fix(flow): retain terminal evidence across interrupted runs

Record parent and child process IDs for diagnostics, persist interruption
before cleanup, and ignore repeated termination signals during cleanup.
Close unexpected failures with a terminal result and retain the child's
exit status without treating namespace-local PID absence as completion.

Require a successful terminal run_flow result when checking simulation
logs with an associated result file. Cover repeated cancellation and
reject running, failed or inconsistent terminal records despite PASS text.

This supports the TINY-R2-P7 campaign. Interrupted and timed-out runs remain
non-acceptance evidence; UART output alone cannot qualify a Tiny simulation.

scripts/check_simulation.py

@@ -39,6 +39,19 @@ def main() -> int:
     if default_failure:
         rejected.append(default_failure.group(0))
 
+    flow_result_path = args.result.with_name("result-sim.json")
+    flow_status = None
+    if flow_result_path.is_file():
+        try:
+            flow_result = json.loads(flow_result_path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError):
+            flow_result = {}
+        flow_status = flow_result.get("status", "invalid")
+        # PIDs are namespace-local and can be reused. A durable terminal
+        # verdict, not a liveness guess from this process, establishes success.
+        if flow_status != "passed" or flow_result.get("exit_code") != 0:
+            missing.append("successful terminal run_flow result")
+
     passed = not missing and not rejected
     report = {
         "schema_version": 1,
@@ -48,6 +61,8 @@ def main() -> int:
         "missing_markers": missing,
         "rejected_matches": sorted(set(rejected)),
     }
+    if flow_status is not None:
+        report["flow_status"] = flow_status
     atomic_write(args.result, json.dumps(report, indent=2, sort_keys=True) + "\n")
     print(f"simulation check: {report['status']} ({args.result.resolve()})")
     return int(not passed)

scripts/run_flow.py

@@ -124,24 +124,22 @@ def main() -> int:
     args.result.parent.mkdir(parents=True, exist_ok=True)
     started = datetime.now(timezone.utc)
     start_clock = time.monotonic()
-    atomic_write(
-        args.result,
-        json.dumps(
-            {
+    running_result = {
         "schema_version": 1,
         "tool": args.tool,
         "command": command,
         "started_at": started.isoformat(),
         "status": "running",
+        "parent_pid": os.getpid(),
         "log": str(args.log.resolve()),
-            },
-            indent=2,
-            sort_keys=True,
-        )
-        + "\n",
-    )
+    }
+    atomic_write(args.result, json.dumps(running_result, indent=2, sort_keys=True) + "\n")
 
     def interrupt_flow(_signum: int, _frame: object) -> None:
+        # A launcher can deliver SIGINT followed by SIGTERM. Do not let a
+        # second signal interrupt process-group cleanup or result persistence.
+        signal.signal(signal.SIGINT, signal.SIG_IGN)
+        signal.signal(signal.SIGTERM, signal.SIG_IGN)
         raise KeyboardInterrupt
 
     signal.signal(signal.SIGINT, interrupt_flow)
@@ -169,6 +167,8 @@ def main() -> int:
                 pass_fds=jobserver_fds(),
                 start_new_session=True,
             )
+            running_result["child_pid"] = process.pid
+            atomic_write(args.result, json.dumps(running_result, indent=2, sort_keys=True) + "\n")
             assert process.stdout is not None
             if args.stream_bytes:
                 while chunk := os.read(process.stdout.fileno(), 4096):
@@ -193,11 +193,25 @@ def main() -> int:
                 returncode = process.wait()
     except KeyboardInterrupt:
         error_message = "interrupted"
+        # Persist cancellation before the bounded cleanup wait. If the outer
+        # executor subsequently kills this wrapper, the file must not imply
+        # that the job is still running or that its child finished normally.
+        cancelled = dict(running_result, status="interrupted", exit_code=130,
+                         interrupted_at=datetime.now(timezone.utc).isoformat(),
+                         child_cleanup_pending=process is not None)
+        atomic_write(args.result, json.dumps(cancelled, indent=2, sort_keys=True) + "\n")
         if process is not None:
             terminate_process_group(process)
         returncode = 130
     except (OSError, subprocess.SubprocessError) as error:
         error_message = str(error)
+        if process is not None:
+            terminate_process_group(process)
+    except BaseException as error:
+        error_message = f"unexpected {type(error).__name__}: {error}"
+        if process is not None:
+            terminate_process_group(process)
+        returncode = 1
 
     finished = datetime.now(timezone.utc)
     result = {
@@ -211,8 +225,13 @@ def main() -> int:
         "peak_rss_kib": resource.getrusage(resource.RUSAGE_CHILDREN).ru_maxrss,
         "exit_code": returncode,
         "status": "interrupted" if returncode == 130 else ("passed" if returncode == 0 else "failed"),
+        "parent_pid": os.getpid(),
         "log": str(args.log.resolve()),
     }
+    if process is not None:
+        result["child_pid"] = process.pid
+        result["child_exit_code"] = process.returncode
+        result["child_cleanup_pending"] = process.returncode is None
     if error_message:
         result["error"] = error_message
     if args.success_marker:

tests/test_script_tools.py

@@ -1969,6 +1969,8 @@ def test_run_flow_writes_structured_result(tmp_path: Path) -> None:
     assert data["status"] == "passed"
     assert data["exit_code"] == 0
     assert data["duration_seconds"] >= 0
+    assert isinstance(data["parent_pid"], int)
+    assert isinstance(data["child_pid"], int)
     assert isinstance(data["peak_rss_kib"], int)
     assert data["peak_rss_kib"] > 0
     assert log.read_text(encoding="utf-8") == "flow output\n"
@@ -2120,6 +2122,41 @@ def test_run_flow_streams_bytes_without_waiting_for_newline(tmp_path: Path) -> N
     assert json.loads(result.read_text(encoding="utf-8"))["status"] == "passed"
 
 
+def test_run_flow_persists_cancel_before_cleanup_and_ignores_second_signal(tmp_path: Path) -> None:
+    result = tmp_path / "cancelled.json"
+    process = subprocess.Popen([
+        sys.executable, str(ROOT / "scripts/run_flow.py"), "--tool", "unit",
+        "--log", str(tmp_path / "cancelled.log"), "--result", str(result), "--",
+        sys.executable, "-u", "-c",
+        "import signal,time; signal.signal(signal.SIGTERM,signal.SIG_IGN); "
+        "print('READY',flush=True); time.sleep(30)",
+    ], text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    try:
+        assert process.stdout is not None
+        assert process.stdout.readline().strip() == "READY"
+        process.send_signal(signal.SIGINT)
+        for _ in range(100):
+            checkpoint = json.loads(result.read_text())
+            if checkpoint["status"] == "interrupted":
+                break
+            time.sleep(0.01)
+        assert checkpoint["status"] == "interrupted"
+        assert checkpoint["child_cleanup_pending"] is True
+        assert "finished_at" not in checkpoint
+        process.send_signal(signal.SIGTERM)
+        _, stderr = process.communicate(timeout=10)
+        assert process.returncode == 130, stderr
+        final = json.loads(result.read_text())
+        assert final["status"] == "interrupted"
+        assert final["child_cleanup_pending"] is False
+        assert final["child_exit_code"] == -signal.SIGKILL
+        assert final["finished_at"] >= final["started_at"]
+    finally:
+        if process.poll() is None:
+            process.kill()
+            process.wait(timeout=5)
+
+
 def test_run_flow_records_interruption(tmp_path: Path) -> None:
     log = tmp_path / "interrupted.log"
     result = tmp_path / "interrupted.json"
@@ -2192,6 +2229,27 @@ def test_simulation_success_marker_and_failure_detection(tmp_path: Path) -> None
     assert json.loads(result.read_text(encoding="utf-8"))["status"] == "failed"
 
 
+@pytest.mark.parametrize("status,exit_code", [("running", None), ("failed", 1),
+                                              ("passed", 124), ("invalid", None)])
+def test_check_simulation_requires_terminal_flow(tmp_path: Path, status: str,
+                                                exit_code: int | None) -> None:
+    log = tmp_path / "sim.log"
+    result = tmp_path / "result-sim-check.json"
+    log.write_text("SIM_TEST_PASS code=0\n", encoding="utf-8")
+    (tmp_path / "result-sim.json").write_text(
+        json.dumps({"status": status, "exit_code": exit_code}),
+        encoding="utf-8",
+    )
+    checked = subprocess.run(
+        [sys.executable, str(ROOT / "scripts/check_simulation.py"), "--log", str(log),
+         "--result", str(result)], text=True, capture_output=True,
+    )
+    report = json.loads(result.read_text(encoding="utf-8"))
+    assert checked.returncode != 0
+    assert report["status"] == "failed"
+    assert report["flow_status"] == status
+
+
 def test_warning_baseline_rejects_new_signature(tmp_path: Path) -> None:
     profile = "unit-profile"
     log = tmp_path / "variant/sim/verilator/verilating.log"