From 335003e7cbbe5e9596d9c7004a2ac0375043ddad Mon Sep 17 00:00:00 2001
From: Sebastien Tardif <sebtardif@ncf.ca>
Date: Wed, 2 Sep 2026 10:01:19 -0700
Subject: [PATCH] fix: time out unbounded lane prepare hook

Lane prepare used subprocess.run with no timeout. A stuck
CLAWBENCH_LANE_PREPARE_CMD blocked lane startup and the
EvalWorker. Bound the hook to CLAWBENCH_LANE_PREPARE_TIMEOUT_SECONDS
(default 180, matching gateway health) and fail the lane on expiry.

Signed-off-by: Sebastien Tardif <sebtardif@ncf.ca>
---
 clawbench/worker.py  |  8 ++++++-
 tests/test_worker.py | 57 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 64 insertions(+), 1 deletion(-)

diff --git a/clawbench/worker.py b/clawbench/worker.py
index 295c9e2..a0dcfd9 100644
--- a/clawbench/worker.py
+++ b/clawbench/worker.py
@@ -629,7 +629,13 @@ def _run_lane_prepare_hook(self, lane: ParallelLane) -> None:
             "CLAWBENCH_LANE_PORT": str(lane.port),
         }
         logger.info("Running lane %d prepare hook", lane.index + 1)
-        subprocess.run([hook], env=hook_env, check=True)
+        timeout_seconds = int(os.environ.get("CLAWBENCH_LANE_PREPARE_TIMEOUT_SECONDS", "180"))
+        try:
+            subprocess.run([hook], env=hook_env, check=True, timeout=timeout_seconds)
+        except subprocess.TimeoutExpired as exc:
+            raise RuntimeError(
+                f"Lane {lane.index + 1} prepare hook timed out after {timeout_seconds}s"
+            ) from exc
 
     def _seed_lane_state_dir(self, target_state_dir: Path) -> None:
         source_state_dir = Path(os.environ.get("OPENCLAW_STATE_DIR", os.path.expanduser("~/.openclaw")))
diff --git a/tests/test_worker.py b/tests/test_worker.py
index 523742f..0b83a2e 100644
--- a/tests/test_worker.py
+++ b/tests/test_worker.py
@@ -1,5 +1,9 @@
 import asyncio
 import json
+import os
+import signal
+import subprocess
+import threading
 from pathlib import Path
 from types import SimpleNamespace
 
@@ -452,3 +456,56 @@ async def fake_update_progress(job_id: str, **kwargs) -> None:
             },
         )
     ]
+
+
+def test_run_lane_prepare_hook_kills_hung_hook(tmp_path: Path, monkeypatch):
+    hook = tmp_path / "hung-hook"
+    hook.write_text(
+        '#!/bin/sh\necho $$ > "$HOME/hook.pid"\nexec sleep 1000\n',
+        encoding="utf-8",
+    )
+    hook.chmod(0o755)
+
+    state_dir = tmp_path / "lane" / "state"
+    state_dir.mkdir(parents=True)
+    (state_dir.parent / "home").mkdir(parents=True)
+    pid_path = state_dir.parent / "home" / "hook.pid"
+
+    monkeypatch.setenv("CLAWBENCH_LANE_PREPARE_CMD", str(hook))
+    monkeypatch.setenv("CLAWBENCH_LANE_PREPARE_TIMEOUT_SECONDS", "1")
+
+    worker = EvalWorker(JobQueue())
+    lane = ParallelLane(index=0, tasks=[DummyTask("t1", "tier1", "coding")])
+    lane.state_dir = state_dir
+    lane.port = GATEWAY_PORT
+
+    outcome: list[BaseException | None] = []
+
+    def run() -> None:
+        try:
+            worker._run_lane_prepare_hook(lane)
+            outcome.append(None)
+        except BaseException as exc:
+            outcome.append(exc)
+
+    thread = threading.Thread(target=run, daemon=True)
+    try:
+        thread.start()
+        thread.join(5)
+        assert not thread.is_alive(), "prepare hook hung without timeout"
+        assert outcome
+        exc = outcome[0]
+        assert exc is not None
+        assert isinstance(exc, (subprocess.TimeoutExpired, RuntimeError))
+        if isinstance(exc, RuntimeError):
+            assert "timed out" in str(exc).lower()
+        assert pid_path.is_file()
+        pid = int(pid_path.read_text(encoding="utf-8").strip())
+        with pytest.raises(ProcessLookupError):
+            os.kill(pid, 0)
+    finally:
+        if pid_path.is_file():
+            try:
+                os.kill(int(pid_path.read_text(encoding="utf-8").strip()), signal.SIGKILL)
+            except (ValueError, ProcessLookupError, OSError):
+                pass