lesavka/scripts/manual/run_upstream_av_sync.sh

#!/usr/bin/env bash
# scripts/manual/run_upstream_av_sync.sh
# Manual: upstream A/V sync hardware probe; not part of CI.
#
# Manual: capture the real Tethys UVC/UAC endpoints while the Lesavka server
# generates paired probe media locally and feeds its own UVC/UAC output sinks.
# This intentionally measures only the server-to-host output-device skew.

set -euo pipefail

SCRIPT_DIR="$(cd -- "$(dirname "${BASH_SOURCE[0]}")" >/dev/null 2>&1 && pwd)"
REPO_ROOT="$(cd -- "${SCRIPT_DIR}/../.." >/dev/null 2>&1 && pwd)"

TETHYS_HOST=${TETHYS_HOST:-tethys}
LESAVKA_SERVER_HOST=${LESAVKA_SERVER_HOST:-theia}
LESAVKA_SERVER_CONNECT_HOST=${LESAVKA_SERVER_CONNECT_HOST:-38.28.125.112}
LESAVKA_SERVER_ADDR=${LESAVKA_SERVER_ADDR:-auto}
LESAVKA_SERVER_SCHEME=${LESAVKA_SERVER_SCHEME:-https}
LESAVKA_TLS_DOMAIN=${LESAVKA_TLS_DOMAIN:-lesavka-server}
PROBE_DURATION_SECONDS=${PROBE_DURATION_SECONDS:-20}
PROBE_WARMUP_SECONDS=${PROBE_WARMUP_SECONDS:-4}
PROBE_START_GRACE_SECONDS=${PROBE_START_GRACE_SECONDS:-20}
PROBE_TIMEOUT_SECONDS=${PROBE_TIMEOUT_SECONDS:-$((PROBE_DURATION_SECONDS + PROBE_WARMUP_SECONDS + PROBE_START_GRACE_SECONDS))}
PROBE_PULSE_PERIOD_MS=${PROBE_PULSE_PERIOD_MS:-1000}
PROBE_PULSE_WIDTH_MS=${PROBE_PULSE_WIDTH_MS:-120}
PROBE_EVENT_WIDTH_CODES=${PROBE_EVENT_WIDTH_CODES:-1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16}
PROBE_CONDITIONING_SECONDS=${PROBE_CONDITIONING_SECONDS:-0}
PROBE_CONDITIONING_WARMUP_SECONDS=${PROBE_CONDITIONING_WARMUP_SECONDS:-1}
PROBE_CONDITIONING_GAP_SECONDS=${PROBE_CONDITIONING_GAP_SECONDS:-1}
PROBE_CONDITIONING_EVENT_WIDTH_CODES=${PROBE_CONDITIONING_EVENT_WIDTH_CODES:-${PROBE_EVENT_WIDTH_CODES}}
PROBE_CONDITIONING_TIMEOUT_SECONDS=${PROBE_CONDITIONING_TIMEOUT_SECONDS:-$((PROBE_CONDITIONING_SECONDS + PROBE_CONDITIONING_WARMUP_SECONDS + PROBE_START_GRACE_SECONDS))}
LESAVKA_OUTPUT_DELAY_PROBE_AUDIO_DELAY_US=${LESAVKA_OUTPUT_DELAY_PROBE_AUDIO_DELAY_US:-0}
LESAVKA_OUTPUT_DELAY_PROBE_VIDEO_DELAY_US=${LESAVKA_OUTPUT_DELAY_PROBE_VIDEO_DELAY_US:-0}
# Do not open the UVC host capture far ahead of the probe. The gadget side only
# has frames once the sync probe is feeding the server, and some hosts time out
# VIDIOC_STREAMON if the camera is starved during pre-roll.
LEAD_IN_SECONDS=${LEAD_IN_SECONDS:-0}
TAIL_SECONDS=${TAIL_SECONDS:-2}
CAPTURE_START_GRACE_SECONDS=${CAPTURE_START_GRACE_SECONDS:-12}
CONDITIONING_CAPTURE_SECONDS=$((PROBE_CONDITIONING_SECONDS > 0 ? PROBE_CONDITIONING_SECONDS + PROBE_CONDITIONING_WARMUP_SECONDS + PROBE_CONDITIONING_GAP_SECONDS : 0))
CAPTURE_SECONDS=${CAPTURE_SECONDS:-$((PROBE_DURATION_SECONDS + PROBE_WARMUP_SECONDS + CONDITIONING_CAPTURE_SECONDS + CAPTURE_START_GRACE_SECONDS + LEAD_IN_SECONDS + TAIL_SECONDS))}
LOCAL_OUTPUT_DIR=${LOCAL_OUTPUT_DIR:-/tmp}
REMOTE_VIDEO_DEVICE=${REMOTE_VIDEO_DEVICE:-auto}
VIDEO_SIZE=${VIDEO_SIZE:-auto}
VIDEO_FPS=${VIDEO_FPS:-auto}
VIDEO_FORMAT=${VIDEO_FORMAT:-mjpeg}
REMOTE_CAPTURE_STACK=${REMOTE_CAPTURE_STACK:-pulse}
REMOTE_PULSE_CAPTURE_TOOL=${REMOTE_PULSE_CAPTURE_TOOL:-gst}
REMOTE_PULSE_VIDEO_MODE=${REMOTE_PULSE_VIDEO_MODE:-copy}
REMOTE_PULSE_AUDIO_ANCHOR_SILENCE=${REMOTE_PULSE_AUDIO_ANCHOR_SILENCE:-1}
REMOTE_AUDIO_SOURCE=${REMOTE_AUDIO_SOURCE:-auto}
REMOTE_AUDIO_QUIESCE_USER_AUDIO=${REMOTE_AUDIO_QUIESCE_USER_AUDIO:-auto}
REMOTE_CAPTURE_ALLOW_ALSA_FALLBACK=${REMOTE_CAPTURE_ALLOW_ALSA_FALLBACK:-0}
REMOTE_CAPTURE_PREROLL_DISCARD_SECONDS=${REMOTE_CAPTURE_PREROLL_DISCARD_SECONDS:-0}
REMOTE_CAPTURE_READY_SETTLE_SECONDS=${REMOTE_CAPTURE_READY_SETTLE_SECONDS:-1}
ANALYSIS_NORMALIZE=${ANALYSIS_NORMALIZE:-0}
ANALYSIS_SCALE_WIDTH=${ANALYSIS_SCALE_WIDTH:-1280}
ANALYSIS_TIMELINE_WINDOW=${ANALYSIS_TIMELINE_WINDOW:-0}
ANALYSIS_TIMELINE_WINDOW_PADDING_SECONDS=${ANALYSIS_TIMELINE_WINDOW_PADDING_SECONDS:-1.0}
SSH_OPTS=${SSH_OPTS:-"-o BatchMode=yes -o ConnectTimeout=30"}
PROBE_PREBUILD=${PROBE_PREBUILD:-1}
ANALYZE_BIN=${ANALYZE_BIN:-"${REPO_ROOT}/target/debug/lesavka-sync-analyze"}
REMOTE_ANALYZE=${REMOTE_ANALYZE:-1}
REMOTE_ANALYZE_BIN=${REMOTE_ANALYZE_BIN:-/tmp/lesavka-sync-analyze}
REMOTE_ANALYZE_COPY=${REMOTE_ANALYZE_COPY:-1}
FETCH_CAPTURE=${FETCH_CAPTURE:-1}
REMOTE_ARTIFACT_RETRIES=${REMOTE_ARTIFACT_RETRIES:-3}
REMOTE_ARTIFACT_RETRY_DELAY_SECONDS=${REMOTE_ARTIFACT_RETRY_DELAY_SECONDS:-5}
REMOTE_SERVER_PREFLIGHT=${REMOTE_SERVER_PREFLIGHT:-1}
REMOTE_EXPECT_CAM_OUTPUT=${REMOTE_EXPECT_CAM_OUTPUT:-uvc}
REMOTE_EXPECT_UVC_CODEC=${REMOTE_EXPECT_UVC_CODEC:-mjpeg}
REMOTE_EXPECT_UVC_WIDTH=${REMOTE_EXPECT_UVC_WIDTH:-}
REMOTE_EXPECT_UVC_HEIGHT=${REMOTE_EXPECT_UVC_HEIGHT:-}
REMOTE_EXPECT_UVC_FPS=${REMOTE_EXPECT_UVC_FPS:-}
LESAVKA_OUTPUT_DELAY_CALIBRATION=${LESAVKA_OUTPUT_DELAY_CALIBRATION:-1}
LESAVKA_OUTPUT_DELAY_APPLY=${LESAVKA_OUTPUT_DELAY_APPLY:-0}
LESAVKA_OUTPUT_DELAY_APPLY_MODE=${LESAVKA_OUTPUT_DELAY_APPLY_MODE:-absolute}
LESAVKA_OUTPUT_DELAY_CONFIRM=${LESAVKA_OUTPUT_DELAY_CONFIRM:-1}
LESAVKA_OUTPUT_DELAY_SAVE=${LESAVKA_OUTPUT_DELAY_SAVE:-0}
LESAVKA_OUTPUT_REQUIRE_SYNC_PASS=${LESAVKA_OUTPUT_REQUIRE_SYNC_PASS:-0}
LESAVKA_OUTPUT_DELAY_TARGET=${LESAVKA_OUTPUT_DELAY_TARGET:-video}
LESAVKA_OUTPUT_DELAY_MIN_PAIRS=${LESAVKA_OUTPUT_DELAY_MIN_PAIRS:-13}
LESAVKA_OUTPUT_DELAY_MAX_ABS_SKEW_MS=${LESAVKA_OUTPUT_DELAY_MAX_ABS_SKEW_MS:-5000}
LESAVKA_OUTPUT_DELAY_MAX_DRIFT_MS=${LESAVKA_OUTPUT_DELAY_MAX_DRIFT_MS:-80}
LESAVKA_OUTPUT_DELAY_GAIN=${LESAVKA_OUTPUT_DELAY_GAIN:-1.0}
LESAVKA_OUTPUT_DELAY_MAX_STEP_US=${LESAVKA_OUTPUT_DELAY_MAX_STEP_US:-1500000}
LESAVKA_OUTPUT_FRESHNESS_MAX_AGE_MS=${LESAVKA_OUTPUT_FRESHNESS_MAX_AGE_MS:-1000}
LESAVKA_OUTPUT_FRESHNESS_MAX_DRIFT_MS=${LESAVKA_OUTPUT_FRESHNESS_MAX_DRIFT_MS:-100}
LESAVKA_OUTPUT_FRESHNESS_MAX_CLOCK_UNCERTAINTY_MS=${LESAVKA_OUTPUT_FRESHNESS_MAX_CLOCK_UNCERTAINTY_MS:-250}
LESAVKA_OUTPUT_FRESHNESS_MIN_PAIRS=${LESAVKA_OUTPUT_FRESHNESS_MIN_PAIRS:-${LESAVKA_OUTPUT_DELAY_MIN_PAIRS}}
LESAVKA_CLOCK_ALIGNMENT_SAMPLES=${LESAVKA_CLOCK_ALIGNMENT_SAMPLES:-5}
CAPTURE_READY_MARKER="__LESAVKA_CAPTURE_READY__"

STAMP="$(date +%Y%m%d-%H%M%S)"
REMOTE_CAPTURE=${REMOTE_CAPTURE:-"/tmp/lesavka-output-delay-probe-${STAMP}.mkv"}
LOCAL_REPORT_DIR="${LOCAL_OUTPUT_DIR%/}/lesavka-output-delay-probe-${STAMP}"
LOCAL_CAPTURE="${LOCAL_REPORT_DIR}/capture.mkv"
LOCAL_ANALYSIS_JSON="${LOCAL_REPORT_DIR}/report.json"
LOCAL_ANALYSIS_ERROR_LOG="${LOCAL_REPORT_DIR}/analysis-error.log"
LOCAL_REPORT_TXT="${LOCAL_REPORT_DIR}/report.txt"
LOCAL_EVENTS_CSV="${LOCAL_REPORT_DIR}/events.csv"
LOCAL_SERVER_PROBE_REPLY="${LOCAL_REPORT_DIR}/server-output-probe-reply.txt"
LOCAL_SERVER_TIMELINE_JSON="${LOCAL_REPORT_DIR}/server-output-timeline.json"
LOCAL_SIGNAL_CONDITIONING_REPLY="${LOCAL_REPORT_DIR}/signal-conditioning-probe-reply.txt"
LOCAL_SIGNAL_CONDITIONING_TIMELINE_JSON="${LOCAL_REPORT_DIR}/signal-conditioning-timeline.json"
LOCAL_OUTPUT_DELAY_CORRELATION_JSON="${LOCAL_REPORT_DIR}/output-delay-correlation.json"
LOCAL_OUTPUT_DELAY_CORRELATION_CSV="${LOCAL_REPORT_DIR}/output-delay-correlation.csv"
LOCAL_OUTPUT_DELAY_CORRELATION_TXT="${LOCAL_REPORT_DIR}/output-delay-correlation.txt"
LOCAL_CLOCK_ALIGNMENT_JSON="${LOCAL_REPORT_DIR}/clock-alignment.json"
LOCAL_OUTPUT_DELAY_JSON="${LOCAL_REPORT_DIR}/output-delay-calibration.json"
LOCAL_OUTPUT_DELAY_ENV="${LOCAL_REPORT_DIR}/output-delay-calibration.env"
LOCAL_CAPTURE_LOG="${LOCAL_REPORT_DIR}/capture.log"
mkdir -p "${LOCAL_REPORT_DIR}"
RESOLVED_LESAVKA_SERVER_ADDR=""
SERVER_TUNNEL_PID=""
SERVER_TUNNEL_REMOTE_PORT=""
SERVER_TUNNEL_LOCAL_PORT=""

cleanup_server_tunnel() {
  if [[ -z "${SERVER_TUNNEL_PID}" ]]; then
    return 0
  fi
  if kill -0 "${SERVER_TUNNEL_PID}" >/dev/null 2>&1; then
    kill "${SERVER_TUNNEL_PID}" >/dev/null 2>&1 || true
    wait "${SERVER_TUNNEL_PID}" >/dev/null 2>&1 || true
  fi
}

trap cleanup_server_tunnel EXIT

pick_local_server_tunnel_port() {
  python3 - <<'PY'
import socket

with socket.socket() as sock:
    sock.bind(("127.0.0.1", 0))
    print(sock.getsockname()[1])
PY
}

sample_host_clock_offset_ns() {
  local host=$1
  local before_ns remote_ns after_ns
  before_ns="$(date +%s%N)"
  remote_ns="$(ssh ${SSH_OPTS} "${host}" 'date +%s%N')"
  after_ns="$(date +%s%N)"
  python3 - <<'PY' "${before_ns}" "${remote_ns}" "${after_ns}"
import sys

before_ns, remote_ns, after_ns = (int(value) for value in sys.argv[1:])
local_mid_ns = (before_ns + after_ns) // 2
print(f"{remote_ns - local_mid_ns} {(after_ns - before_ns) // 2} {after_ns - before_ns}")
PY
}

sample_best_host_clock_offset_ns() {
  local host=$1
  local samples=$2
  if python3 - <<'PY' "${host}" "${samples}" "${SSH_OPTS}"; then
import shlex
import subprocess
import sys
import time

host = sys.argv[1]
try:
    samples = max(1, int(sys.argv[2]))
except Exception:
    samples = 5
ssh_opts = shlex.split(sys.argv[3])
remote_code = (
    "import sys,time\n"
    "for _line in sys.stdin:\n"
    "    print(time.time_ns(), flush=True)\n"
)
cmd = [
    "ssh",
    *ssh_opts,
    host,
    "python3 -u -c " + shlex.quote(remote_code),
]
try:
    proc = subprocess.Popen(
        cmd,
        stdin=subprocess.PIPE,
        stdout=subprocess.PIPE,
        stderr=subprocess.DEVNULL,
        text=True,
        bufsize=1,
    )
except Exception:
    raise SystemExit(1)

rows = []
try:
    for _ in range(samples):
        if proc.stdin is None or proc.stdout is None:
            break
        before_ns = time.time_ns()
        try:
            proc.stdin.write("x\n")
            proc.stdin.flush()
            line = proc.stdout.readline()
        except (BrokenPipeError, OSError):
            break
        after_ns = time.time_ns()
        if not line:
            break
        try:
            remote_ns = int(line.strip())
        except Exception:
            continue
        local_mid_ns = (before_ns + after_ns) // 2
        rtt_ns = after_ns - before_ns
        rows.append((rtt_ns // 2, remote_ns - local_mid_ns, rtt_ns))
finally:
    try:
        if proc.stdin is not None:
            proc.stdin.close()
    except Exception:
        pass
    try:
        proc.wait(timeout=2)
    except Exception:
        try:
            proc.kill()
        except Exception:
            pass

if not rows:
    raise SystemExit(1)
uncertainty_ns, offset_ns, rtt_ns = min(rows)
print(f"{offset_ns} {uncertainty_ns} {rtt_ns} {len(rows)} persistent-ssh-python")
PY
    return 0
  fi

  local tmp
  tmp="$(mktemp)"
  local i=0
  while (( i < samples )); do
    sample_host_clock_offset_ns "${host}" >>"${tmp}" 2>/dev/null || true
    ((i += 1))
  done
  python3 - <<'PY' "${tmp}"
import pathlib
import sys

rows = []
for line in pathlib.Path(sys.argv[1]).read_text().splitlines():
    try:
        offset_ns, uncertainty_ns, rtt_ns = (int(value) for value in line.split())
    except Exception:
        continue
    rows.append((uncertainty_ns, offset_ns, rtt_ns))
if not rows:
    raise SystemExit(1)
uncertainty_ns, offset_ns, rtt_ns = min(rows)
print(f"{offset_ns} {uncertainty_ns} {rtt_ns} {len(rows)} ssh-date")
PY
  local rc=$?
  rm -f "${tmp}"
  return "${rc}"
}

sample_server_to_capture_clock_offset_ns() {
  local server_host=$1
  local capture_host=$2
  local samples=$3
  python3 - <<'PY' "${server_host}" "${capture_host}" "${samples}" "${SSH_OPTS}"
import shlex
import subprocess
import sys

server_host = sys.argv[1]
capture_host = sys.argv[2]
try:
    samples = max(1, int(sys.argv[3]))
except Exception:
    samples = 5
ssh_opts_text = sys.argv[4]
ssh_opts = shlex.split(ssh_opts_text)
remote_code = r'''
import shlex
import subprocess
import sys
import time

capture_host = sys.argv[1]
samples = max(1, int(sys.argv[2]))
ssh_opts = shlex.split(sys.argv[3])
capture_code = (
    "import sys,time\n"
    "for _line in sys.stdin:\n"
    "    print(time.time_ns(), flush=True)\n"
)
cmd = [
    "ssh",
    *ssh_opts,
    capture_host,
    "python3 -u -c " + shlex.quote(capture_code),
]
try:
    proc = subprocess.Popen(
        cmd,
        stdin=subprocess.PIPE,
        stdout=subprocess.PIPE,
        stderr=subprocess.DEVNULL,
        text=True,
        bufsize=1,
    )
except Exception:
    raise SystemExit(1)

rows = []
try:
    for _ in range(samples):
        if proc.stdin is None or proc.stdout is None:
            break
        before_ns = time.time_ns()
        try:
            proc.stdin.write("x\n")
            proc.stdin.flush()
            line = proc.stdout.readline()
        except (BrokenPipeError, OSError):
            break
        after_ns = time.time_ns()
        if not line:
            break
        try:
            remote_ns = int(line.strip())
        except Exception:
            continue
        local_mid_ns = (before_ns + after_ns) // 2
        rtt_ns = after_ns - before_ns
        rows.append((rtt_ns // 2, remote_ns - local_mid_ns, rtt_ns))
finally:
    try:
        if proc.stdin is not None:
            proc.stdin.close()
    except Exception:
        pass
    try:
        proc.wait(timeout=2)
    except Exception:
        try:
            proc.kill()
        except Exception:
            pass

if not rows:
    raise SystemExit(1)
uncertainty_ns, offset_ns, rtt_ns = min(rows)
print(f"{offset_ns} {uncertainty_ns} {rtt_ns} {len(rows)} server-to-capture-persistent-ssh-python")
'''
remote_cmd = (
    "python3 -u -c "
    + shlex.quote(remote_code)
    + " "
    + shlex.quote(capture_host)
    + " "
    + shlex.quote(str(samples))
    + " "
    + shlex.quote(ssh_opts_text)
)
result = subprocess.run(
    ["ssh", *ssh_opts, server_host, remote_cmd],
    check=False,
    stdout=subprocess.PIPE,
    stderr=subprocess.DEVNULL,
    text=True,
)
if result.returncode != 0:
    raise SystemExit(result.returncode)
for line in result.stdout.splitlines():
    if line.strip():
        print(line.strip())
        raise SystemExit(0)
raise SystemExit(1)
PY
}

write_clock_alignment() {
  echo "==> sampling Theia/Tethys clock alignment for freshness"
  local direct_sample
  if direct_sample="$(sample_server_to_capture_clock_offset_ns "${LESAVKA_SERVER_HOST}" "${TETHYS_HOST}" "${LESAVKA_CLOCK_ALIGNMENT_SAMPLES}")"; then
    python3 - <<'PY' \
      "${LESAVKA_SERVER_HOST}" \
      "${TETHYS_HOST}" \
      "${direct_sample}" \
      "${LOCAL_CLOCK_ALIGNMENT_JSON}"
import json
import pathlib
import sys

server_host, capture_host, sample, output_path = sys.argv[1:]
parts = sample.split()
offset_ns, uncertainty_ns, rtt_ns, sample_count = (int(value) for value in parts[:4])
method = parts[4] if len(parts) > 4 else "server-to-capture-ssh"
artifact = {
    "schema": "lesavka.clock-alignment.v1",
    "available": True,
    "method": "server host to capture host persistent ssh midpoint",
    "server_host": server_host,
    "capture_host": capture_host,
    "server_clock_offset_from_local_ns": None,
    "capture_clock_offset_from_local_ns": None,
    "theia_to_tethys_offset_ns": offset_ns,
    "uncertainty_ns": uncertainty_ns,
    "uncertainty_ms": uncertainty_ns / 1_000_000.0,
    "server_sample_rtt_ns": 0,
    "capture_sample_rtt_ns": rtt_ns,
    "server_samples": sample_count,
    "capture_samples": sample_count,
    "server_sample_method": "server-host-local-clock",
    "capture_sample_method": method,
}
pathlib.Path(output_path).write_text(json.dumps(artifact, indent=2, sort_keys=True) + "\n")
print(f"   ↪ theia_to_tethys_offset_ms={offset_ns / 1_000_000.0:+.3f}")
print(f"   ↪ clock_alignment_uncertainty_ms={uncertainty_ns / 1_000_000.0:.3f}")
PY
    return 0
  fi
  echo "   ↪ server-to-capture clock alignment unavailable; falling back to client-mediated SSH samples"

  local theia_sample tethys_sample
  if ! theia_sample="$(sample_best_host_clock_offset_ns "${LESAVKA_SERVER_HOST}" "${LESAVKA_CLOCK_ALIGNMENT_SAMPLES}")"; then
    echo "   ↪ clock alignment unavailable: failed to sample ${LESAVKA_SERVER_HOST}"
    printf '{"schema":"lesavka.clock-alignment.v1","available":false,"reason":"failed to sample server host"}\n' >"${LOCAL_CLOCK_ALIGNMENT_JSON}"
    return 0
  fi
  if ! tethys_sample="$(sample_best_host_clock_offset_ns "${TETHYS_HOST}" "${LESAVKA_CLOCK_ALIGNMENT_SAMPLES}")"; then
    echo "   ↪ clock alignment unavailable: failed to sample ${TETHYS_HOST}"
    printf '{"schema":"lesavka.clock-alignment.v1","available":false,"reason":"failed to sample capture host"}\n' >"${LOCAL_CLOCK_ALIGNMENT_JSON}"
    return 0
  fi

  python3 - <<'PY' \
    "${LESAVKA_SERVER_HOST}" \
    "${TETHYS_HOST}" \
    "${theia_sample}" \
    "${tethys_sample}" \
    "${LOCAL_CLOCK_ALIGNMENT_JSON}"
import json
import pathlib
import sys

server_host, capture_host, server_sample, capture_sample, output_path = sys.argv[1:]
server_parts = server_sample.split()
capture_parts = capture_sample.split()
server_offset_ns, server_uncertainty_ns, server_rtt_ns, server_samples = (int(value) for value in server_parts[:4])
capture_offset_ns, capture_uncertainty_ns, capture_rtt_ns, capture_samples = (int(value) for value in capture_parts[:4])
server_method = server_parts[4] if len(server_parts) > 4 else "ssh-date"
capture_method = capture_parts[4] if len(capture_parts) > 4 else "ssh-date"
theia_to_tethys_offset_ns = capture_offset_ns - server_offset_ns
uncertainty_ns = server_uncertainty_ns + capture_uncertainty_ns
artifact = {
    "schema": "lesavka.clock-alignment.v1",
    "available": True,
    "method": "persistent ssh midpoint" if server_method == capture_method == "persistent-ssh-python" else "ssh midpoint",
    "server_host": server_host,
    "capture_host": capture_host,
    "server_clock_offset_from_local_ns": server_offset_ns,
    "capture_clock_offset_from_local_ns": capture_offset_ns,
    "theia_to_tethys_offset_ns": theia_to_tethys_offset_ns,
    "uncertainty_ns": uncertainty_ns,
    "uncertainty_ms": uncertainty_ns / 1_000_000.0,
    "server_sample_rtt_ns": server_rtt_ns,
    "capture_sample_rtt_ns": capture_rtt_ns,
    "server_samples": server_samples,
    "capture_samples": capture_samples,
    "server_sample_method": server_method,
    "capture_sample_method": capture_method,
}
pathlib.Path(output_path).write_text(json.dumps(artifact, indent=2, sort_keys=True) + "\n")
print(
    f"   ↪ theia_to_tethys_offset_ms={theia_to_tethys_offset_ns / 1_000_000.0:+.3f}"
)
print(f"   ↪ clock_alignment_uncertainty_ms={uncertainty_ns / 1_000_000.0:.3f}")
PY
}

wait_for_server_tunnel() {
  local local_port=$1
  local tries=50
  local i=0
  while (( i < tries )); do
    if nc -z 127.0.0.1 "${local_port}" >/dev/null 2>&1; then
      return 0
    fi
    if ! kill -0 "${SERVER_TUNNEL_PID}" >/dev/null 2>&1; then
      wait "${SERVER_TUNNEL_PID}" >/dev/null 2>&1 || true
      echo "SSH tunnel to ${LESAVKA_SERVER_HOST} exited before becoming ready" >&2
      exit 88
    fi
    sleep 0.1
    ((i += 1))
  done

  echo "SSH tunnel to ${LESAVKA_SERVER_HOST} did not become ready on localhost:${local_port}" >&2
  exit 89
}

start_server_tunnel() {
  local remote_port=$1
  local local_port
  local_port="$(pick_local_server_tunnel_port)"
  echo "==> opening SSH tunnel to ${LESAVKA_SERVER_HOST}:127.0.0.1:${remote_port} on localhost:${local_port}"
  ssh ${SSH_OPTS} -o ExitOnForwardFailure=yes \
    -N \
    -L "127.0.0.1:${local_port}:127.0.0.1:${remote_port}" \
    "${LESAVKA_SERVER_HOST}" &
  SERVER_TUNNEL_PID=$!
  SERVER_TUNNEL_REMOTE_PORT="${remote_port}"
  SERVER_TUNNEL_LOCAL_PORT="${local_port}"
  wait_for_server_tunnel "${local_port}"
}

retry_remote_artifact_command() {
  local description=$1
  shift
  local attempt status
  for attempt in $(seq 1 "${REMOTE_ARTIFACT_RETRIES}"); do
    set +e
    "$@"
    status=$?
    set -e
    if [[ "${status}" -eq 0 ]]; then
      return 0
    fi
    if [[ "${attempt}" -lt "${REMOTE_ARTIFACT_RETRIES}" ]]; then
      printf '%s failed with status %s; retrying %s/%s in %ss\n' \
        "${description}" "${status}" "$((attempt + 1))" "${REMOTE_ARTIFACT_RETRIES}" \
        "${REMOTE_ARTIFACT_RETRY_DELAY_SECONDS}" >&2
      sleep "${REMOTE_ARTIFACT_RETRY_DELAY_SECONDS}"
    fi
  done
  printf '%s failed after %s attempt(s)\n' "${description}" "${REMOTE_ARTIFACT_RETRIES}" >&2
  return "${status}"
}

run_remote_sync_analysis_once() {
  local output_path=$1
  local error_path=${2:-/dev/null}
  ssh ${SSH_OPTS} "${TETHYS_HOST}" \
    "chmod +x '${REMOTE_ANALYZE_BIN}' && '${REMOTE_ANALYZE_BIN}' '${remote_fetch_capture}' --json --event-width-codes '${PROBE_EVENT_WIDTH_CODES}' ${analysis_window_arg}" \
    > "${output_path}" 2> "${error_path}"
}

resolve_server_addr() {
  if [[ "${LESAVKA_SERVER_ADDR}" != "auto" ]]; then
    RESOLVED_LESAVKA_SERVER_ADDR="${LESAVKA_SERVER_ADDR}"
    return 0
  fi

  local bind_addr port
  bind_addr="$(
    ssh ${SSH_OPTS} "${LESAVKA_SERVER_HOST}" \
      "grep -E '^LESAVKA_SERVER_BIND_ADDR=' /etc/lesavka/server.env 2>/dev/null | tail -n1 | cut -d= -f2-" \
      2>/dev/null || true
  )"
  port="${bind_addr##*:}"
  if [[ "${port}" =~ ^[0-9]+$ ]]; then
    start_server_tunnel "${port}"
    RESOLVED_LESAVKA_SERVER_ADDR="${LESAVKA_SERVER_SCHEME}://127.0.0.1:${SERVER_TUNNEL_LOCAL_PORT}"
    return 0
  fi

  start_server_tunnel "50051"
  RESOLVED_LESAVKA_SERVER_ADDR="${LESAVKA_SERVER_SCHEME}://127.0.0.1:${SERVER_TUNNEL_LOCAL_PORT}"
}

preflight_server_path() {
  [[ "${REMOTE_SERVER_PREFLIGHT}" != "0" ]] || return 0

  echo "==> verifying Lesavka server path on ${LESAVKA_SERVER_HOST}"
  ssh ${SSH_OPTS} "${LESAVKA_SERVER_HOST}" bash -s -- \
    "${REMOTE_EXPECT_CAM_OUTPUT}" \
    "${REMOTE_EXPECT_UVC_CODEC}" \
    "${REMOTE_EXPECT_UVC_WIDTH}" \
    "${REMOTE_EXPECT_UVC_HEIGHT}" \
    "${REMOTE_EXPECT_UVC_FPS}" <<'REMOTE_PREFLIGHT'
set -euo pipefail
expect_cam_output=${1:-}
expect_uvc_codec=${2:-}
expect_uvc_width=${3:-}
expect_uvc_height=${4:-}
expect_uvc_fps=${5:-}

read_env_value() {
  local key=$1
  local file=$2
  local value=""
  value=$(grep -E "^${key}=" "$file" 2>/dev/null | tail -n1 | cut -d= -f2- || true)
  printf '%s\n' "$value"
}

cam_output=$(read_env_value "LESAVKA_CAM_OUTPUT" /etc/lesavka/server.env)
server_uvc_codec=$(read_env_value "LESAVKA_UVC_CODEC" /etc/lesavka/server.env)
runtime_uvc_codec=$(read_env_value "LESAVKA_UVC_CODEC" /etc/lesavka/uvc.env)
server_uvc_width=$(read_env_value "LESAVKA_UVC_WIDTH" /etc/lesavka/server.env)
server_uvc_height=$(read_env_value "LESAVKA_UVC_HEIGHT" /etc/lesavka/server.env)
server_uvc_fps=$(read_env_value "LESAVKA_UVC_FPS" /etc/lesavka/server.env)
runtime_uvc_width=$(read_env_value "LESAVKA_UVC_WIDTH" /etc/lesavka/uvc.env)
runtime_uvc_height=$(read_env_value "LESAVKA_UVC_HEIGHT" /etc/lesavka/uvc.env)
runtime_uvc_fps=$(read_env_value "LESAVKA_UVC_FPS" /etc/lesavka/uvc.env)
effective_uvc_codec=${server_uvc_codec:-$runtime_uvc_codec}
effective_uvc_width=${server_uvc_width:-$runtime_uvc_width}
effective_uvc_height=${server_uvc_height:-$runtime_uvc_height}
effective_uvc_fps=${server_uvc_fps:-$runtime_uvc_fps}

printf '   ↪ server.env CAM_OUTPUT=%s\n' "${cam_output:-<unset>}"
printf '   ↪ server.env UVC_CODEC=%s\n' "${server_uvc_codec:-<unset>}"
printf '   ↪ uvc.env    UVC_CODEC=%s\n' "${runtime_uvc_codec:-<unset>}"
printf '   ↪ server.env UVC_MODE=%sx%s@%s\n' "${server_uvc_width:-<unset>}" "${server_uvc_height:-<unset>}" "${server_uvc_fps:-<unset>}"
printf '   ↪ uvc.env    UVC_MODE=%sx%s@%s\n' "${runtime_uvc_width:-<unset>}" "${runtime_uvc_height:-<unset>}" "${runtime_uvc_fps:-<unset>}"
printf '   ↪ effective UVC_MODE=%sx%s@%s\n' "${effective_uvc_width:-<unset>}" "${effective_uvc_height:-<unset>}" "${effective_uvc_fps:-<unset>}"

if [[ -n "${expect_cam_output}" && "${cam_output}" != "${expect_cam_output}" ]]; then
  printf 'expected CAM_OUTPUT=%s but found %s\n' "${expect_cam_output}" "${cam_output:-<unset>}" >&2
  exit 64
fi
if [[ -n "${expect_uvc_codec}" && "${effective_uvc_codec}" != "${expect_uvc_codec}" ]]; then
  printf 'expected effective UVC_CODEC=%s but found %s\n' "${expect_uvc_codec}" "${effective_uvc_codec:-<unset>}" >&2
  exit 65
fi
if [[ -n "${expect_uvc_codec}" && "${runtime_uvc_codec}" != "${expect_uvc_codec}" ]]; then
  printf 'expected uvc.env UVC_CODEC=%s but found %s\n' "${expect_uvc_codec}" "${runtime_uvc_codec:-<unset>}" >&2
  exit 66
fi
if [[ -n "${expect_uvc_width}" && "${effective_uvc_width}" != "${expect_uvc_width}" ]]; then
  printf 'expected effective UVC_WIDTH=%s but found %s\n' "${expect_uvc_width}" "${effective_uvc_width:-<unset>}" >&2
  exit 67
fi
if [[ -n "${expect_uvc_height}" && "${effective_uvc_height}" != "${expect_uvc_height}" ]]; then
  printf 'expected effective UVC_HEIGHT=%s but found %s\n' "${expect_uvc_height}" "${effective_uvc_height:-<unset>}" >&2
  exit 68
fi
if [[ -n "${expect_uvc_fps}" && "${effective_uvc_fps}" != "${expect_uvc_fps}" ]]; then
  printf 'expected effective UVC_FPS=%s but found %s\n' "${expect_uvc_fps}" "${effective_uvc_fps:-<unset>}" >&2
  exit 69
fi
if [[ -n "${expect_uvc_width}" && "${runtime_uvc_width}" != "${expect_uvc_width}" ]]; then
  printf 'expected uvc.env UVC_WIDTH=%s but found %s\n' "${expect_uvc_width}" "${runtime_uvc_width:-<unset>}" >&2
  exit 70
fi
if [[ -n "${expect_uvc_height}" && "${runtime_uvc_height}" != "${expect_uvc_height}" ]]; then
  printf 'expected uvc.env UVC_HEIGHT=%s but found %s\n' "${expect_uvc_height}" "${runtime_uvc_height:-<unset>}" >&2
  exit 71
fi
if [[ -n "${expect_uvc_fps}" && "${runtime_uvc_fps}" != "${expect_uvc_fps}" ]]; then
  printf 'expected uvc.env UVC_FPS=%s but found %s\n' "${expect_uvc_fps}" "${runtime_uvc_fps:-<unset>}" >&2
  exit 72
fi

systemctl is-active lesavka-server lesavka-uvc lesavka-core >/dev/null
REMOTE_PREFLIGHT
}

print_lesavka_versions() {
  echo "==> Lesavka versions under test"
  if [[ ! -x "${REPO_ROOT}/target/debug/lesavka-relayctl" ]]; then
    (cd "${REPO_ROOT}" && cargo build -p lesavka_client --bin lesavka-relayctl >/dev/null)
  fi
  local version_output
  if ! version_output="$(
    LESAVKA_TLS_DOMAIN="${LESAVKA_TLS_DOMAIN}" \
      "${REPO_ROOT}/target/debug/lesavka-relayctl" \
      --server "${RESOLVED_LESAVKA_SERVER_ADDR}" \
      version 2>&1
  )"; then
    echo "failed to query Lesavka versions through ${RESOLVED_LESAVKA_SERVER_ADDR}" >&2
    echo "${version_output}" >&2
    return 1
  fi
  if ! grep -q "^client_version=" <<<"${version_output}"; then
    echo "Lesavka version query did not report client_version=; refusing to run an unattributed probe" >&2
    echo "${version_output}" >&2
    return 1
  fi
  if grep -q "^client_full_version=" <<<"${version_output}"; then
    echo "Lesavka version query reported a combined version+revision; refusing ambiguous probe attribution" >&2
    echo "${version_output}" >&2
    return 1
  fi
  if ! grep -q "^client_revision=" <<<"${version_output}"; then
    echo "Lesavka version query did not report client_revision=; refusing to run an unattributed probe" >&2
    echo "${version_output}" >&2
    return 1
  fi
  if ! grep -q "^server_version=" <<<"${version_output}"; then
    echo "Lesavka version query did not report server_version=; refusing to run an unattributed probe" >&2
    echo "${version_output}" >&2
    return 1
  fi
  if ! grep -q "^server_revision=" <<<"${version_output}"; then
    echo "Lesavka version query did not report server_revision=; refusing to run an unattributed probe" >&2
    echo "${version_output}" >&2
    return 1
  fi
  while IFS= read -r line; do
    [[ -n "${line}" ]] && echo "   ↪ ${line}"
  done <<<"${version_output}"
}

write_output_delay_calibration() {
  [[ "${LESAVKA_OUTPUT_DELAY_CALIBRATION}" != "0" ]] || return 0
  [[ -f "${LOCAL_ANALYSIS_JSON}" ]] || return 0

  echo "==> deriving UVC/UAC output-delay calibration"
  python3 - <<'PY' \
    "${LOCAL_ANALYSIS_JSON}" \
    "${LOCAL_OUTPUT_DELAY_JSON}" \
    "${LOCAL_OUTPUT_DELAY_ENV}" \
    "${LOCAL_OUTPUT_DELAY_CORRELATION_JSON}" \
    "${LESAVKA_OUTPUT_DELAY_TARGET}" \
    "${LESAVKA_OUTPUT_DELAY_MIN_PAIRS}" \
    "${LESAVKA_OUTPUT_DELAY_MAX_ABS_SKEW_MS}" \
    "${LESAVKA_OUTPUT_DELAY_MAX_DRIFT_MS}" \
    "${LESAVKA_OUTPUT_DELAY_GAIN}" \
    "${LESAVKA_OUTPUT_DELAY_MAX_STEP_US}" \
    "${LESAVKA_OUTPUT_DELAY_APPLY}" \
    "${LESAVKA_OUTPUT_DELAY_APPLY_MODE}" \
    "${LESAVKA_OUTPUT_DELAY_SAVE}" \
    "${LESAVKA_OUTPUT_DELAY_PROBE_AUDIO_DELAY_US}" \
    "${LESAVKA_OUTPUT_DELAY_PROBE_VIDEO_DELAY_US}"
import json
import math
import pathlib
import shlex
import sys

(
    report_path,
    output_json_path,
    output_env_path,
    correlation_path,
    target,
    min_pairs_raw,
    max_abs_skew_raw,
    max_drift_raw,
    gain_raw,
    max_step_raw,
    apply_raw,
    apply_mode_raw,
    save_raw,
    active_audio_raw,
    active_video_raw,
) = sys.argv[1:]


def as_int(value, default):
    try:
        return int(str(value).strip())
    except Exception:
        return default


def as_float(value, default):
    try:
        result = float(str(value).strip())
    except Exception:
        return default
    return result if math.isfinite(result) else default


def as_bool(value):
    return str(value).strip().lower() not in {"", "0", "false", "no", "off"}


def env_line(key, value):
    return f"{key}={shlex.quote(str(value))}\n"


report = json.loads(pathlib.Path(report_path).read_text())
verdict = report.get("verdict") or {}
try:
    correlation = json.loads(pathlib.Path(correlation_path).read_text())
except Exception:
    correlation = {}
freshness = correlation.get("freshness") or {}
capture_timebase = freshness.get("capture_timebase") or {}

target = target.strip().lower()
apply_mode = apply_mode_raw.strip().lower()
min_pairs = max(1, as_int(min_pairs_raw, 13))
max_abs_skew_ms = max(1.0, as_float(max_abs_skew_raw, 5000.0))
max_drift_ms = max(0.0, as_float(max_drift_raw, 80.0))
gain = min(max(as_float(gain_raw, 1.0), 0.01), 1.0)
max_step_us = max(1, as_int(max_step_raw, 1_500_000))
active_audio_offset_us = as_int(active_audio_raw, 0)
active_video_offset_us = as_int(active_video_raw, 0)

paired = as_int(report.get("paired_event_count"), 0)
median_skew_ms = as_float(report.get("median_skew_ms"), 0.0)
p95_abs_skew_ms = as_float(
    verdict.get("p95_abs_skew_ms"),
    as_float(report.get("max_abs_skew_ms"), 0.0),
)
max_abs_observed_ms = as_float(report.get("max_abs_skew_ms"), p95_abs_skew_ms)
drift_ms = as_float(report.get("drift_ms"), 0.0)

raw_device_delta_us = int(round(median_skew_ms * 1000.0))
scaled_delta_us = int(round(raw_device_delta_us * gain))
bounded_delta_us = max(-max_step_us, min(max_step_us, scaled_delta_us))

audio_delta_us = 0
video_delta_us = 0
refusal_reasons = []

if target == "video":
    video_delta_us = bounded_delta_us
elif target == "audio":
    audio_delta_us = -bounded_delta_us
else:
    refusal_reasons.append(f"unsupported target {target!r}; use video or audio")

if apply_mode not in {"absolute", "relative"}:
    refusal_reasons.append(
        f"unsupported apply mode {apply_mode!r}; use absolute or relative"
    )

if paired < min_pairs:
    refusal_reasons.append(f"paired_event_count {paired} < {min_pairs}")
if max_abs_observed_ms > max_abs_skew_ms:
    refusal_reasons.append(
        f"max_abs_skew_ms {max_abs_observed_ms:.1f} > {max_abs_skew_ms:.1f}"
    )
if abs(drift_ms) > max_drift_ms:
    refusal_reasons.append(f"abs(drift_ms) {abs(drift_ms):.1f} > {max_drift_ms:.1f}")
if capture_timebase and capture_timebase.get("valid") is not True:
    refusal_reasons.append(
        "capture timebase invalid for calibration: "
        f"{capture_timebase.get('status', 'unknown')} - {capture_timebase.get('reason', '')}"
    )
elif freshness.get("status") == "invalid":
    refusal_reasons.append(
        f"freshness invalid for calibration: {freshness.get('reason', '')}"
    )

audio_target_offset_us = active_audio_offset_us + audio_delta_us
video_target_offset_us = active_video_offset_us + video_delta_us
ready = not refusal_reasons
decision = "ready" if ready else "refused"
note = (
    "direct UVC/UAC output-delay calibration: "
    f"median device skew {median_skew_ms:+.1f}ms, target={target}, "
    f"audio {active_audio_offset_us:+d}->{audio_target_offset_us:+d}us "
    f"(delta {audio_delta_us:+d}us), "
    f"video {active_video_offset_us:+d}->{video_target_offset_us:+d}us "
    f"(delta {video_delta_us:+d}us)"
)
if not ready:
    note = f"direct UVC/UAC output-delay calibration refused: {'; '.join(refusal_reasons)}"

artifact = {
    "schema": "lesavka.output-delay-calibration.v1",
    "source": "direct-uvc-uac-output-probe",
    "scope": "server-output-static-baseline",
    "applies_to": "server UVC/UAC gadget output path",
    "measurement_host_role": "lab-attached USB host",
    "probe_media_origin": "server-generated",
    "probe_media_path": "server generated signatures -> UVC/UAC sinks -> lab host capture",
    "injection_scope": "server-final-output-handoff",
    "server_pipeline_reference": "generated media PTS before intentional sync delay",
    "sink_handoff_path": "video CameraRelay::feed; audio Voice::push",
    "client_uplink_included": False,
    "report_json": report_path,
    "correlation_json": correlation_path if pathlib.Path(correlation_path).exists() else "",
    "audio_after_video_positive": True,
    "target": target,
    "ready": ready,
    "decision": decision,
    "apply_enabled": as_bool(apply_raw),
    "apply_mode": apply_mode,
    "save_enabled": as_bool(save_raw),
    "paired_event_count": paired,
    "min_pairs": min_pairs,
    "measured_device_skew_ms": median_skew_ms,
    "p95_abs_skew_ms": p95_abs_skew_ms,
    "max_abs_skew_ms": max_abs_observed_ms,
    "max_abs_skew_limit_ms": max_abs_skew_ms,
    "drift_ms": drift_ms,
    "max_drift_ms": max_drift_ms,
    "gain": gain,
    "max_step_us": max_step_us,
    "active_audio_offset_us": active_audio_offset_us,
    "active_video_offset_us": active_video_offset_us,
    "raw_device_delta_us": raw_device_delta_us,
    "bounded_device_delta_us": bounded_delta_us,
    "audio_offset_adjust_us": audio_delta_us,
    "video_offset_adjust_us": video_delta_us,
    "audio_target_offset_us": audio_target_offset_us,
    "video_target_offset_us": video_target_offset_us,
    "refusal_reasons": refusal_reasons,
    "note": note,
}

pathlib.Path(output_json_path).write_text(json.dumps(artifact, indent=2, sort_keys=True) + "\n")
env_values = {
    "output_delay_ready": str(ready).lower(),
    "output_delay_decision": decision,
    "output_delay_target": target,
    "output_delay_audio_delta_us": audio_delta_us,
    "output_delay_video_delta_us": video_delta_us,
    "output_delay_active_audio_offset_us": active_audio_offset_us,
    "output_delay_active_video_offset_us": active_video_offset_us,
    "output_delay_audio_target_offset_us": audio_target_offset_us,
    "output_delay_video_target_offset_us": video_target_offset_us,
    "output_delay_measured_skew_ms": f"{median_skew_ms:.3f}",
    "output_delay_paired_event_count": paired,
    "output_delay_drift_ms": f"{drift_ms:.3f}",
    "output_delay_apply_mode": apply_mode,
    "output_delay_note": note,
}
with pathlib.Path(output_env_path).open("w") as handle:
    for key, value in env_values.items():
        handle.write(env_line(key, value))
PY
}

extract_timeline_from_reply() {
  local reply_path=$1
  local timeline_path=$2
  [[ -f "${reply_path}" ]] || return 0

  python3 - <<'PY' "${reply_path}" "${timeline_path}"
import json
import pathlib
import sys

reply_path = pathlib.Path(sys.argv[1])
timeline_path = pathlib.Path(sys.argv[2])
prefix = "server_timeline_json="
raw = ""
for line in reply_path.read_text(errors="replace").splitlines():
    if line.startswith(prefix):
        raw = line[len(prefix):].strip()
if not raw:
    raise SystemExit(0)
timeline = json.loads(raw)
timeline_path.write_text(json.dumps(timeline, indent=2, sort_keys=True) + "\n")
PY
}

extract_server_timeline() {
  extract_timeline_from_reply "${LOCAL_SERVER_PROBE_REPLY}" "${LOCAL_SERVER_TIMELINE_JSON}"
}

extract_signal_conditioning_timeline() {
  extract_timeline_from_reply "${LOCAL_SIGNAL_CONDITIONING_REPLY}" "${LOCAL_SIGNAL_CONDITIONING_TIMELINE_JSON}"
}

compute_analysis_window_arg() {
  [[ "${ANALYSIS_TIMELINE_WINDOW}" != "0" ]] || return 0
  [[ -f "${LOCAL_SERVER_TIMELINE_JSON}" ]] || return 0
  [[ -f "${LOCAL_CAPTURE_LOG}" ]] || return 0
  [[ -f "${LOCAL_CLOCK_ALIGNMENT_JSON}" ]] || return 0

  python3 - <<'PY' \
    "${LOCAL_SERVER_TIMELINE_JSON}" \
    "${LOCAL_CAPTURE_LOG}" \
    "${LOCAL_CLOCK_ALIGNMENT_JSON}" \
    "${ANALYSIS_TIMELINE_WINDOW_PADDING_SECONDS}" || true
import json
import math
import pathlib
import sys

timeline_path, capture_log_path, clock_path, padding_raw = sys.argv[1:]


def as_int(value):
    try:
        return int(str(value).strip())
    except Exception:
        return None


def as_float(value, default):
    try:
        parsed = float(str(value).strip())
    except Exception:
        return default
    return parsed if math.isfinite(parsed) else default


def capture_start_ns(path):
    try:
        lines = pathlib.Path(path).read_text(errors="replace").splitlines()
    except Exception:
        return None
    for line in lines:
        if line.startswith("capture_start_unix_ns="):
            return as_int(line.split("=", 1)[1])
    return None


try:
    timeline = json.loads(pathlib.Path(timeline_path).read_text())
    clock = json.loads(pathlib.Path(clock_path).read_text())
except Exception:
    raise SystemExit(0)

capture_start = capture_start_ns(capture_log_path)
offset = as_int(clock.get("theia_to_tethys_offset_ns")) if clock.get("available") else None
if capture_start is None or offset is None:
    raise SystemExit(0)

times = []
for event in timeline.get("events", []):
    for key in ("audio_push_unix_ns", "video_feed_unix_ns"):
        event_ns = as_int(event.get(key))
        if event_ns is not None:
            times.append((event_ns + offset - capture_start) / 1_000_000_000.0)

if not times:
    raise SystemExit(0)

padding = max(0.0, as_float(padding_raw, 1.0))
start = max(0.0, min(times) - padding)
end = max(times) + padding
if end <= start:
    raise SystemExit(0)
print(f"--analysis-window-s {start:.3f}:{end:.3f}")
PY
}

write_output_delay_correlation() {
  [[ -f "${LOCAL_ANALYSIS_JSON}" ]] || return 0
  [[ -f "${LOCAL_SERVER_TIMELINE_JSON}" ]] || return 0

  echo "==> correlating Theia feed timing with Tethys observations"
  python3 - <<'PY' \
    "${LOCAL_ANALYSIS_JSON}" \
    "${LOCAL_SERVER_TIMELINE_JSON}" \
    "${LOCAL_OUTPUT_DELAY_CORRELATION_JSON}" \
    "${LOCAL_OUTPUT_DELAY_CORRELATION_CSV}" \
    "${LOCAL_OUTPUT_DELAY_CORRELATION_TXT}" \
    "${LOCAL_CAPTURE}" \
    "${LOCAL_CAPTURE_LOG}" \
    "${LOCAL_CLOCK_ALIGNMENT_JSON}" \
    "${LESAVKA_OUTPUT_FRESHNESS_MAX_AGE_MS}" \
    "${LESAVKA_OUTPUT_FRESHNESS_MAX_DRIFT_MS}" \
    "${LESAVKA_OUTPUT_FRESHNESS_MAX_CLOCK_UNCERTAINTY_MS}" \
    "${LESAVKA_OUTPUT_FRESHNESS_MIN_PAIRS}" \
    "${CAPTURE_SECONDS}" \
    "${CAPTURE_START_GRACE_SECONDS}"
import csv
import json
import math
import pathlib
import statistics
import struct
import subprocess
import sys

(
    report_path,
    timeline_path,
    output_json_path,
    output_csv_path,
    output_txt_path,
    capture_path,
    capture_log_path,
    clock_alignment_path,
    max_freshness_age_raw,
    max_freshness_drift_raw,
    max_clock_uncertainty_raw,
    min_freshness_pairs_raw,
    capture_seconds_raw,
    capture_start_grace_seconds_raw,
) = sys.argv[1:]
report = json.loads(pathlib.Path(report_path).read_text())
timeline = json.loads(pathlib.Path(timeline_path).read_text())
server_event_list = timeline.get("events", [])
server_events = {int(event["event_id"]): event for event in server_event_list}


def finite(value):
    try:
        result = float(value)
    except Exception:
        return None
    return result if math.isfinite(result) else None


def as_float(value, default):
    try:
        result = float(str(value).strip())
    except Exception:
        return default
    return result if math.isfinite(result) else default


def as_int_or_none(value):
    try:
        return int(str(value).strip())
    except Exception:
        return None


server_events_by_code = {}
for event in server_event_list:
    code = as_int_or_none(event.get("code"))
    if code is None:
        continue
    server_events_by_code.setdefault(code, []).append(event)
unique_server_events_by_code = {
    code: events[0] for code, events in server_events_by_code.items() if len(events) == 1
}


def load_json_or_empty(path):
    try:
        return json.loads(pathlib.Path(path).read_text())
    except Exception:
        return {}


def run_json(command, description):
    try:
        output = subprocess.check_output(command, stderr=subprocess.PIPE)
    except Exception as error:
        return {"available": False, "error": f"{description} failed: {error}"}
    try:
        return json.loads(output)
    except Exception as error:
        return {"available": False, "error": f"{description} JSON parse failed: {error}"}


def run_bytes(command, description):
    try:
        return subprocess.check_output(command, stderr=subprocess.PIPE)
    except Exception as error:
        raise RuntimeError(f"{description} failed: {error}") from error


def parse_capture_start_unix_ns(path):
    try:
        lines = pathlib.Path(path).read_text(errors="replace").splitlines()
    except Exception:
        return None
    for line in lines:
        if line.startswith("capture_start_unix_ns="):
            return as_int_or_none(line.split("=", 1)[1])
    return None


def percentile(values, pct):
    values = sorted(value for value in values if value is not None and math.isfinite(value))
    if not values:
        return None
    if len(values) == 1:
        return values[0]
    rank = (len(values) - 1) * (pct / 100.0)
    lower = math.floor(rank)
    upper = math.ceil(rank)
    if lower == upper:
        return values[int(rank)]
    fraction = rank - lower
    return values[lower] * (1.0 - fraction) + values[upper] * fraction


def numeric_stats(values, suffix=""):
    values = [value for value in values if value is not None and math.isfinite(value)]
    if not values:
        return {
            "available": False,
            "count": 0,
            f"median{suffix}": None,
            f"p95{suffix}": None,
            f"max{suffix}": None,
        }
    return {
        "available": True,
        "count": len(values),
        f"median{suffix}": percentile(values, 50.0),
        f"p95{suffix}": percentile(values, 95.0),
        f"max{suffix}": max(values),
    }


def stats(rows, key):
    values = [row.get(key) for row in rows if row.get(key) is not None]
    values = [value for value in values if math.isfinite(value)]
    if not values:
        return {
            "available": False,
            "count": 0,
            "first_ms": None,
            "last_ms": None,
            "mean_ms": None,
            "min_ms": None,
            "median_ms": None,
            "p95_ms": None,
            "max_ms": None,
        }
    return {
        "available": True,
        "count": len(values),
        "first_ms": values[0],
        "last_ms": values[-1],
        "mean_ms": sum(values) / len(values),
        "min_ms": min(values),
        "median_ms": percentile(values, 50.0),
        "p95_ms": percentile(values, 95.0),
        "max_ms": max(values),
    }


def shift_stats(base, delta_ms):
    shifted = dict(base)
    if not shifted.get("available"):
        return shifted
    for key in ["first_ms", "last_ms", "mean_ms", "min_ms", "median_ms", "p95_ms", "max_ms"]:
        value = shifted.get(key)
        shifted[key] = value + delta_ms if value is not None else None
    shifted["intentional_delay_ms"] = delta_ms
    return shifted


def fit_linear(rows, key):
    points = [(row["event_time_s"], row[key]) for row in rows if row.get(key) is not None]
    if len(points) < 2:
        return {
            "available": False,
            "intercept_ms": 0.0,
            "slope_ms_per_s": 0.0,
            "r2": 0.0,
            "drift_ms": 0.0,
        }
    xs = [point[0] for point in points]
    ys = [point[1] for point in points]
    mean_x = sum(xs) / len(xs)
    mean_y = sum(ys) / len(ys)
    denom = sum((x - mean_x) ** 2 for x in xs)
    slope = 0.0 if denom == 0 else sum((x - mean_x) * (y - mean_y) for x, y in points) / denom
    intercept = mean_y - slope * mean_x
    predicted = [intercept + slope * x for x in xs]
    ss_tot = sum((y - mean_y) ** 2 for y in ys)
    ss_res = sum((y - y_hat) ** 2 for y, y_hat in zip(ys, predicted))
    r2 = 1.0 if ss_tot == 0.0 else max(0.0, min(1.0, 1.0 - (ss_res / ss_tot)))
    drift = (intercept + slope * xs[-1]) - (intercept + slope * xs[0])
    return {
        "available": True,
        "intercept_ms": intercept,
        "slope_ms_per_s": slope,
        "r2": r2,
        "drift_ms": drift,
        "first_fit_ms": intercept + slope * xs[0],
        "last_fit_ms": intercept + slope * xs[-1],
    }


def correlation(rows, left_key, right_key):
    pairs = [
        (row[left_key], row[right_key])
        for row in rows
        if row.get(left_key) is not None and row.get(right_key) is not None
    ]
    if len(pairs) < 2:
        return 0.0
    xs = [pair[0] for pair in pairs]
    ys = [pair[1] for pair in pairs]
    mean_x = sum(xs) / len(xs)
    mean_y = sum(ys) / len(ys)
    denom_x = sum((x - mean_x) ** 2 for x in xs)
    denom_y = sum((y - mean_y) ** 2 for y in ys)
    if denom_x <= 0.0 or denom_y <= 0.0:
        return 0.0
    return sum((x - mean_x) * (y - mean_y) for x, y in pairs) / math.sqrt(denom_x * denom_y)


def ffprobe_frame_timestamps(path):
    data = run_json(
        [
            "ffprobe",
            "-hide_banner",
            "-loglevel",
            "error",
            "-select_streams",
            "v:0",
            "-show_frames",
            "-show_entries",
            "frame=best_effort_timestamp_time",
            "-of",
            "json",
            path,
        ],
        "ffprobe video frames",
    )
    return [
        float(frame["best_effort_timestamp_time"])
        for frame in data.get("frames", [])
        if frame.get("best_effort_timestamp_time") not in (None, "N/A")
    ]


def ffprobe_packet_times(path, stream):
    data = run_json(
        [
            "ffprobe",
            "-hide_banner",
            "-loglevel",
            "error",
            "-select_streams",
            stream,
            "-show_packets",
            "-show_entries",
            "packet=pts_time,duration_time",
            "-of",
            "json",
            path,
        ],
        f"ffprobe {stream} packets",
    )
    packets = []
    for packet in data.get("packets", []):
        try:
            pts = float(packet["pts_time"])
        except Exception:
            continue
        duration = None
        try:
            duration = float(packet.get("duration_time"))
        except Exception:
            pass
        packets.append((pts, duration))
    return packets


def continuity_frame_ids(path):
    width = 320
    height = 32
    blocks = 20
    try:
        raw = run_bytes(
            [
                "ffmpeg",
                "-hide_banner",
                "-loglevel",
                "error",
                "-i",
                path,
                "-map",
                "0:v:0",
                "-vf",
                f"crop=iw:32:0:ih-32,scale={width}:{height}:flags=area,format=gray",
                "-f",
                "rawvideo",
                "-pix_fmt",
                "gray",
                "-",
            ],
            "ffmpeg video continuity extraction",
        )
    except RuntimeError:
        return []
    frame_pixels = width * height
    if not raw or len(raw) % frame_pixels != 0:
        return []
    block_width = width // blocks
    ids = []
    for offset in range(0, len(raw), frame_pixels):
        frame = raw[offset : offset + frame_pixels]
        averages = []
        for block in range(blocks):
            x_start = block * block_width
            x_end = width if block + 1 == blocks else (block + 1) * block_width
            total = 0
            count = 0
            for y in range(height):
                row = y * width
                for x in range(x_start, x_end):
                    total += frame[row + x]
                    count += 1
            averages.append(total / max(1, count))
        white = averages[0]
        black = averages[1]
        if white < 150 or black > 105 or white - black < 70:
            ids.append(None)
            continue
        threshold = (white + black) / 2.0
        value = 0
        for block in range(2, 18):
            value = (value << 1) | int(averages[block] > threshold)
        parity = bool(averages[18] > threshold)
        inverse = bool(averages[19] > threshold)
        if parity == inverse or parity != bool(value.bit_count() & 1):
            ids.append(None)
            continue
        ids.append(value)
    return ids


def sequence_smoothness(ids):
    decoded = [value for value in ids if value is not None]
    duplicates = 0
    missing = 0
    regressions = 0
    jumps = []
    previous = None
    for value in decoded:
        if previous is None:
            previous = value
            continue
        diff = (value - previous) & 0xFFFF
        jumps.append(diff)
        if diff == 0:
            duplicates += 1
        elif 1 < diff < 32768:
            missing += diff - 1
        elif diff >= 32768:
            regressions += 1
        previous = value
    return {
        "decoded_frames": len(decoded),
        "undecodable_frames": len(ids) - len(decoded),
        "unique_frames": len(set(decoded)),
        "duplicate_frames": duplicates,
        "estimated_missing_frames": missing,
        "sequence_regressions": regressions,
        "largest_forward_jump": max(jumps) if jumps else 0,
    }


def interval_smoothness(timestamps, expected_ms, window_start_s=None, window_end_s=None):
    if window_start_s is not None and window_end_s is not None:
        timestamps = [ts for ts in timestamps if window_start_s <= ts <= window_end_s]
    intervals = [
        (right - left) * 1000.0
        for left, right in zip(timestamps, timestamps[1:])
        if right > left
    ]
    if (expected_ms is None or expected_ms <= 0.0) and intervals:
        expected_ms = statistics.median(intervals)
    jitter = [abs(value - expected_ms) for value in intervals] if expected_ms else []
    hiccup_threshold = expected_ms * 1.5 if expected_ms else float("inf")
    short_threshold = expected_ms * 0.5 if expected_ms else -1.0
    return {
        "timestamps": len(timestamps),
        "expected_interval_ms": expected_ms,
        "interval_stats": numeric_stats(intervals, "_interval_ms"),
        "jitter_stats": numeric_stats(jitter, "_jitter_ms"),
        "hiccup_count": sum(1 for value in intervals if value > hiccup_threshold),
        "short_interval_count": sum(1 for value in intervals if value < short_threshold),
        "estimated_missing_by_timestamp": sum(
            max(0, round(value / expected_ms) - 1)
            for value in intervals
            if expected_ms and value > hiccup_threshold
        ),
    }


def audio_rms_smoothness(path, window_start_s, window_end_s):
    sample_rate = 48_000
    window_ms = 10
    samples_per_window = sample_rate * window_ms // 1000
    try:
        raw = run_bytes(
            [
                "ffmpeg",
                "-hide_banner",
                "-loglevel",
                "error",
                "-i",
                path,
                "-map",
                "0:a:0",
                "-ac",
                "1",
                "-ar",
                str(sample_rate),
                "-f",
                "s16le",
                "-acodec",
                "pcm_s16le",
                "-",
            ],
            "ffmpeg audio continuity extraction",
        )
    except RuntimeError as error:
        return {"available": False, "error": str(error)}
    start_sample = max(0, int(window_start_s * sample_rate))
    end_sample = min(len(raw) // 2, int(window_end_s * sample_rate))
    if end_sample <= start_sample:
        return {"available": False, "error": "empty audio smoothness window"}
    sample_count = end_sample - start_sample
    samples = struct.unpack(
        f"<{sample_count}h",
        raw[start_sample * 2 : end_sample * 2],
    )
    rms_values = []
    for index in range(0, len(samples) - samples_per_window + 1, samples_per_window):
        chunk = samples[index : index + samples_per_window]
        rms_values.append(math.sqrt(sum(value * value for value in chunk) / len(chunk)))
    low_threshold = 90.0
    return {
        "available": bool(rms_values),
        "window_ms": window_ms,
        "rms_stats": numeric_stats(rms_values, "_rms"),
        "low_rms_window_count": sum(1 for value in rms_values if value < low_threshold),
        "low_rms_threshold": low_threshold,
    }


def analyze_smoothness(path, report, timeline):
    camera_fps = as_float(timeline.get("camera_fps"), 0.0)
    duration_s = as_float(timeline.get("duration_us"), 0.0) / 1_000_000.0
    warmup_s = as_float(timeline.get("warmup_us"), 0.0) / 1_000_000.0
    paired_times = [
        value
        for event in report.get("paired_events", [])
        for value in [finite(event.get("video_time_s")), finite(event.get("audio_time_s"))]
        if value is not None
    ]
    if paired_times:
        window_start_s = max(0.0, min(paired_times) - warmup_s)
        window_end_s = window_start_s + duration_s
    else:
        window_start_s = None
        window_end_s = None
    video_timestamps = ffprobe_frame_timestamps(path)
    audio_packets = ffprobe_packet_times(path, "a:0")
    audio_packet_timestamps = [pts for pts, _duration in audio_packets]
    media_timeline = {
        "video_first_s": min(video_timestamps) if video_timestamps else None,
        "video_last_s": max(video_timestamps) if video_timestamps else None,
        "audio_first_s": min(audio_packet_timestamps) if audio_packet_timestamps else None,
        "audio_last_s": max(audio_packet_timestamps) if audio_packet_timestamps else None,
    }
    media_timeline["video_span_s"] = (
        media_timeline["video_last_s"] - media_timeline["video_first_s"]
        if media_timeline["video_first_s"] is not None and media_timeline["video_last_s"] is not None
        else None
    )
    media_timeline["audio_span_s"] = (
        media_timeline["audio_last_s"] - media_timeline["audio_first_s"]
        if media_timeline["audio_first_s"] is not None and media_timeline["audio_last_s"] is not None
        else None
    )
    media_timeline["audio_video_span_gap_s"] = (
        media_timeline["video_span_s"] - media_timeline["audio_span_s"]
        if media_timeline["video_span_s"] is not None and media_timeline["audio_span_s"] is not None
        else None
    )
    video_ids = continuity_frame_ids(path)
    if window_start_s is not None and window_end_s is not None and len(video_ids) == len(video_timestamps):
        video_ids = [
            frame_id
            for frame_id, timestamp in zip(video_ids, video_timestamps)
            if window_start_s <= timestamp <= window_end_s
        ]
    expected_video_ms = 1000.0 / camera_fps if camera_fps > 0.0 else None
    audio_durations = [
        duration * 1000.0
        for _pts, duration in audio_packets
        if duration is not None and math.isfinite(duration) and duration > 0.0
    ]
    expected_audio_ms = statistics.median(audio_durations) if audio_durations else None
    audio_window_end = window_end_s if window_end_s is not None else max(audio_packet_timestamps or [0.0])
    return {
        "schema": "lesavka.output-smoothness-summary.v1",
        "scope": "captured RC target media cadence and continuity over the server-generated probe window",
        "window_start_s": window_start_s,
        "window_end_s": window_end_s,
        "media_timeline": media_timeline,
        "video": {
            **interval_smoothness(video_timestamps, expected_video_ms, window_start_s, window_end_s),
            **sequence_smoothness(video_ids),
        },
        "audio": {
            "packet_cadence": interval_smoothness(
                audio_packet_timestamps,
                expected_audio_ms,
                window_start_s,
                window_end_s,
            ),
            "rms_continuity": audio_rms_smoothness(path, window_start_s or 0.0, audio_window_end),
        },
    }


capture_start_unix_ns = parse_capture_start_unix_ns(capture_log_path)
clock_alignment = load_json_or_empty(clock_alignment_path)
clock_alignment_available = bool(clock_alignment.get("available"))
theia_to_tethys_offset_ns = (
    as_int_or_none(clock_alignment.get("theia_to_tethys_offset_ns"))
    if clock_alignment_available
    else None
)
clock_uncertainty_ms = as_float(clock_alignment.get("uncertainty_ms"), 0.0)
audio_delay_ms = as_float(timeline.get("audio_delay_us"), 0.0) / 1000.0
video_delay_ms = as_float(timeline.get("video_delay_us"), 0.0) / 1000.0


def corrected_server_capture_time_s(server, key):
    server_unix_ns = as_int_or_none(server.get(key))
    if (
        server_unix_ns is None
        or theia_to_tethys_offset_ns is None
        or capture_start_unix_ns is None
    ):
        return None
    return (server_unix_ns + theia_to_tethys_offset_ns - capture_start_unix_ns) / 1_000_000_000.0


def corrected_capture_time_s_from_unix_ns(server_unix_ns):
    if (
        server_unix_ns is None
        or theia_to_tethys_offset_ns is None
        or capture_start_unix_ns is None
    ):
        return None
    return (server_unix_ns + theia_to_tethys_offset_ns - capture_start_unix_ns) / 1_000_000_000.0


def shifted_unix_ns(unix_ns, delta_ms):
    if unix_ns is None:
        return None
    return unix_ns + int(round(delta_ms * 1_000_000.0))


def nearest_server_event_for_observation(observed, used_event_ids):
    observed_video_s = finite(observed.get("video_time_s"))
    observed_audio_s = finite(observed.get("audio_time_s"))
    if observed_video_s is None and observed_audio_s is None:
        return None, None, None
    best = None
    for server in server_event_list:
        event_id = as_int_or_none(server.get("event_id"))
        if event_id in used_event_ids:
            continue
        expected_video_s = corrected_server_capture_time_s(server, "video_feed_unix_ns")
        expected_audio_s = corrected_server_capture_time_s(server, "audio_push_unix_ns")
        score = 0.0
        parts = 0
        if observed_video_s is not None and expected_video_s is not None:
            score += abs(observed_video_s - expected_video_s)
            parts += 1
        if observed_audio_s is not None and expected_audio_s is not None:
            score += abs(observed_audio_s - expected_audio_s)
            parts += 1
        if parts == 0:
            continue
        normalized_score = score / parts
        if best is None or normalized_score < best[0]:
            best = (normalized_score, event_id, server)
    if best is None:
        return None, None, None
    return best[2], "nearest_clocked_time", best[0] * 1000.0


joined = []
used_server_event_ids = set()
for observed in report.get("paired_events", []):
    observed_event_id = as_int_or_none(observed.get("event_id"))
    observed_server_event_id = as_int_or_none(observed.get("server_event_id"))
    observed_event_code = as_int_or_none(observed.get("event_code"))
    server = None
    match_method = None
    match_score_ms = None
    if observed_server_event_id is not None:
        server = server_events.get(observed_server_event_id)
        match_method = "analyzer_server_event_id" if server else None
    if server is None and observed_event_code is not None:
        server = unique_server_events_by_code.get(observed_event_code)
        match_method = "unique_event_code" if server else None
    if server is None:
        server, match_method, match_score_ms = nearest_server_event_for_observation(
            observed,
            used_server_event_ids,
        )
    if server is None and observed_event_code is None and observed_event_id is not None:
        server = server_events.get(observed_event_id)
        match_method = "legacy_pair_event_id" if server else None
    if not server:
        continue
    event_id = int(server.get("event_id", -1))
    used_server_event_ids.add(event_id)
    observed_skew_ms = finite(observed.get("skew_ms"))
    server_feed_delta_ms = finite(server.get("server_feed_delta_ms"))
    tethys_video_time_s = finite(observed.get("video_time_s"))
    tethys_audio_time_s = finite(observed.get("audio_time_s"))
    server_video_feed_s = finite(server.get("video_feed_monotonic_us"))
    server_audio_push_s = finite(server.get("audio_push_monotonic_us"))
    if server_video_feed_s is not None:
        server_video_feed_s /= 1_000_000.0
    if server_audio_push_s is not None:
        server_audio_push_s /= 1_000_000.0
    server_video_feed_unix_ns = as_int_or_none(server.get("video_feed_unix_ns"))
    server_audio_push_unix_ns = as_int_or_none(server.get("audio_push_unix_ns"))
    server_video_pipeline_unix_ns = shifted_unix_ns(
        server_video_feed_unix_ns,
        -video_delay_ms,
    )
    server_audio_pipeline_unix_ns = shifted_unix_ns(
        server_audio_push_unix_ns,
        -audio_delay_ms,
    )
    tethys_video_unix_ns = (
        capture_start_unix_ns + int(round(tethys_video_time_s * 1_000_000_000.0))
        if capture_start_unix_ns is not None and tethys_video_time_s is not None
        else None
    )
    tethys_audio_unix_ns = (
        capture_start_unix_ns + int(round(tethys_audio_time_s * 1_000_000_000.0))
        if capture_start_unix_ns is not None and tethys_audio_time_s is not None
        else None
    )
    corrected_server_video_feed_unix_ns = (
        server_video_feed_unix_ns + theia_to_tethys_offset_ns
        if server_video_feed_unix_ns is not None and theia_to_tethys_offset_ns is not None
        else None
    )
    corrected_server_audio_push_unix_ns = (
        server_audio_push_unix_ns + theia_to_tethys_offset_ns
        if server_audio_push_unix_ns is not None and theia_to_tethys_offset_ns is not None
        else None
    )
    corrected_server_video_pipeline_unix_ns = (
        server_video_pipeline_unix_ns + theia_to_tethys_offset_ns
        if server_video_pipeline_unix_ns is not None and theia_to_tethys_offset_ns is not None
        else None
    )
    corrected_server_audio_pipeline_unix_ns = (
        server_audio_pipeline_unix_ns + theia_to_tethys_offset_ns
        if server_audio_pipeline_unix_ns is not None and theia_to_tethys_offset_ns is not None
        else None
    )
    video_sink_handoff_overhead_ms = (
        (tethys_video_unix_ns - corrected_server_video_feed_unix_ns) / 1_000_000.0
        if tethys_video_unix_ns is not None and corrected_server_video_feed_unix_ns is not None
        else None
    )
    audio_sink_handoff_overhead_ms = (
        (tethys_audio_unix_ns - corrected_server_audio_push_unix_ns) / 1_000_000.0
        if tethys_audio_unix_ns is not None and corrected_server_audio_push_unix_ns is not None
        else None
    )
    video_freshness_ms = (
        (tethys_video_unix_ns - corrected_server_video_pipeline_unix_ns) / 1_000_000.0
        if tethys_video_unix_ns is not None and corrected_server_video_pipeline_unix_ns is not None
        else None
    )
    audio_freshness_ms = (
        (tethys_audio_unix_ns - corrected_server_audio_pipeline_unix_ns) / 1_000_000.0
        if tethys_audio_unix_ns is not None and corrected_server_audio_pipeline_unix_ns is not None
        else None
    )
    video_event_age_ms = video_freshness_ms
    audio_event_age_ms = audio_freshness_ms
    residual_path_skew_ms = (
        observed_skew_ms - server_feed_delta_ms
        if observed_skew_ms is not None and server_feed_delta_ms is not None
        else None
    )
    planned_start_us = int(server.get("planned_start_us", 0))
    joined.append({
        "event_id": event_id,
        "observed_event_id": observed_event_id,
        "observed_server_event_id": observed_server_event_id,
        "observed_event_code": observed_event_code,
        "match_method": match_method,
        "match_score_ms": match_score_ms,
        "code": int(server.get("code", 0)),
        "event_time_s": planned_start_us / 1_000_000.0,
        "planned_start_us": planned_start_us,
        "planned_end_us": int(server.get("planned_end_us", 0)),
        "tethys_video_time_s": tethys_video_time_s,
        "tethys_audio_time_s": tethys_audio_time_s,
        "observed_skew_ms": observed_skew_ms,
        "server_video_feed_monotonic_us": server.get("video_feed_monotonic_us"),
        "server_audio_push_monotonic_us": server.get("audio_push_monotonic_us"),
        "server_video_feed_unix_ns": server_video_feed_unix_ns,
        "server_audio_push_unix_ns": server_audio_push_unix_ns,
        "server_video_pipeline_unix_ns": server_video_pipeline_unix_ns,
        "server_audio_pipeline_unix_ns": server_audio_pipeline_unix_ns,
        "tethys_video_unix_ns": tethys_video_unix_ns,
        "tethys_audio_unix_ns": tethys_audio_unix_ns,
        "server_video_feed_s": server_video_feed_s,
        "server_audio_push_s": server_audio_push_s,
        "video_freshness_ms": video_freshness_ms,
        "audio_freshness_ms": audio_freshness_ms,
        "video_sink_handoff_overhead_ms": video_sink_handoff_overhead_ms,
        "audio_sink_handoff_overhead_ms": audio_sink_handoff_overhead_ms,
        "video_event_age_ms": video_event_age_ms,
        "audio_event_age_ms": audio_event_age_ms,
        "server_feed_delta_ms": server_feed_delta_ms,
        "residual_path_skew_ms": residual_path_skew_ms,
        "confidence": finite(observed.get("confidence")),
    })

first_event_time_s = joined[0]["event_time_s"] if joined else 0.0
for row in joined:
    row["relative_event_time_s"] = row["event_time_s"] - first_event_time_s
    row["event_time_s"] = row["relative_event_time_s"]

observed_model = fit_linear(joined, "observed_skew_ms")
server_model = fit_linear(joined, "server_feed_delta_ms")
residual_model = fit_linear(joined, "residual_path_skew_ms")
video_freshness_model = fit_linear(joined, "video_freshness_ms")
audio_freshness_model = fit_linear(joined, "audio_freshness_ms")
video_sink_handoff_model = fit_linear(joined, "video_sink_handoff_overhead_ms")
audio_sink_handoff_model = fit_linear(joined, "audio_sink_handoff_overhead_ms")
video_event_age_model = fit_linear(joined, "video_event_age_ms")
audio_event_age_model = fit_linear(joined, "audio_event_age_ms")
video_freshness_stats = stats(joined, "video_freshness_ms")
audio_freshness_stats = stats(joined, "audio_freshness_ms")
video_sink_handoff_stats = stats(joined, "video_sink_handoff_overhead_ms")
audio_sink_handoff_stats = stats(joined, "audio_sink_handoff_overhead_ms")
video_event_age_stats = dict(video_freshness_stats)
audio_event_age_stats = dict(audio_freshness_stats)
video_event_age_stats["intentional_delay_ms"] = video_delay_ms
audio_event_age_stats["intentional_delay_ms"] = audio_delay_ms
server_observed_correlation = correlation(joined, "server_feed_delta_ms", "observed_skew_ms")
sync_verdict = report.get("verdict") or {}
sync_passed = sync_verdict.get("passed") is True

observed_drift = observed_model.get("drift_ms", 0.0)
server_drift = server_model.get("drift_ms", 0.0)
residual_drift = residual_model.get("drift_ms", 0.0)
same_direction = observed_drift == 0.0 or (observed_drift > 0) == (server_drift > 0)
server_share = 0.0 if abs(observed_drift) < 1e-6 else abs(server_drift) / abs(observed_drift)
if same_direction and server_share >= 0.5 and abs(server_drift) >= 20.0:
    dominant_layer = "server_feed_timing"
else:
    dominant_layer = "post_server_output_or_tethys_capture"

correction_mode = (
    "linear_function_candidate"
    if abs(residual_drift) >= 20.0
    else "scalar_candidate"
)

max_freshness_age_ms = max(1.0, as_float(max_freshness_age_raw, 1000.0))
max_freshness_drift_ms = max(0.0, as_float(max_freshness_drift_raw, 100.0))
max_clock_uncertainty_ms = max(0.0, as_float(max_clock_uncertainty_raw, 100.0))
min_freshness_pairs = max(1, as_int_or_none(min_freshness_pairs_raw) or 1)
configured_capture_duration_s = as_float(capture_seconds_raw, 0.0)
capture_start_grace_s = as_float(capture_start_grace_seconds_raw, 0.0)
freshness_p95_values = [
    value
    for value in [
        video_freshness_stats.get("p95_ms"),
        audio_freshness_stats.get("p95_ms"),
    ]
    if value is not None
]
event_age_p95_values = [
    value
    for value in [
        video_event_age_stats.get("p95_ms"),
        audio_event_age_stats.get("p95_ms"),
    ]
    if value is not None
]
freshness_drift_values = [
    abs(value)
    for value in [
        video_event_age_model.get("drift_ms"),
        audio_event_age_model.get("drift_ms"),
    ]
    if value is not None and math.isfinite(value)
]
freshness_worst_p95_ms = max(freshness_p95_values) if freshness_p95_values else None
freshness_worst_event_p95_ms = max(event_age_p95_values) if event_age_p95_values else None
freshness_worst_drift_ms = max(freshness_drift_values) if freshness_drift_values else None
freshness_worst_event_with_uncertainty_ms = (
    freshness_worst_event_p95_ms + clock_uncertainty_ms
    if freshness_worst_event_p95_ms is not None
    else None
)
event_age_min_values = [
    value
    for value in [
        video_event_age_stats.get("min_ms"),
        audio_event_age_stats.get("min_ms"),
    ]
    if value is not None and math.isfinite(value)
]
freshness_min_event_age_ms = min(event_age_min_values) if event_age_min_values else None
smoothness = analyze_smoothness(capture_path, report, timeline)
media_timeline = smoothness.get("media_timeline") or {}
observed_capture_duration_values = [
    value
    for value in [
        finite(media_timeline.get("video_last_s")),
        finite(media_timeline.get("audio_last_s")),
    ]
    if value is not None
]
observed_capture_duration_s = (
    max(observed_capture_duration_values) if observed_capture_duration_values else None
)


def negative_rows(rows, key, limit_ms):
    return [
        row
        for row in rows
        if row.get(key) is not None
        and math.isfinite(row[key])
        and row[key] < -limit_ms
    ]


def non_monotonic_rows(rows, key):
    bad = []
    previous = None
    for row in rows:
        value = finite(row.get(key))
        if value is None:
            continue
        if previous is not None and value + 1e-9 < previous:
            bad.append(row)
        previous = value
    return bad


video_negative_rows = negative_rows(joined, "video_freshness_ms", clock_uncertainty_ms)
audio_negative_rows = negative_rows(joined, "audio_freshness_ms", clock_uncertainty_ms)
monotonic_checks = {
    "server video pipeline": non_monotonic_rows(joined, "server_video_pipeline_unix_ns"),
    "server audio pipeline": non_monotonic_rows(joined, "server_audio_pipeline_unix_ns"),
    "Tethys video observation": non_monotonic_rows(joined, "tethys_video_time_s"),
    "Tethys audio observation": non_monotonic_rows(joined, "tethys_audio_time_s"),
}
non_monotonic_labels = [
    f"{label} ({len(rows)} reversal{'s' if len(rows) != 1 else ''})"
    for label, rows in monotonic_checks.items()
    if rows
]
capture_timebase_status = "valid"
capture_timebase_reason = "Tethys observation timestamps are monotonic against server pipeline references"
capture_timebase_reasons = []
capture_timebase_warnings = []
if not joined:
    capture_timebase_status = "unknown"
    capture_timebase_reason = "no joined coded events are available"
elif not clock_alignment_available:
    capture_timebase_status = "unknown"
    capture_timebase_reason = "server/capture clock alignment is unavailable"
elif video_negative_rows or audio_negative_rows:
    capture_timebase_status = "invalid"
    if video_negative_rows:
        capture_timebase_reasons.append(
            f"video observed before server pipeline injection {len(video_negative_rows)} time(s)"
        )
    if audio_negative_rows:
        capture_timebase_reasons.append(
            f"audio observed before server pipeline injection {len(audio_negative_rows)} time(s)"
        )
if non_monotonic_labels:
    capture_timebase_status = "invalid"
    capture_timebase_reasons.append(
        "non-monotonic event timestamps: " + ", ".join(non_monotonic_labels)
    )

span_gap_s = finite(media_timeline.get("audio_video_span_gap_s"))
if span_gap_s is not None and abs(span_gap_s) > 1.0:
    relation = "shorter" if span_gap_s > 0 else "longer"
    capture_timebase_warnings.append(
        f"captured audio stream is {abs(span_gap_s):.3f}s {relation} than video; "
        "treating whole-file media span as recorder coverage, not event freshness"
    )
if capture_timebase_reasons:
    capture_timebase_reason = "; ".join(capture_timebase_reasons)

server_start_capture_time_s = corrected_capture_time_s_from_unix_ns(
    as_int_or_none(timeline.get("server_start_unix_ns"))
)
first_timing_row = joined[0] if joined else {}
first_event_timing = {
    "event_id": first_timing_row.get("event_id"),
    "code": first_timing_row.get("code"),
    "server_audio_pipeline_capture_s": corrected_capture_time_s_from_unix_ns(
        first_timing_row.get("server_audio_pipeline_unix_ns")
    ),
    "server_video_pipeline_capture_s": corrected_capture_time_s_from_unix_ns(
        first_timing_row.get("server_video_pipeline_unix_ns")
    ),
    "server_audio_sink_push_capture_s": corrected_capture_time_s_from_unix_ns(
        first_timing_row.get("server_audio_push_unix_ns")
    ),
    "server_video_sink_feed_capture_s": corrected_capture_time_s_from_unix_ns(
        first_timing_row.get("server_video_feed_unix_ns")
    ),
    "tethys_audio_observed_s": first_timing_row.get("tethys_audio_time_s"),
    "tethys_video_observed_s": first_timing_row.get("tethys_video_time_s"),
    "audio_pipeline_freshness_ms": first_timing_row.get("audio_freshness_ms"),
    "video_pipeline_freshness_ms": first_timing_row.get("video_freshness_ms"),
}


def nearest_observation(expected_s, observations_s, tolerance_s):
    if expected_s is None:
        return None
    best = None
    for observed_s in observations_s:
        observed = finite(observed_s)
        if observed is None:
            continue
        delta_s = observed - expected_s
        if abs(delta_s) > tolerance_s:
            continue
        if best is None or abs(delta_s) < abs(best["delta_s"]):
            best = {
                "time_s": observed,
                "delta_ms": delta_s * 1000.0,
                "delta_s": delta_s,
            }
    return best


def build_event_visibility():
    video_onsets = [value for value in (finite(v) for v in report.get("video_onsets_s") or []) if value is not None]
    audio_onsets = [value for value in (finite(v) for v in report.get("audio_onsets_s") or []) if value is not None]
    raw_first_video = finite(report.get("raw_first_video_activity_s"))
    raw_first_audio = finite(report.get("raw_first_audio_activity_s"))
    pulse_period_s = max(0.001, as_float(timeline.get("pulse_period_ms"), 1000.0) / 1000.0)
    tolerance_s = max(0.20, min(0.45, pulse_period_s * 0.45))
    paired_by_server_event = {
        int(row["event_id"]): row for row in joined if row.get("event_id") is not None
    }
    events = []
    summary = {
        "paired": 0,
        "unpaired": 0,
        "video_only": 0,
        "audio_only": 0,
        "missing": 0,
    }
    for server in server_event_list:
        event_id = as_int_or_none(server.get("event_id"))
        if event_id is None:
            continue
        code = as_int_or_none(server.get("code"))
        video_pipeline_s = corrected_capture_time_s_from_unix_ns(
            shifted_unix_ns(as_int_or_none(server.get("video_feed_unix_ns")), -video_delay_ms)
        )
        audio_pipeline_s = corrected_capture_time_s_from_unix_ns(
            shifted_unix_ns(as_int_or_none(server.get("audio_push_unix_ns")), -audio_delay_ms)
        )
        video_match = nearest_observation(video_pipeline_s, video_onsets, tolerance_s)
        audio_match = nearest_observation(audio_pipeline_s, audio_onsets, tolerance_s)
        raw_video_match = (
            video_match is None
            and raw_first_video is not None
            and video_pipeline_s is not None
            and abs(raw_first_video - video_pipeline_s) <= tolerance_s
        )
        raw_audio_match = (
            audio_match is None
            and raw_first_audio is not None
            and audio_pipeline_s is not None
            and abs(raw_first_audio - audio_pipeline_s) <= tolerance_s
        )
        paired = paired_by_server_event.get(event_id)
        video_seen = video_match is not None or raw_video_match
        audio_seen = audio_match is not None or raw_audio_match
        if paired is not None:
            status = "paired"
        elif video_seen and audio_seen:
            status = "unpaired"
        elif video_seen:
            status = "video_only"
        elif audio_seen:
            status = "audio_only"
        else:
            status = "missing"
        summary[status] += 1
        events.append({
            "event_id": event_id,
            "code": code,
            "status": status,
            "paired": paired is not None,
            "expected_video_pipeline_s": video_pipeline_s,
            "expected_audio_pipeline_s": audio_pipeline_s,
            "video_seen": video_seen,
            "audio_seen": audio_seen,
            "video_source": "coded_onset" if video_match else ("raw_activity" if raw_video_match else None),
            "audio_source": "coded_onset" if audio_match else ("raw_activity" if raw_audio_match else None),
            "nearest_video_time_s": (video_match or {}).get("time_s"),
            "nearest_audio_time_s": (audio_match or {}).get("time_s"),
            "nearest_video_delta_ms": (video_match or {}).get("delta_ms"),
            "nearest_audio_delta_ms": (audio_match or {}).get("delta_ms"),
            "paired_video_time_s": paired.get("tethys_video_time_s") if paired else None,
            "paired_audio_time_s": paired.get("tethys_audio_time_s") if paired else None,
            "paired_skew_ms": paired.get("observed_skew_ms") if paired else None,
            "confidence": paired.get("confidence") if paired else None,
        })
    return {
        "schema": "lesavka.coded-event-visibility.v1",
        "scope": "server event slots classified against Tethys coded onsets and raw first activity",
        "match_tolerance_s": tolerance_s,
        "summary": summary,
        "events": events,
    }


event_visibility = build_event_visibility()
timing_map = {
    "schema": "lesavka.output-timing-map.v1",
    "scope": "capture-relative map of server pipeline reference, sink handoff, and Tethys observation times",
    "injection_scope": timeline.get("injection_scope", "server-final-output-handoff"),
    "server_pipeline_reference": timeline.get(
        "server_pipeline_reference",
        "generated media PTS before intentional sync delay",
    ),
    "sink_handoff_path": timeline.get(
        "sink_handoff_path",
        "video CameraRelay::feed; audio Voice::push",
    ),
    "client_uplink_included": bool(timeline.get("client_uplink_included", False)),
    "freshness_clock_starts_at": "server pipeline reference before intentional sync delay",
    "capture_start_unix_ns": capture_start_unix_ns,
    "server_start_unix_ns": as_int_or_none(timeline.get("server_start_unix_ns")),
    "server_start_capture_time_s": server_start_capture_time_s,
    "capture_to_server_start_ms": (
        server_start_capture_time_s * 1000.0 if server_start_capture_time_s is not None else None
    ),
    "probe_warmup_ms": as_float(timeline.get("warmup_us"), 0.0) / 1000.0,
    "probe_duration_ms": as_float(timeline.get("duration_us"), 0.0) / 1000.0,
    "active_probe_media_ms": (
        as_float(timeline.get("warmup_us"), 0.0) + as_float(timeline.get("duration_us"), 0.0)
    ) / 1000.0,
    "configured_capture_duration_s": configured_capture_duration_s,
    "observed_capture_duration_s": observed_capture_duration_s,
    "capture_start_grace_s": capture_start_grace_s,
    "intentional_audio_delay_ms": audio_delay_ms,
    "intentional_video_delay_ms": video_delay_ms,
    "first_event": first_event_timing,
}
if not event_age_p95_values:
    freshness_status = "invalid"
    freshness_reason = "clock-aligned server pipeline references and Tethys observations were not available"
elif len(joined) < min_freshness_pairs:
    freshness_status = "invalid"
    freshness_reason = (
        f"paired coded events {len(joined)} < {min_freshness_pairs}; "
        "freshness requires enough matched injected/observed events"
    )
elif not clock_alignment_available or clock_uncertainty_ms > max_clock_uncertainty_ms:
    freshness_status = "invalid"
    freshness_reason = (
        "server/capture clock alignment unavailable or too uncertain: "
        f"{clock_uncertainty_ms:.1f} ms exceeds {max_clock_uncertainty_ms:.1f} ms limit"
    )
elif capture_timebase_status == "invalid":
    freshness_status = "invalid"
    freshness_reason = f"capture timebase invalid: {capture_timebase_reason}"
elif freshness_min_event_age_ms is not None and freshness_min_event_age_ms < -clock_uncertainty_ms:
    freshness_status = "invalid"
    freshness_reason = (
        f"one media stream has impossible negative pipeline freshness beyond clock uncertainty: "
        f"minimum freshness {freshness_min_event_age_ms:.1f} ms, uncertainty "
        f"{clock_uncertainty_ms:.1f} ms"
    )
elif freshness_worst_event_p95_ms < -clock_uncertainty_ms:
    freshness_status = "invalid"
    freshness_reason = (
        f"freshness was negative beyond clock uncertainty: "
        f"worst pipeline p95 {freshness_worst_event_p95_ms:.1f} ms, uncertainty "
        f"{clock_uncertainty_ms:.1f} ms"
    )
elif not sync_passed:
    freshness_status = "unknown"
    freshness_reason = (
        "sync did not pass, so freshness from paired signatures is not trustworthy: "
        f"{sync_verdict.get('status', 'unknown')} - {sync_verdict.get('reason', '')}"
    )
elif freshness_worst_event_with_uncertainty_ms <= max_freshness_age_ms and (
    freshness_worst_drift_ms is None or freshness_worst_drift_ms <= max_freshness_drift_ms
):
    freshness_status = "pass"
    freshness_reason = (
        f"worst pipeline p95 {freshness_worst_event_p95_ms:.1f} ms + clock uncertainty "
        f"{clock_uncertainty_ms:.1f} ms <= {max_freshness_age_ms:.1f} ms and worst freshness drift "
        f"{(freshness_worst_drift_ms or 0.0):.1f} ms <= {max_freshness_drift_ms:.1f} ms"
    )
else:
    freshness_status = "fail"
    freshness_reason = (
        f"worst pipeline p95 "
        f"{freshness_worst_event_p95_ms if freshness_worst_event_p95_ms is not None else 0.0:.1f} ms "
        f"+ clock uncertainty {clock_uncertainty_ms:.1f} ms "
        f"(limit {max_freshness_age_ms:.1f} ms), worst freshness drift "
        f"{freshness_worst_drift_ms if freshness_worst_drift_ms is not None else 0.0:.1f} ms "
        f"(limit {max_freshness_drift_ms:.1f} ms)"
    )

artifact = {
    "schema": "lesavka.output-delay-correlation.v1",
    "report_json": report_path,
    "server_timeline_json": timeline_path,
    "joined_event_count": len(joined),
    "audio_after_video_positive": True,
    "timing_map": timing_map,
    "observed_skew_model": observed_model,
    "server_feed_delta_model": server_model,
    "residual_path_skew_model": residual_model,
    "freshness": {
        "schema": "lesavka.output-freshness-summary.v1",
        "status": freshness_status,
        "reason": freshness_reason,
        "scope": "clock-corrected server pipeline reference to Tethys observed playback from the same paired signatures",
        "capture_start_unix_ns": capture_start_unix_ns,
        "capture_timebase": {
            "status": capture_timebase_status,
            "reason": capture_timebase_reason,
            "valid": capture_timebase_status == "valid",
            "warnings": capture_timebase_warnings,
            "video_observed_before_injection_count": len(video_negative_rows),
            "audio_observed_before_injection_count": len(audio_negative_rows),
            "non_monotonic_event_timestamps": non_monotonic_labels,
        },
        "clock_alignment": clock_alignment,
        "clock_uncertainty_ms": clock_uncertainty_ms,
        "max_age_limit_ms": max_freshness_age_ms,
        "max_drift_limit_ms": max_freshness_drift_ms,
        "max_clock_uncertainty_ms": max_clock_uncertainty_ms,
        "min_paired_events": min_freshness_pairs,
        "paired_event_count": len(joined),
        "intentional_audio_delay_ms": audio_delay_ms,
        "intentional_video_delay_ms": video_delay_ms,
        "worst_p95_sink_handoff_overhead_ms": max(
            [
                value
                for value in [
                    video_sink_handoff_stats.get("p95_ms"),
                    audio_sink_handoff_stats.get("p95_ms"),
                ]
                if value is not None
            ],
            default=None,
        ),
        "worst_p95_pipeline_freshness_ms": freshness_worst_p95_ms,
        "worst_event_age_p95_ms": freshness_worst_event_p95_ms,
        "worst_event_age_with_uncertainty_ms": freshness_worst_event_with_uncertainty_ms,
        "minimum_event_age_ms": freshness_min_event_age_ms,
        "minimum_pipeline_freshness_ms": freshness_min_event_age_ms,
        "sync_passed": sync_passed,
        "worst_p95_freshness_ms": freshness_worst_event_p95_ms,
        "worst_freshness_drift_ms": freshness_worst_drift_ms,
        "video_freshness_stats": video_freshness_stats,
        "audio_freshness_stats": audio_freshness_stats,
        "video_sink_handoff_overhead_stats": video_sink_handoff_stats,
        "audio_sink_handoff_overhead_stats": audio_sink_handoff_stats,
        "video_event_age_stats": video_event_age_stats,
        "audio_event_age_stats": audio_event_age_stats,
        "video_freshness_model": video_freshness_model,
        "audio_freshness_model": audio_freshness_model,
        "video_sink_handoff_overhead_model": video_sink_handoff_model,
        "audio_sink_handoff_overhead_model": audio_sink_handoff_model,
        "video_event_age_model": video_event_age_model,
        "audio_event_age_model": audio_event_age_model,
    },
    "smoothness": smoothness,
    "event_visibility": event_visibility,
    "server_observed_correlation": server_observed_correlation,
    "server_drift_share_of_observed": server_share,
    "dominant_layer": dominant_layer,
    "correction_mode": correction_mode,
    "video_delay_function_candidate": {
        "units": "microseconds",
        "end_to_end": {
            "intercept_us": round(observed_model.get("intercept_ms", 0.0) * 1000.0),
            "slope_us_per_s": round(observed_model.get("slope_ms_per_s", 0.0) * 1000.0),
            "formula": "video_delay_us(t) = intercept_us + slope_us_per_s * seconds_since_first_event",
        },
        "output_path_only": {
            "intercept_us": round(residual_model.get("intercept_ms", 0.0) * 1000.0),
            "slope_us_per_s": round(residual_model.get("slope_ms_per_s", 0.0) * 1000.0),
            "formula": "video_delay_us(t) = intercept_us + slope_us_per_s * seconds_since_first_event",
        },
    },
    "events": joined,
}

pathlib.Path(output_json_path).write_text(json.dumps(artifact, indent=2, sort_keys=True) + "\n")
with pathlib.Path(output_csv_path).open("w", newline="", encoding="utf-8") as handle:
    fieldnames = [
        "event_id",
        "observed_event_id",
        "observed_server_event_id",
        "observed_event_code",
        "match_method",
        "match_score_ms",
        "code",
        "event_time_s",
        "tethys_video_time_s",
        "tethys_audio_time_s",
        "observed_skew_ms",
        "server_video_feed_s",
        "server_audio_push_s",
        "server_video_feed_unix_ns",
        "server_audio_push_unix_ns",
        "server_video_pipeline_unix_ns",
        "server_audio_pipeline_unix_ns",
        "tethys_video_unix_ns",
        "tethys_audio_unix_ns",
        "video_freshness_ms",
        "audio_freshness_ms",
        "video_sink_handoff_overhead_ms",
        "audio_sink_handoff_overhead_ms",
        "video_event_age_ms",
        "audio_event_age_ms",
        "server_feed_delta_ms",
        "residual_path_skew_ms",
        "server_video_feed_monotonic_us",
        "server_audio_push_monotonic_us",
        "confidence",
    ]
    writer = csv.DictWriter(handle, fieldnames=fieldnames)
    writer.writeheader()
    for row in joined:
        writer.writerow({key: row.get(key) for key in fieldnames})


def fmt_number(value, unit="", precision=1):
    if value is None:
        return "n/a"
    try:
        value = float(value)
    except Exception:
        return "n/a"
    if not math.isfinite(value):
        return "n/a"
    return f"{value:.{precision}f}{unit}"


visibility_summary = event_visibility.get("summary") or {}
visibility_lines = [
    "Coded event visibility",
    f"- paired={visibility_summary.get('paired', 0)} unpaired={visibility_summary.get('unpaired', 0)} video_only={visibility_summary.get('video_only', 0)} audio_only={visibility_summary.get('audio_only', 0)} missing={visibility_summary.get('missing', 0)}",
]
for event in event_visibility.get("events", []):
    status = event.get("status", "unknown")
    if status == "paired":
        detail = (
            f"skew={fmt_number(event.get('paired_skew_ms'), ' ms')} "
            f"conf={fmt_number(event.get('confidence'), '', 3)}"
        )
    else:
        video_source = event.get("video_source") or "not_seen"
        audio_source = event.get("audio_source") or "not_seen"
        detail = f"video={video_source} audio={audio_source}"
    visibility_lines.append(
        f"- code {event.get('code')} event {event.get('event_id')}: {status} ({detail})"
    )


lines = [
    f"Output-delay correlation for {report_path}",
    f"- joined events: {len(joined)}",
    f"- dominant layer: {dominant_layer}",
    f"- correction mode: {correction_mode}",
    f"- observed skew model: {observed_model.get('intercept_ms', 0.0):+.3f} ms + {observed_model.get('slope_ms_per_s', 0.0):+.3f} ms/s * t",
    f"- server feed model: {server_model.get('intercept_ms', 0.0):+.3f} ms + {server_model.get('slope_ms_per_s', 0.0):+.3f} ms/s * t",
    f"- residual path model: {residual_model.get('intercept_ms', 0.0):+.3f} ms + {residual_model.get('slope_ms_per_s', 0.0):+.3f} ms/s * t",
    f"- server/observed correlation: {server_observed_correlation:+.3f}",
    f"- server drift share of observed: {server_share:.3f}",
    "",
    "Timing map",
    "- scope: capture-relative server pipeline reference, sink handoff, and Tethys observation",
    f"- injection point: {timing_map.get('injection_scope')} ({timing_map.get('sink_handoff_path')}); client uplink included={timing_map.get('client_uplink_included')}",
    f"- freshness clock starts at: {timing_map.get('freshness_clock_starts_at')}",
    f"- capture to server probe start: {fmt_number(timing_map.get('capture_to_server_start_ms'), ' ms')} (setup/pre-roll, not media latency)",
    f"- probe warmup before first coded event: {fmt_number(timing_map.get('probe_warmup_ms'), ' ms')}",
    f"- first event server pipeline refs: audio {fmt_number(first_event_timing.get('server_audio_pipeline_capture_s'), 's', 3)}, video {fmt_number(first_event_timing.get('server_video_pipeline_capture_s'), 's', 3)}",
    f"- first event sink handoff: audio {fmt_number(first_event_timing.get('server_audio_sink_push_capture_s'), 's', 3)}, video {fmt_number(first_event_timing.get('server_video_sink_feed_capture_s'), 's', 3)}",
    f"- first event Tethys observation: audio {fmt_number(first_event_timing.get('tethys_audio_observed_s'), 's', 3)}, video {fmt_number(first_event_timing.get('tethys_video_observed_s'), 's', 3)}",
    f"- first event pipeline freshness: audio {fmt_number(first_event_timing.get('audio_pipeline_freshness_ms'), ' ms')}, video {fmt_number(first_event_timing.get('video_pipeline_freshness_ms'), ' ms')}",
    "",
    "Freshness timeline",
    f"- capture duration: {fmt_number(observed_capture_duration_s if observed_capture_duration_s is not None else configured_capture_duration_s, 's', 3)} (recorder coverage, not media latency; configured {fmt_number(configured_capture_duration_s, 's', 3)})",
    f"- active probe media: {fmt_number(timing_map.get('active_probe_media_ms'), ' ms')} (warmup + coded events)",
    f"- setup before probe: {fmt_number(timing_map.get('capture_to_server_start_ms'), ' ms')} (capture readiness/tunnel/probe startup, not media latency)",
    f"- measured event freshness: video p95 {fmt_number(video_freshness_stats.get('p95_ms'), ' ms')}, audio p95 {fmt_number(audio_freshness_stats.get('p95_ms'), ' ms')}",
    f"- freshness verdict: {freshness_status} ({freshness_reason})",
    "",
    *visibility_lines,
    "",
    f"Output freshness for {report_path}",
    "- scope: clock-corrected server pipeline reference to Tethys observed playback from the same paired signatures",
    f"- freshness status: {freshness_status} ({freshness_reason})",
    f"- capture timebase: {capture_timebase_status} ({capture_timebase_reason})",
    *[f"- capture timebase warning: {warning}" for warning in capture_timebase_warnings],
    f"- clock uncertainty: +/-{clock_uncertainty_ms:.1f} ms",
    f"- intentional sync delays: audio {audio_delay_ms:+.1f} ms, video {video_delay_ms:+.1f} ms",
    f"- sink handoff overhead: video median {fmt_number(video_sink_handoff_stats.get('median_ms'), ' ms')} / p95 {fmt_number(video_sink_handoff_stats.get('p95_ms'), ' ms')} / max {fmt_number(video_sink_handoff_stats.get('max_ms'), ' ms')}; audio median {fmt_number(audio_sink_handoff_stats.get('median_ms'), ' ms')} / p95 {fmt_number(audio_sink_handoff_stats.get('p95_ms'), ' ms')} / max {fmt_number(audio_sink_handoff_stats.get('max_ms'), ' ms')}",
    f"- pipeline freshness: video median {fmt_number(video_freshness_stats.get('median_ms'), ' ms')} / p95 {fmt_number(video_freshness_stats.get('p95_ms'), ' ms')} / max {fmt_number(video_freshness_stats.get('max_ms'), ' ms')}; audio median {fmt_number(audio_freshness_stats.get('median_ms'), ' ms')} / p95 {fmt_number(audio_freshness_stats.get('p95_ms'), ' ms')} / max {fmt_number(audio_freshness_stats.get('max_ms'), ' ms')}",
    f"- freshness budget: worst pipeline p95 {fmt_number(freshness_worst_event_p95_ms, ' ms')} + clock uncertainty {clock_uncertainty_ms:.1f} ms = {fmt_number(freshness_worst_event_with_uncertainty_ms, ' ms')} vs limit {max_freshness_age_ms:.1f} ms",
    f"- video freshness drift: {video_event_age_model.get('drift_ms', 0.0):+.1f} ms over paired events ({video_event_age_model.get('slope_ms_per_s', 0.0):+.3f} ms/s)",
    f"- audio freshness drift: {audio_event_age_model.get('drift_ms', 0.0):+.1f} ms over paired events ({audio_event_age_model.get('slope_ms_per_s', 0.0):+.3f} ms/s)",
    "",
    "Output smoothness",
    f"- window: {smoothness.get('window_start_s') or 0.0:.3f}s to {smoothness.get('window_end_s') or 0.0:.3f}s",
    f"- capture media timeline: video span {fmt_number(media_timeline.get('video_span_s'), 's', 3)}, audio span {fmt_number(media_timeline.get('audio_span_s'), 's', 3)}, audio/video span gap {fmt_number(media_timeline.get('audio_video_span_gap_s'), 's', 3)}",
    f"- video cadence: frames {smoothness.get('video', {}).get('timestamps', 0)}, expected interval {(smoothness.get('video', {}).get('expected_interval_ms') or 0.0):.1f} ms, p95 jitter {(smoothness.get('video', {}).get('jitter_stats', {}).get('p95_jitter_ms') or 0.0):.1f} ms, max interval {(smoothness.get('video', {}).get('interval_stats', {}).get('max_interval_ms') or 0.0):.1f} ms, hiccups {smoothness.get('video', {}).get('hiccup_count', 0)}",
    f"- video continuity: decoded {smoothness.get('video', {}).get('decoded_frames', 0)}, duplicates {smoothness.get('video', {}).get('duplicate_frames', 0)}, estimated missing {smoothness.get('video', {}).get('estimated_missing_frames', 0)}, undecodable {smoothness.get('video', {}).get('undecodable_frames', 0)}",
    f"- audio packet cadence: packets {smoothness.get('audio', {}).get('packet_cadence', {}).get('timestamps', 0)}, p95 jitter {(smoothness.get('audio', {}).get('packet_cadence', {}).get('jitter_stats', {}).get('p95_jitter_ms') or 0.0):.1f} ms, max interval {(smoothness.get('audio', {}).get('packet_cadence', {}).get('interval_stats', {}).get('max_interval_ms') or 0.0):.1f} ms, hiccups {smoothness.get('audio', {}).get('packet_cadence', {}).get('hiccup_count', 0)}",
    f"- audio pilot continuity: low-RMS windows {smoothness.get('audio', {}).get('rms_continuity', {}).get('low_rms_window_count', 0)}, median RMS {(smoothness.get('audio', {}).get('rms_continuity', {}).get('rms_stats', {}).get('median_rms') or 0.0):.1f}",
]
summary = "\n".join(lines) + "\n"
pathlib.Path(output_txt_path).write_text(summary)
print(summary, end="")
PY
}

maybe_apply_output_delay_calibration() {
  [[ "${LESAVKA_OUTPUT_DELAY_CALIBRATION}" != "0" ]] || return 0
  [[ -f "${LOCAL_OUTPUT_DELAY_ENV}" ]] || return 0

  # shellcheck disable=SC1090
  source "${LOCAL_OUTPUT_DELAY_ENV}"

  echo "==> UVC/UAC output-delay calibration decision"
  echo "   ↪ output_delay_calibration_json=${LOCAL_OUTPUT_DELAY_JSON}"
  echo "   ↪ output_delay_ready=${output_delay_ready:-false}"
  echo "   ↪ output_delay_decision=${output_delay_decision:-unknown}"
  echo "   ↪ output_delay_target=${output_delay_target:-unknown}"
  echo "   ↪ output_delay_paired_event_count=${output_delay_paired_event_count:-0}"
  echo "   ↪ output_delay_measured_skew_ms=${output_delay_measured_skew_ms:-0.0}"
  echo "   ↪ output_delay_drift_ms=${output_delay_drift_ms:-0.0}"
  echo "   ↪ output_delay_audio_delta_us=${output_delay_audio_delta_us:-0}"
  echo "   ↪ output_delay_video_delta_us=${output_delay_video_delta_us:-0}"
  echo "   ↪ output_delay_active_audio_offset_us=${output_delay_active_audio_offset_us:-0}"
  echo "   ↪ output_delay_active_video_offset_us=${output_delay_active_video_offset_us:-0}"
  echo "   ↪ output_delay_audio_target_offset_us=${output_delay_audio_target_offset_us:-0}"
  echo "   ↪ output_delay_video_target_offset_us=${output_delay_video_target_offset_us:-0}"
  echo "   ↪ output_delay_apply_mode=${output_delay_apply_mode:-${LESAVKA_OUTPUT_DELAY_APPLY_MODE}}"
  echo "   ↪ output_delay_note=${output_delay_note:-}"

  if [[ "${output_delay_ready:-false}" != "true" ]]; then
    echo "   ↪ output delay calibration apply refused: ${output_delay_note:-not ready}"
    return 0
  fi

  if [[ "${LESAVKA_OUTPUT_DELAY_APPLY}" == "0" ]]; then
    echo "   ↪ output delay calibration apply disabled"
    return 0
  fi

  local apply_mode="${output_delay_apply_mode:-${LESAVKA_OUTPUT_DELAY_APPLY_MODE}}"
  local apply_audio_delta="${output_delay_audio_delta_us:-0}"
  local apply_video_delta="${output_delay_video_delta_us:-0}"

  if [[ "${apply_mode}" == "absolute" ]]; then
    local calibration_output current_audio_offset_us current_video_offset_us
    if ! calibration_output="$(
      LESAVKA_TLS_DOMAIN="${LESAVKA_TLS_DOMAIN}" \
        "${REPO_ROOT}/target/debug/lesavka-relayctl" \
          --server "${RESOLVED_LESAVKA_SERVER_ADDR}" \
          calibration 2>&1
    )"; then
      echo "   ↪ output delay calibration apply refused: current calibration query failed"
      echo "${calibration_output}" >&2
      return 0
    fi
    current_audio_offset_us="$(awk -F= '/^calibration_active_audio_offset_us=/{print $2; exit}' <<<"${calibration_output}")"
    current_video_offset_us="$(awk -F= '/^calibration_active_video_offset_us=/{print $2; exit}' <<<"${calibration_output}")"
    if [[ ! "${current_audio_offset_us}" =~ ^-?[0-9]+$ || ! "${current_video_offset_us}" =~ ^-?[0-9]+$ ]]; then
      echo "   ↪ output delay calibration apply refused: could not parse active calibration offsets"
      echo "${calibration_output}" >&2
      return 0
    fi
    apply_audio_delta=$(( ${output_delay_audio_target_offset_us:-0} - current_audio_offset_us ))
    apply_video_delta=$(( ${output_delay_video_target_offset_us:-0} - current_video_offset_us ))
    echo "   ↪ current_active_audio_offset_us=${current_audio_offset_us}"
    echo "   ↪ current_active_video_offset_us=${current_video_offset_us}"
    echo "   ↪ absolute_target_audio_offset_us=${output_delay_audio_target_offset_us:-0}"
    echo "   ↪ absolute_target_video_offset_us=${output_delay_video_target_offset_us:-0}"
  elif [[ "${apply_mode}" != "relative" ]]; then
    echo "   ↪ output delay calibration apply refused: unsupported apply mode ${apply_mode}"
    return 0
  fi

  echo "   ↪ calibration_apply_audio_delta_us=${apply_audio_delta}"
  echo "   ↪ calibration_apply_video_delta_us=${apply_video_delta}"
  echo "==> applying measured UVC/UAC output-delay calibration"
  LESAVKA_TLS_DOMAIN="${LESAVKA_TLS_DOMAIN}" \
    "${REPO_ROOT}/target/debug/lesavka-relayctl" \
      --server "${RESOLVED_LESAVKA_SERVER_ADDR}" \
      calibrate \
      "${apply_audio_delta}" \
      "${apply_video_delta}" \
      "${output_delay_note:-direct UVC/UAC output-delay calibration}"

  if [[ "${LESAVKA_OUTPUT_DELAY_SAVE}" != "0" ]]; then
    echo "==> saving measured UVC/UAC output-delay calibration as default"
    LESAVKA_TLS_DOMAIN="${LESAVKA_TLS_DOMAIN}" \
      "${REPO_ROOT}/target/debug/lesavka-relayctl" \
        --server "${RESOLVED_LESAVKA_SERVER_ADDR}" \
        calibration-save-default
  fi
}

maybe_run_output_delay_confirmation() {
  [[ "${LESAVKA_OUTPUT_DELAY_CONFIRM}" != "0" ]] || return 0
  [[ "${LESAVKA_OUTPUT_DELAY_APPLY}" != "0" ]] || return 0
  [[ "${LESAVKA_OUTPUT_DELAY_CONFIRMING:-0}" != "1" ]] || return 0
  [[ "${output_delay_ready:-false}" == "true" ]] || return 0

  local confirm_audio_delay="${output_delay_audio_target_offset_us:-0}"
  local confirm_video_delay="${output_delay_video_target_offset_us:-0}"
  local confirm_output_dir="${LOCAL_REPORT_DIR}/confirmation"
  mkdir -p "${confirm_output_dir}"

  echo "==> confirming fixed UVC/UAC output-delay calibration"
  echo "   ↪ confirmation_audio_delay_us=${confirm_audio_delay}"
  echo "   ↪ confirmation_video_delay_us=${confirm_video_delay}"
  echo "   ↪ confirmation_requires_sync_pass=1"

  LESAVKA_OUTPUT_DELAY_CONFIRMING=1 \
  LESAVKA_OUTPUT_DELAY_CONFIRM=0 \
  LESAVKA_OUTPUT_DELAY_APPLY=0 \
  LESAVKA_OUTPUT_DELAY_SAVE=0 \
  LESAVKA_OUTPUT_REQUIRE_SYNC_PASS=1 \
  LESAVKA_OUTPUT_DELAY_PROBE_AUDIO_DELAY_US="${confirm_audio_delay}" \
  LESAVKA_OUTPUT_DELAY_PROBE_VIDEO_DELAY_US="${confirm_video_delay}" \
  PROBE_PREBUILD=0 \
  LOCAL_OUTPUT_DIR="${confirm_output_dir}" \
    "${SCRIPT_DIR}/run_upstream_av_sync.sh"
}

enforce_sync_verdict() {
  [[ "${LESAVKA_OUTPUT_REQUIRE_SYNC_PASS}" != "0" ]] || return 0
  [[ -f "${LOCAL_ANALYSIS_JSON}" ]] || {
    echo "required sync pass unavailable: missing ${LOCAL_ANALYSIS_JSON}" >&2
    exit 94
  }

  python3 - <<'PY' "${LOCAL_ANALYSIS_JSON}"
import json
import pathlib
import sys

report = json.loads(pathlib.Path(sys.argv[1]).read_text())
verdict = report.get("verdict") or {}
if verdict.get("passed") is True:
    raise SystemExit(0)
print(
    "required sync pass failed: "
    f"{verdict.get('status', 'unknown')} - {verdict.get('reason', '')}",
    file=sys.stderr,
)
raise SystemExit(94)
PY
}

if [[ "${PROBE_PREBUILD}" != "0" ]]; then
  echo "==> prebuilding relay control/analyzer before opening the capture window"
  (
    cd "${REPO_ROOT}"
    cargo build -p lesavka_client --bin lesavka-sync-analyze --bin lesavka-relayctl
  )
fi

if [[ ! -x "${ANALYZE_BIN}" ]]; then
  echo "sync analyzer binary not found at ${ANALYZE_BIN}" >&2
  exit 1
fi

resolve_server_addr
echo "==> resolved Lesavka server addr: ${RESOLVED_LESAVKA_SERVER_ADDR}"
if [[ -n "${SERVER_TUNNEL_PID}" ]]; then
  echo "   ↪ tunneled to ${LESAVKA_SERVER_HOST}:127.0.0.1:${SERVER_TUNNEL_REMOTE_PORT}"
fi

print_lesavka_versions
preflight_server_path
write_clock_alignment

echo "==> starting Tethys capture on ${TETHYS_HOST}"
ssh ${SSH_OPTS} "${TETHYS_HOST}" bash -s -- \
  "${REMOTE_CAPTURE}" \
  "${CAPTURE_SECONDS}" \
  "${REMOTE_VIDEO_DEVICE}" \
  "${VIDEO_SIZE}" \
  "${VIDEO_FPS}" \
  "${VIDEO_FORMAT}" \
  "${REMOTE_CAPTURE_STACK}" \
  "${REMOTE_PULSE_CAPTURE_TOOL}" \
  "${REMOTE_PULSE_VIDEO_MODE}" \
  "${REMOTE_PULSE_AUDIO_ANCHOR_SILENCE}" \
  "${REMOTE_AUDIO_SOURCE}" \
  "${REMOTE_AUDIO_QUIESCE_USER_AUDIO}" \
  "${REMOTE_CAPTURE_ALLOW_ALSA_FALLBACK}" \
  "${REMOTE_CAPTURE_PREROLL_DISCARD_SECONDS}" \
  "${REMOTE_CAPTURE_READY_SETTLE_SECONDS}" \
  > >(tee "${LOCAL_CAPTURE_LOG}") \
  2> >(tee -a "${LOCAL_CAPTURE_LOG}" >&2) <<'REMOTE_CAPTURE_SCRIPT' &
set -euo pipefail
remote_capture=$1
capture_seconds=$2
remote_video_device=$3
video_size=$4
video_fps=$5
video_format=$6
remote_capture_stack=$7
remote_pulse_capture_tool=$8
remote_pulse_video_mode=$9
remote_pulse_audio_anchor_silence=${10}
remote_audio_source=${11}
remote_audio_quiesce_user_audio=${12}
remote_capture_allow_alsa_fallback=${13}
remote_capture_preroll_discard_seconds=${14}
remote_capture_ready_settle_seconds=${15}

rm -f "${remote_capture}"

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_video_device() {
  local requested=$1
  if [[ "${requested}" != "auto" ]]; then
    printf '%s\n' "${requested}"
    return 0
  fi

  local by_id
  by_id=$(find /dev/v4l/by-id -maxdepth 1 -type l -name '*Lesavka*video-index0' 2>/dev/null | head -n1 || true)
  if [[ -n "${by_id}" ]]; then
    printf '%s\n' "${by_id}"
    return 0
  fi

  if command -v v4l2-ctl >/dev/null 2>&1; then
    local resolved
    resolved="$(
      v4l2-ctl --list-devices 2>/dev/null \
        | awk '
            BEGIN { want=0 }
            /Lesavka Composite: UVC Camera/ { want=1; next }
            /^[^ \t]/ { want=0 }
            want && /^[ \t]+\/dev\/video[0-9]+/ {
              gsub(/^[ \t]+/, "", $0)
              print
              exit
            }
          '
    )"
    if [[ -n "${resolved}" ]]; then
      printf '%s\n' "${resolved}"
      return 0
    fi
  fi

  printf 'Lesavka UVC video device not found on Tethys; refusing to fall back to an unrelated webcam/capture card.\n' >&2
  exit 64
}

resolve_pulse_source() {
  if ! command -v pactl >/dev/null 2>&1; then
    return 1
  fi
  pactl list short sources 2>/dev/null \
    | awk '
      /alsa_input\..*Lesavka_Composite/ { print $2; found=1; exit }
      /Lesavka_Composite/ && !fallback { fallback=$2 }
      END {
        if (found) exit 0
        if (fallback != "") { print fallback; exit 0 }
        exit 1
      }
    '
}

resolve_alsa_audio_device() {
  if ! command -v arecord >/dev/null 2>&1; then
    return 1
  fi

  arecord -l 2>/dev/null | awk '
    /^card [0-9]+:/ && ($0 ~ /Lesavka|UAC2_Gadget|UAC2Gadget|Composite/) {
      card=$2
      sub(":", "", card)
      for (i = 1; i <= NF; i++) {
        if ($i == "device") {
          dev=$(i + 1)
          sub(":", "", dev)
          printf "hw:%s,%s\n", card, dev
          found=1
          exit 0
        }
      }
    }
    END {
      if (!found) exit 1
    }
  '
}

gst_video_source_caps() {
  case "${video_format}" in
    ""|mjpeg|MJPG)
      printf 'image/jpeg,width=%s,height=%s,framerate=%s/1' \
        "${resolved_video_size%x*}" \
        "${resolved_video_size#*x}" \
        "${resolved_video_fps}"
      ;;
    yuyv422|YUYV|yuyv)
      printf 'video/x-raw,format=YUY2,width=%s,height=%s,framerate=%s/1' \
        "${resolved_video_size%x*}" \
        "${resolved_video_size#*x}" \
        "${resolved_video_fps}"
      ;;
    *)
      printf 'unsupported gst video_format=%s\n' "${video_format}" >&2
      exit 64
      ;;
  esac
}

gst_video_decode_chain() {
  case "${video_format}" in
    ""|mjpeg|MJPG)
      printf 'jpegdec ! '
      ;;
    yuyv422|YUYV|yuyv)
      printf ''
      ;;
    *)
      printf 'unsupported gst video_format=%s\n' "${video_format}" >&2
      exit 64
      ;;
  esac
}

gst_audio_mixer_element() {
  if gst-inspect-1.0 audiomixer 2>/dev/null | grep -q 'ignore-inactive-pads'; then
    printf 'audiomixer name=amix ignore-inactive-pads=true'
  else
    printf 'audiomixer name=amix'
  fi
}

current_video_profile() {
  if ! command -v v4l2-ctl >/dev/null 2>&1; then
    return 1
  fi

  v4l2-ctl -d "${resolved_video_device}" --all 2>/dev/null \
    | awk '
        /Width\/Height[[:space:]]*:/ {
          split($0, a, ":")
          gsub(/^[ \t]+/, "", a[2])
          split(a[2], wh, "/")
          width=wh[1]
          height=wh[2]
          next
        }
        /Frames per second:/ {
          fps=$4
          sub(/\..*/, "", fps)
        }
        END {
          if (width != "" && height != "") {
            print "size=" width "x" height
          }
          if (fps != "") {
            print "fps=" fps
          }
        }
      '
}

resolve_video_size() {
  local requested=$1
  if [[ "${requested}" != "auto" ]]; then
    printf '%s\n' "${requested}"
    return 0
  fi
  local current_profile
  current_profile="$(current_video_profile || true)"
  local current_size
  current_size="$(awk -F= '/^size=/{print $2; exit}' <<<"${current_profile}")"
  if [[ -n "${current_size}" ]]; then
    printf '%s\n' "${current_size}"
    return 0
  fi
  if ! command -v v4l2-ctl >/dev/null 2>&1; then
    printf '640x480\n'
    return 0
  fi
  local listing
  listing="$(v4l2-ctl -d "${resolved_video_device}" --list-formats-ext 2>/dev/null || true)"
  local preferred
  for preferred in 1920x1080 1360x768 1280x720; do
    if grep -q "Size: Discrete ${preferred}" <<<"${listing}"; then
      printf '%s\n' "${preferred}"
      return 0
    fi
  done
  local first_size
  first_size="$(grep -m1 -o 'Size: Discrete [0-9]\+x[0-9]\+' <<<"${listing}" | awk '{print $3}' || true)"
  if [[ -n "${first_size}" ]]; then
    printf '%s\n' "${first_size}"
    return 0
  fi
  printf '640x480\n'
}

resolve_video_fps() {
  local requested=$1
  if [[ "${requested}" != "auto" ]]; then
    printf '%s\n' "${requested}"
    return 0
  fi
  local current_profile
  current_profile="$(current_video_profile || true)"
  local current_fps
  current_fps="$(awk -F= '/^fps=/{print $2; exit}' <<<"${current_profile}")"
  if [[ -n "${current_fps}" ]]; then
    printf '%s\n' "${current_fps}"
    return 0
  fi
  if ! command -v v4l2-ctl >/dev/null 2>&1; then
    printf '20\n'
    return 0
  fi
  local listing
  listing="$(v4l2-ctl -d "${resolved_video_device}" --list-formats-ext 2>/dev/null || true)"
  local first_fps
  first_fps="$(grep -m1 -o '[0-9]\+\.[0-9]\+ fps' <<<"${listing}" | awk '{sub(/\..*/, "", $1); print $1}' || true)"
  if [[ -n "${first_fps}" ]]; then
    printf '%s\n' "${first_fps}"
    return 0
  fi
  printf '20\n'
}

resolve_pw_audio_target() {
  if ! command -v pw-dump >/dev/null 2>&1 || ! command -v python3 >/dev/null 2>&1; then
    return 1
  fi
  pw-dump | python3 -c '
import json
import sys

try:
    objs = json.load(sys.stdin)
except Exception:
    raise SystemExit(1)

for obj in objs:
    if obj.get("type") != "PipeWire:Interface:Node":
        continue
    props = (obj.get("info") or {}).get("props") or {}
    if props.get("media.class") != "Audio/Source":
        continue
    serial = props.get("object.serial")
    name = props.get("node.name", "")
    desc = props.get("node.description", "")
    if serial is None:
        continue
    if "Lesavka_Composite" in name or "Lesavka Composite" in desc:
        print(serial)
        raise SystemExit(0)

raise SystemExit(1)
'
}

capture_mode=""
alsa_audio_dev="hw:3,0"
pulse_source=""
pw_audio_target=""

case "${remote_capture_stack}" in
  auto)
    if [[ "${remote_audio_source}" == "auto" ]]; then
      if pulse_source="$(resolve_pulse_source)"; then
        capture_mode="pulse"
      elif command -v pw-record >/dev/null 2>&1 \
        && command -v pw-v4l2 >/dev/null 2>&1 \
        && pw_audio_target="$(resolve_pw_audio_target)"; then
        capture_mode="pwpipe"
      elif [[ "${remote_capture_allow_alsa_fallback}" == "1" ]] && alsa_audio_dev="$(resolve_alsa_audio_device)"; then
        capture_mode="alsa"
        printf 'PipeWire Lesavka source not found; using explicit diagnostic ALSA fallback device %s\n' "${alsa_audio_dev}" >&2
      else
        printf 'Lesavka Pulse/PipeWire audio source not found; refusing raw ALSA fallback for timing-sensitive capture.\n' >&2
        printf 'Set REMOTE_CAPTURE_STACK=alsa or REMOTE_CAPTURE_ALLOW_ALSA_FALLBACK=1 only for diagnostic signal-presence checks.\n' >&2
        exit 64
      fi
    elif [[ "${remote_audio_source}" == pulse:* ]]; then
      capture_mode="pulse"
      pulse_source="${remote_audio_source#pulse:}"
    elif [[ "${remote_audio_source}" == alsa:* ]]; then
      capture_mode="alsa"
      alsa_audio_dev="${remote_audio_source#alsa:}"
    else
      printf 'unsupported REMOTE_AUDIO_SOURCE=%s\n' "${remote_audio_source}" >&2
      exit 64
    fi
    ;;
  pwpipe)
    if ! command -v pw-record >/dev/null 2>&1 || ! command -v pw-v4l2 >/dev/null 2>&1; then
      printf 'REMOTE_CAPTURE_STACK=pwpipe requires pw-record and pw-v4l2\n' >&2
      exit 64
    fi
    pw_audio_target="$(resolve_pw_audio_target)" || {
      printf 'PipeWire Lesavka capture target not found for REMOTE_CAPTURE_STACK=pwpipe\n' >&2
      exit 64
    }
    capture_mode="pwpipe"
    ;;
  pulse)
    if [[ "${remote_audio_source}" == pulse:* ]]; then
      pulse_source="${remote_audio_source#pulse:}"
    elif [[ "${remote_audio_source}" == "auto" ]]; then
      pulse_source="$(resolve_pulse_source)" || {
        printf 'PipeWire Lesavka source not found for REMOTE_CAPTURE_STACK=pulse\n' >&2
        exit 64
      }
    else
      pulse_source="${remote_audio_source}"
    fi
    capture_mode="pulse"
    ;;
  alsa)
    if [[ "${remote_audio_source}" == alsa:* ]]; then
      alsa_audio_dev="${remote_audio_source#alsa:}"
    elif [[ "${remote_audio_source}" != "auto" ]]; then
      alsa_audio_dev="${remote_audio_source}"
    fi
    capture_mode="alsa"
    ;;
  *)
    printf 'unsupported REMOTE_CAPTURE_STACK=%s\n' "${remote_capture_stack}" >&2
    exit 64
    ;;
esac

resolved_video_device="$(resolve_video_device "${remote_video_device}")"
resolved_video_size="$(resolve_video_size "${video_size}")"
resolved_video_fps="$(resolve_video_fps "${video_fps}")"
printf 'using video device: %s\n' "${resolved_video_device}" >&2
printf 'using video mode: %s @ %s fps (%s)\n' "${resolved_video_size}" "${resolved_video_fps}" "${video_format:-driver-default}" >&2
video_args=(-f video4linux2 -framerate "${resolved_video_fps}" -video_size "${resolved_video_size}")
if [[ -n "${video_format}" ]]; then
  video_args+=(-input_format "${video_format}")
fi
gst_source_caps="$(gst_video_source_caps)"
gst_decode_chain="$(gst_video_decode_chain)"
gst_audio_mixer="$(gst_audio_mixer_element)"

run_ffmpeg_capture() {
  local rc=0
  announce_capture_start
  timeout --kill-after=5 --signal=INT "$((capture_seconds + 5))" "$@" </dev/null &
  local capture_pid=$!
  signal_capture_ready
  wait "${capture_pid}" || rc=$?
  case "${rc}" in
    0|124|130)
      return 0
      ;;
    *)
      return "${rc}"
      ;;
  esac
}

announce_capture_start() {
  printf 'capture_start_unix_ns=%s\n' "$(date +%s%N)" >&2
}

signal_capture_ready() {
  if [[ "${remote_capture_ready_settle_seconds}" =~ ^[0-9]+([.][0-9]+)?$ ]]; then
    sleep "${remote_capture_ready_settle_seconds}"
  fi
  printf '%s\n' "__LESAVKA_CAPTURE_READY__"
}

run_tolerant_capture() {
  announce_capture_start
  "$@" &
  local capture_pid=$!
  signal_capture_ready
  wait "${capture_pid}" || true
}

quiesce_for_alsa=0
case "${remote_audio_quiesce_user_audio}" in
  1|true|yes)
    quiesce_for_alsa=1
    ;;
  auto)
    if [[ "${capture_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 [[ "${capture_mode}" == "alsa" && "${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

discard_preroll_capture() {
  local seconds=$1
  [[ "${seconds}" =~ ^[0-9]+$ ]] || return 0
  [[ "${seconds}" -gt 0 ]] || return 0
  local discard="/tmp/lesavka-output-delay-preroll-discard-$RANDOM.mkv"
  printf 'discarding %ss of post-enumeration capture before probe\n' "${seconds}" >&2
  case "${capture_mode}" in
    pulse)
      if command -v ffmpeg >/dev/null 2>&1; then
        timeout --kill-after=3 --signal=INT "$((seconds + 5))" ffmpeg -nostdin -hide_banner -loglevel error -y \
          -thread_queue_size 1024 \
          "${video_args[@]}" \
          -i "${resolved_video_device}" \
          -thread_queue_size 1024 \
          -f pulse \
          -i "${pulse_source}" \
          -t "${seconds}" \
          -c:v copy \
          -c:a pcm_s16le \
          "${discard}" </dev/null >/dev/null 2>&1 || true
      fi
      ;;
    alsa)
      if command -v ffmpeg >/dev/null 2>&1; then
        timeout --kill-after=3 --signal=INT "$((seconds + 5))" ffmpeg -nostdin -hide_banner -loglevel error -y \
          -thread_queue_size 1024 \
          "${video_args[@]}" \
          -i "${resolved_video_device}" \
          -thread_queue_size 1024 \
          -f alsa -ac 2 -ar 48000 \
          -i "${alsa_audio_dev}" \
          -t "${seconds}" \
          -c:v copy \
          -c:a pcm_s16le \
          "${discard}" </dev/null >/dev/null 2>&1 || true
      fi
      ;;
    *)
      sleep "${seconds}"
      ;;
  esac
  rm -f "${discard}"
}

discard_preroll_capture "${remote_capture_preroll_discard_seconds}"

if [[ "${capture_mode}" == "pwpipe" ]]; then
  printf 'using PipeWire-native mux capture target serial: %s\n' "${pw_audio_target}" >&2
  announce_capture_start
  (
    timeout "${capture_seconds}" pw-record \
      --target "${pw_audio_target}" \
      --rate 48000 \
      --channels 2 \
      --format s16 \
      --raw - \
      | pw-v4l2 ffmpeg -hide_banner -loglevel error -y \
        -thread_queue_size 1024 \
        "${video_args[@]}" \
        -i "${resolved_video_device}" \
        -thread_queue_size 1024 \
        -f s16le -ar 48000 -ac 2 \
        -i pipe:0 \
        -t "${capture_seconds}" \
        -c:v copy \
        -c:a pcm_s16le \
        "${remote_capture}"
  ) &
  capture_pid=$!
  signal_capture_ready
  wait "${capture_pid}"
elif [[ "${capture_mode}" == "pulse" ]]; then
  printf 'using Pulse source: %s\n' "${pulse_source}" >&2
  case "${remote_pulse_capture_tool}" in
    ffmpeg)
      case "${remote_pulse_video_mode}" in
        copy)
          run_ffmpeg_capture ffmpeg -hide_banner -loglevel error -y \
            -thread_queue_size 1024 \
            "${video_args[@]}" \
            -i "${resolved_video_device}" \
            -thread_queue_size 1024 \
            -f pulse \
            -i "${pulse_source}" \
            -map 0:v:0 \
            -map 1:a:0 \
            -t "${capture_seconds}" \
            -c:v copy \
            -c:a pcm_s16le \
            "${remote_capture}"
          ;;
        cfr)
          run_ffmpeg_capture ffmpeg -hide_banner -loglevel error -y \
            -thread_queue_size 1024 \
            "${video_args[@]}" \
            -i "${resolved_video_device}" \
            -thread_queue_size 1024 \
            -f pulse \
            -i "${pulse_source}" \
            -map 0:v:0 \
            -map 1:a:0 \
            -t "${capture_seconds}" \
            -vf "fps=${resolved_video_fps}" \
            -c:v libx264 -preset ultrafast -crf 12 -g 1 -pix_fmt yuv420p \
            -c:a pcm_s16le \
            "${remote_capture}"
          ;;
        *)
          printf 'unsupported REMOTE_PULSE_VIDEO_MODE=%s\n' "${remote_pulse_video_mode}" >&2
          exit 64
          ;;
      esac
      ;;
      gst)
        if [[ "${remote_pulse_audio_anchor_silence}" == "1" ]]; then
          printf 'anchoring Pulse capture audio timeline with generated silence\n' >&2
        fi
        case "${remote_pulse_video_mode}" in
          copy)
            if [[ "${video_format}" != "mjpeg" && "${video_format}" != "MJPG" && -n "${video_format}" ]]; then
              printf 'gst copy mode only supports mjpeg input, got %s\n' "${video_format}" >&2
              exit 64
            fi
            if [[ "${remote_pulse_audio_anchor_silence}" == "1" ]]; then
              run_tolerant_capture timeout --kill-after=5 --signal=INT "$((capture_seconds + 3))" \
                gst-launch-1.0 -q -e \
                  matroskamux name=mux ! filesink location="${remote_capture}" \
                  v4l2src device="${resolved_video_device}" do-timestamp=true ! \
                    ${gst_source_caps} ! \
                    queue ! mux. \
                  ${gst_audio_mixer} ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    queue ! mux. \
                  audiotestsrc wave=silence is-live=true do-timestamp=true ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    queue ! amix. \
                  pulsesrc device="${pulse_source}" do-timestamp=true ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    audioconvert ! audioresample ! audio/x-raw,rate=48000,channels=2 ! \
                    queue ! amix.
            else
              run_tolerant_capture timeout --kill-after=5 --signal=INT "$((capture_seconds + 3))" \
                gst-launch-1.0 -q -e \
                  matroskamux name=mux ! filesink location="${remote_capture}" \
                  v4l2src device="${resolved_video_device}" do-timestamp=true ! \
                    ${gst_source_caps} ! \
                    queue ! mux. \
                  pulsesrc device="${pulse_source}" do-timestamp=true ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    audioconvert ! audioresample ! audio/x-raw,rate=48000,channels=2 ! \
                    queue ! mux.
            fi
            ;;
          cfr)
            if [[ "${remote_pulse_audio_anchor_silence}" == "1" ]]; then
              run_tolerant_capture timeout --kill-after=5 --signal=INT "$((capture_seconds + 3))" \
                gst-launch-1.0 -q -e \
                  matroskamux name=mux ! filesink location="${remote_capture}" \
                  v4l2src device="${resolved_video_device}" do-timestamp=true ! \
                    ${gst_source_caps} ! \
                    ${gst_decode_chain} \
                    videoconvert ! videorate ! video/x-raw,framerate="${resolved_video_fps}"/1 ! \
                    x264enc tune=zerolatency speed-preset=ultrafast key-int-max=1 bitrate=5000 ! \
                    h264parse ! \
                    queue ! mux. \
                  ${gst_audio_mixer} ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    queue ! mux. \
                  audiotestsrc wave=silence is-live=true do-timestamp=true ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    queue ! amix. \
                  pulsesrc device="${pulse_source}" do-timestamp=true ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    audioconvert ! audioresample ! audio/x-raw,rate=48000,channels=2 ! \
                    queue ! amix.
            else
              run_tolerant_capture timeout --kill-after=5 --signal=INT "$((capture_seconds + 3))" \
                gst-launch-1.0 -q -e \
                  matroskamux name=mux ! filesink location="${remote_capture}" \
                  v4l2src device="${resolved_video_device}" do-timestamp=true ! \
                    ${gst_source_caps} ! \
                    ${gst_decode_chain} \
                    videoconvert ! videorate ! video/x-raw,framerate="${resolved_video_fps}"/1 ! \
                    x264enc tune=zerolatency speed-preset=ultrafast key-int-max=1 bitrate=5000 ! \
                    h264parse ! \
                    queue ! mux. \
                  pulsesrc device="${pulse_source}" do-timestamp=true ! \
                    audio/x-raw,rate=48000,channels=2 ! \
                    audioconvert ! audioresample ! audio/x-raw,rate=48000,channels=2 ! \
                    queue ! mux.
            fi
            ;;
        *)
          printf 'unsupported REMOTE_PULSE_VIDEO_MODE=%s\n' "${remote_pulse_video_mode}" >&2
          exit 64
          ;;
      esac
      ;;
    *)
      printf 'unsupported REMOTE_PULSE_CAPTURE_TOOL=%s\n' "${remote_pulse_capture_tool}" >&2
      exit 64
      ;;
  esac
else
  case "${remote_pulse_video_mode}" in
    copy)
      run_ffmpeg_capture ffmpeg -hide_banner -loglevel error -y \
        -thread_queue_size 1024 \
        "${video_args[@]}" \
        -i "${resolved_video_device}" \
        -thread_queue_size 1024 \
        -f alsa -ac 2 -ar 48000 \
        -i "${alsa_audio_dev}" \
        -t "${capture_seconds}" \
        -c:v copy \
        -c:a pcm_s16le \
        "${remote_capture}"
      ;;
    cfr)
      run_ffmpeg_capture ffmpeg -hide_banner -loglevel error -y \
        -thread_queue_size 1024 \
        "${video_args[@]}" \
        -i "${resolved_video_device}" \
        -thread_queue_size 1024 \
        -f alsa -ac 2 -ar 48000 \
        -i "${alsa_audio_dev}" \
        -t "${capture_seconds}" \
        -vf "fps=${resolved_video_fps}" \
        -c:v libx264 -preset ultrafast -crf 12 -g 1 -pix_fmt yuv420p \
        -c:a pcm_s16le \
        "${remote_capture}"
      ;;
    *)
      printf 'unsupported REMOTE_PULSE_VIDEO_MODE=%s\n' "${remote_pulse_video_mode}" >&2
      exit 64
      ;;
  esac
fi
REMOTE_CAPTURE_SCRIPT
capture_pid=$!

wait_for_capture_ready() {
  local tries=100
  local i=0
  while (( i < tries )); do
    if [[ -f "${LOCAL_CAPTURE_LOG}" ]] && grep -q "${CAPTURE_READY_MARKER}" "${LOCAL_CAPTURE_LOG}"; then
      return 0
    fi
    if ! kill -0 "${capture_pid}" >/dev/null 2>&1; then
      capture_status=0
      wait "${capture_pid}" || capture_status=$?
      echo "Tethys capture failed before the sync probe could start; see ${LOCAL_CAPTURE_LOG} for details." >&2
      exit "${capture_status}"
    fi
    sleep 0.1
    ((i += 1))
  done

  echo "Timed out waiting for Tethys capture to become ready; see ${LOCAL_CAPTURE_LOG} for details." >&2
  exit 90
}

wait_for_capture_ready

sleep "${LEAD_IN_SECONDS}"

run_output_delay_probe() {
  local label=$1
  local duration_seconds=$2
  local warmup_seconds=$3
  local event_width_codes=$4
  local audio_delay_us=$5
  local video_delay_us=$6
  local timeout_seconds=$7
  local reply_path=$8

  set +e
  (
    cd "${REPO_ROOT}"
    LESAVKA_TLS_DOMAIN="${LESAVKA_TLS_DOMAIN}" \
      timeout --signal=INT "${timeout_seconds}" \
        "${REPO_ROOT}/target/debug/lesavka-relayctl" \
        --server "${RESOLVED_LESAVKA_SERVER_ADDR}" \
        output-delay-probe \
        "${duration_seconds}" \
        "${warmup_seconds}" \
        "${PROBE_PULSE_PERIOD_MS}" \
        "${PROBE_PULSE_WIDTH_MS}" \
        "${event_width_codes}" \
        "${audio_delay_us}" \
        "${video_delay_us}"
  ) 2>&1 | tee "${reply_path}"
  local status=${PIPESTATUS[0]}
  set -e
  if [[ "${status}" -eq 124 ]]; then
    printf '%s output-delay probe timed out after %ss\n' "${label}" "${timeout_seconds}" >&2
  fi
  return "${status}"
}

conditioning_status=0
conditioning_timed_out=0
if [[ "${PROBE_CONDITIONING_SECONDS}" =~ ^[0-9]+$ && "${PROBE_CONDITIONING_SECONDS}" -gt 0 ]]; then
  echo "==> conditioning Tethys UVC/UAC signal path before measured probe"
  echo "   ↪ conditioning duration=${PROBE_CONDITIONING_SECONDS}s warmup=${PROBE_CONDITIONING_WARMUP_SECONDS}s gap=${PROBE_CONDITIONING_GAP_SECONDS}s"
  run_output_delay_probe \
    "signal-conditioning" \
    "${PROBE_CONDITIONING_SECONDS}" \
    "${PROBE_CONDITIONING_WARMUP_SECONDS}" \
    "${PROBE_CONDITIONING_EVENT_WIDTH_CODES}" \
    "${LESAVKA_OUTPUT_DELAY_PROBE_AUDIO_DELAY_US}" \
    "${LESAVKA_OUTPUT_DELAY_PROBE_VIDEO_DELAY_US}" \
    "${PROBE_CONDITIONING_TIMEOUT_SECONDS}" \
    "${LOCAL_SIGNAL_CONDITIONING_REPLY}" || conditioning_status=$?
  if [[ "${conditioning_status}" -eq 124 ]]; then
    conditioning_timed_out=1
  fi
  extract_signal_conditioning_timeline
  if [[ "${conditioning_status}" -ne 0 ]]; then
    echo "signal-conditioning output-delay probe failed with status ${conditioning_status}" >&2
    exit "${conditioning_status}"
  fi
  sleep "${PROBE_CONDITIONING_GAP_SECONDS}"
fi

echo "==> running server-generated UVC/UAC output-delay probe against ${RESOLVED_LESAVKA_SERVER_ADDR}"
probe_status=0
probe_timed_out=0
run_output_delay_probe \
  "server" \
  "${PROBE_DURATION_SECONDS}" \
  "${PROBE_WARMUP_SECONDS}" \
  "${PROBE_EVENT_WIDTH_CODES}" \
  "${LESAVKA_OUTPUT_DELAY_PROBE_AUDIO_DELAY_US}" \
  "${LESAVKA_OUTPUT_DELAY_PROBE_VIDEO_DELAY_US}" \
  "${PROBE_TIMEOUT_SECONDS}" \
  "${LOCAL_SERVER_PROBE_REPLY}" || probe_status=$?
if [[ "${probe_status}" -eq 124 ]]; then
  probe_timed_out=1
fi
extract_server_timeline

capture_status=0
wait "${capture_pid}" || capture_status=$?
capture_v4l2_fault=0
if [[ -f "${LOCAL_CAPTURE_LOG}" ]] \
  && grep -q 'VIDIOC_QBUF): Bad file descriptor' "${LOCAL_CAPTURE_LOG}"; then
  capture_v4l2_fault=1
fi
capture_streamon_timeout=0
if [[ -f "${LOCAL_CAPTURE_LOG}" ]] \
  && grep -q 'VIDIOC_STREAMON.*Connection timed out' "${LOCAL_CAPTURE_LOG}"; then
  capture_streamon_timeout=1
fi

if retry_remote_artifact_command \
  "checking remote capture on ${TETHYS_HOST}" \
  ssh ${SSH_OPTS} "${TETHYS_HOST}" "test -f '${REMOTE_CAPTURE}'"; then
  remote_fetch_capture="${REMOTE_CAPTURE}"
  analysis_status=0
  if [[ "${ANALYSIS_NORMALIZE}" != "0" ]]; then
    remote_fetch_capture="${REMOTE_CAPTURE%.mkv}-analysis.mkv"
    echo "==> normalizing remote capture to CFR for analysis"
    normalize_status=0
    ssh ${SSH_OPTS} "${TETHYS_HOST}" bash -s -- \
      "${REMOTE_CAPTURE}" \
      "${remote_fetch_capture}" \
      "${VIDEO_FPS}" \
      "${ANALYSIS_SCALE_WIDTH}" <<'REMOTE_NORMALIZE_SCRIPT' || normalize_status=$?
set -euo pipefail
src=$1
dst=$2
fps=$3
scale_width=$4
ffmpeg -hide_banner -loglevel error -y \
  -i "${src}" \
  -vf "fps=${fps},scale='min(${scale_width},iw)':-2" \
  -c:v libx264 -preset ultrafast -crf 12 -g 1 -pix_fmt yuv420p \
  -c:a pcm_s16le \
  "${dst}"
REMOTE_NORMALIZE_SCRIPT
    if [[ "${normalize_status}" -ne 0 ]]; then
      echo "remote CFR normalization failed; falling back to raw capture" >&2
      remote_fetch_capture="${REMOTE_CAPTURE}"
    fi
  fi

  if [[ "${REMOTE_ANALYZE}" != "0" ]]; then
    if [[ "${REMOTE_ANALYZE_COPY}" != "0" ]]; then
      echo "==> copying sync analyzer to ${TETHYS_HOST}:${REMOTE_ANALYZE_BIN}"
      retry_remote_artifact_command \
        "copying sync analyzer to ${TETHYS_HOST}" \
        scp ${SSH_OPTS} "${ANALYZE_BIN}" "${TETHYS_HOST}:${REMOTE_ANALYZE_BIN}"
    fi
    analysis_window_arg="$(compute_analysis_window_arg)"
    if [[ -n "${analysis_window_arg}" ]]; then
      echo "   ↪ analyzer timeline window: ${analysis_window_arg#--analysis-window-s }"
    fi
    echo "==> analyzing capture on ${TETHYS_HOST}"
    analysis_tmp="${LOCAL_ANALYSIS_JSON}.tmp"
    analysis_error_tmp="${LOCAL_ANALYSIS_ERROR_LOG}.tmp"
    rm -f "${analysis_tmp}" "${analysis_error_tmp}" "${LOCAL_ANALYSIS_ERROR_LOG}"
    set +e
    retry_remote_artifact_command \
      "running remote sync analysis on ${TETHYS_HOST}" \
      run_remote_sync_analysis_once "${analysis_tmp}" "${analysis_error_tmp}"
    analysis_status=$?
    set -e
    if [[ -s "${analysis_error_tmp}" ]]; then
      cp "${analysis_error_tmp}" "${LOCAL_ANALYSIS_ERROR_LOG}"
    fi
    if [[ "${analysis_status}" -eq 0 ]]; then
      mv "${analysis_tmp}" "${LOCAL_ANALYSIS_JSON}"
      rm -f "${analysis_error_tmp}"
    else
      rm -f "${analysis_tmp}"
      # Keep the .tmp copy too. If this script exits on analysis failure, the
      # readiness summarizer can still parse whichever error artifact survived.
    fi
  fi

  if [[ "${FETCH_CAPTURE}" != "0" ]]; then
    echo "==> fetching capture back to ${LOCAL_CAPTURE}"
    fetch_capture_status=0
    retry_remote_artifact_command \
      "fetching capture from ${TETHYS_HOST}" \
      scp ${SSH_OPTS} "${TETHYS_HOST}:${remote_fetch_capture}" "${LOCAL_CAPTURE}" \
      || fetch_capture_status=$?
    if [[ "${fetch_capture_status}" -ne 0 ]]; then
      echo "warning: failed to fetch capture artifact; continuing with remote analysis JSON when available" >&2
      if [[ "${REMOTE_ANALYZE}" == "0" ]]; then
        exit "${fetch_capture_status}"
      fi
    fi
  fi
  if [[ "${analysis_status}" -ne 0 ]]; then
    echo "remote analysis failed with status ${analysis_status}; capture preserved at ${LOCAL_CAPTURE}" >&2
    exit "${analysis_status}"
  fi
fi

if [[ "${probe_status}" -ne 0 ]]; then
  if [[ "${probe_timed_out}" -eq 1 ]]; then
    echo "server output-delay probe timed out after ${PROBE_TIMEOUT_SECONDS}s; this usually means one UVC/UAC output sink did not close cleanly." >&2
  fi
  echo "server output-delay probe failed with status ${probe_status}" >&2
  [[ -f "${LOCAL_CAPTURE}" ]] && echo "partial capture preserved at ${LOCAL_CAPTURE}" >&2
  exit "${probe_status}"
fi
if [[ "${capture_status}" -ne 0 ]]; then
  if [[ "${capture_streamon_timeout}" -eq 1 ]]; then
    echo "Tethys capture timed out during VIDIOC_STREAMON; the UVC host opened before MJPEG frames reached the gadget." >&2
    echo "Keep LEAD_IN_SECONDS=0 and restart lesavka-uvc/lesavka-server before retrying if the gadget is wedged from an earlier failed run." >&2
  fi
  if [[ "${capture_status}" -eq 141 && ( -f "${LOCAL_CAPTURE}" || -f "${LOCAL_ANALYSIS_JSON}" ) ]]; then
    echo "Tethys capture ended with PipeWire SIGPIPE after ffmpeg closed; accepting preserved analysis artifacts" >&2
  elif [[ "${capture_status}" -eq 124 && ( -f "${LOCAL_CAPTURE}" || -f "${LOCAL_ANALYSIS_JSON}" ) ]]; then
    echo "Tethys capture timed out after preserving analysis artifacts; accepting the run for analysis" >&2
  else
  echo "Tethys capture failed with status ${capture_status}" >&2
  [[ -f "${LOCAL_CAPTURE}" ]] && echo "partial capture preserved at ${LOCAL_CAPTURE}" >&2
  exit "${capture_status}"
  fi
fi

if [[ "${REMOTE_ANALYZE}" != "0" ]]; then
  if [[ ! -f "${LOCAL_ANALYSIS_JSON}" ]]; then
    echo "remote analysis did not produce ${LOCAL_ANALYSIS_JSON}" >&2
    exit 92
  fi
  echo "==> remote analysis summary"
  python - <<'PY' "${LOCAL_ANALYSIS_JSON}" "${LOCAL_REPORT_TXT}" "${LOCAL_EVENTS_CSV}"
import csv
import json
import pathlib
import sys

report = json.loads(pathlib.Path(sys.argv[1]).read_text())
verdict = report.get('verdict', {})
cal = report.get('calibration', {})
lines = [
    f"A/V sync report for {sys.argv[1]}",
    f"- verdict: {verdict.get('status', 'unknown')} ({'pass' if verdict.get('passed') else 'fail'})",
    f"- verdict reason: {verdict.get('reason', '')}",
    f"- p95 abs skew: {float(verdict.get('p95_abs_skew_ms', 0.0)):.1f} ms",
    f"- video onsets: {report['video_event_count']}",
    f"- audio onsets: {report['audio_event_count']}",
    f"- paired pulses: {report['paired_event_count']}",
    f"- activity start delta: {report.get('activity_start_delta_ms', 0.0):+.1f} ms (audio after video is positive)",
    f"- first skew: {report['first_skew_ms']:+.1f} ms (audio after video is positive)",
    f"- last skew: {report['last_skew_ms']:+.1f} ms",
    f"- mean skew: {report['mean_skew_ms']:+.1f} ms",
    f"- median skew: {report['median_skew_ms']:+.1f} ms",
    f"- max abs skew: {report['max_abs_skew_ms']:.1f} ms",
    f"- drift: {report['drift_ms']:+.1f} ms",
    f"- calibration ready: {cal.get('ready')}",
    f"- recommended audio offset adjust: {int(cal.get('recommended_audio_offset_adjust_us', 0)):+d} us",
    f"- alternative video offset adjust: {int(cal.get('recommended_video_offset_adjust_us', 0)):+d} us",
    f"- calibration note: {cal.get('note', '')}",
]
summary = "\n".join(lines) + "\n"
pathlib.Path(sys.argv[2]).write_text(summary)
with pathlib.Path(sys.argv[3]).open("w", newline="") as handle:
    writer = csv.DictWriter(
        handle,
        fieldnames=[
            "event_id",
            "server_event_id",
            "event_code",
            "video_time_s",
            "audio_time_s",
            "skew_ms",
            "confidence",
        ],
    )
    writer.writeheader()
    for event in report.get("paired_events", []):
        writer.writerow({
            "event_id": event.get("event_id"),
            "server_event_id": event.get("server_event_id"),
            "event_code": event.get("event_code"),
            "video_time_s": event.get("video_time_s"),
            "audio_time_s": event.get("audio_time_s"),
            "skew_ms": event.get("skew_ms"),
            "confidence": event.get("confidence"),
        })
print(summary, end="")
PY
else
  if [[ ! -f "${LOCAL_CAPTURE}" ]]; then
    echo "capture was not fetched and REMOTE_ANALYZE=0 left nothing local to analyze" >&2
    exit 93
  fi
  echo "==> analyzing capture"
  analysis_window_arg="$(compute_analysis_window_arg)"
  if [[ -n "${analysis_window_arg}" ]]; then
    echo "   ↪ analyzer timeline window: ${analysis_window_arg#--analysis-window-s }"
  fi
  (
    cd "${REPO_ROOT}"
    # shellcheck disable=SC2086
    "${ANALYZE_BIN}" "${LOCAL_CAPTURE}" --report-dir "${LOCAL_REPORT_DIR}" --event-width-codes "${PROBE_EVENT_WIDTH_CODES}" ${analysis_window_arg}
  )
fi

write_output_delay_correlation
write_output_delay_calibration
maybe_apply_output_delay_calibration
maybe_run_output_delay_confirmation
enforce_sync_verdict

if [[ "${capture_v4l2_fault}" -eq 1 ]]; then
  echo "warning: Tethys video capture reported VIDIOC_QBUF / Bad file descriptor; treat unstable skew or analyzer failures as host-capture suspect" >&2
fi

echo "==> done"
echo "artifact_dir: ${LOCAL_REPORT_DIR}"
if [[ -f "${LOCAL_CAPTURE}" ]]; then
  echo "capture: ${LOCAL_CAPTURE}"
fi
if [[ -f "${LOCAL_ANALYSIS_JSON}" ]]; then
  echo "report_json: ${LOCAL_ANALYSIS_JSON}"
fi
if [[ -f "${LOCAL_ANALYSIS_ERROR_LOG}" ]]; then
  echo "analysis_error_log: ${LOCAL_ANALYSIS_ERROR_LOG}"
fi
if [[ -f "${LOCAL_REPORT_TXT}" ]]; then
  echo "report_txt: ${LOCAL_REPORT_TXT}"
fi
if [[ -f "${LOCAL_EVENTS_CSV}" ]]; then
  echo "events_csv: ${LOCAL_EVENTS_CSV}"
fi
if [[ -f "${LOCAL_SERVER_TIMELINE_JSON}" ]]; then
  echo "server_timeline_json: ${LOCAL_SERVER_TIMELINE_JSON}"
fi
if [[ -f "${LOCAL_SIGNAL_CONDITIONING_REPLY}" ]]; then
  echo "signal_conditioning_reply: ${LOCAL_SIGNAL_CONDITIONING_REPLY}"
fi
if [[ -f "${LOCAL_SIGNAL_CONDITIONING_TIMELINE_JSON}" ]]; then
  echo "signal_conditioning_timeline_json: ${LOCAL_SIGNAL_CONDITIONING_TIMELINE_JSON}"
fi
if [[ -f "${LOCAL_CLOCK_ALIGNMENT_JSON}" ]]; then
  echo "clock_alignment_json: ${LOCAL_CLOCK_ALIGNMENT_JSON}"
fi
if [[ -f "${LOCAL_OUTPUT_DELAY_CORRELATION_JSON}" ]]; then
  echo "output_delay_correlation_json: ${LOCAL_OUTPUT_DELAY_CORRELATION_JSON}"
fi
if [[ -f "${LOCAL_OUTPUT_DELAY_CORRELATION_CSV}" ]]; then
  echo "output_delay_correlation_csv: ${LOCAL_OUTPUT_DELAY_CORRELATION_CSV}"
fi
if [[ -f "${LOCAL_OUTPUT_DELAY_CORRELATION_TXT}" ]]; then
  echo "output_delay_correlation_txt: ${LOCAL_OUTPUT_DELAY_CORRELATION_TXT}"
fi
if [[ -f "${LOCAL_OUTPUT_DELAY_JSON}" ]]; then
  echo "output_delay_calibration_json: ${LOCAL_OUTPUT_DELAY_JSON}"
fi
if [[ -f "${LOCAL_OUTPUT_DELAY_ENV}" ]]; then
  echo "output_delay_calibration_env: ${LOCAL_OUTPUT_DELAY_ENV}"
fi
if [[ -f "${LOCAL_CAPTURE_LOG}" ]]; then
  echo "capture_log: ${LOCAL_CAPTURE_LOG}"
fi