                &capsfilter,
+                &level,
+                &post_level,
+                &alsa_sink,
+            ])?;
+        }
         appsrc.link(&decodebin)?;
-        gst::Element::link_many([
+        if delay_queue_enabled {
-            &convert,
+            gst::Element::link_many([
-            &resample,
+                &convert,
-            &capsfilter,
+                &resample,
-            &level,
+                &capsfilter,
-            &delay_queue,
+                &level,
-            &post_level,
+                &delay_queue,
-            &alsa_sink,
+                &post_level,
-        ])?;
+                &alsa_sink,
+            ])?;
+        } else {
+            gst::Element::link_many([
+                &convert,
+                &resample,
+                &capsfilter,
+                &level,
+                &post_level,
+                &alsa_sink,
+            ])?;
+        }
 
         /*------------ decodebin autolink ----------------*/
         let convert_sink = convert
@@ -423,4 +458,12 @@ mod voice_sink_timing_tests {
             assert!(super::voice_sink_session_clock_align_enabled());
         });
     }
+
+    #[test]
+    fn delay_queue_turns_on_only_for_positive_compensation() {
+        assert!(!super::voice_sink_delay_queue_enabled(-1));
+        assert!(!super::voice_sink_delay_queue_enabled(0));
+        assert!(super::voice_sink_delay_queue_enabled(1));
+        assert!(super::voice_sink_delay_queue_enabled(90_000));
+    }
 }