From d22fa746fd0a3ed4c834bf9f95f57db490db8091 Mon Sep 17 00:00:00 2001
From: jxxghp <jxxghp@gmail.com>
Date: Fri, 11 Sep 2026 18:57:03 +0800
Subject: [PATCH] fix(agent): ignore expected app-server reader cancellation

---
 scripts/evaluation/codex.py          |  2 +-
 tests/test_agent_evaluation_codex.py | 10 +++++++++-
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/scripts/evaluation/codex.py b/scripts/evaluation/codex.py
index f9b3b55b8..f31de08c1 100644
--- a/scripts/evaluation/codex.py
+++ b/scripts/evaluation/codex.py
@@ -364,7 +364,7 @@ async def _execute_app_server(
         if not stderr_reader.done():
             stderr_reader.cancel()
         reader_result = (await asyncio.gather(stderr_reader, return_exceptions=True))[0]
-        if isinstance(reader_result, BaseException):
+        if isinstance(reader_result, BaseException) and not isinstance(reader_result, asyncio.CancelledError):
             failure = failure or type(reader_result).__name__
     return {
         "returncode": process.returncode,
diff --git a/tests/test_agent_evaluation_codex.py b/tests/test_agent_evaluation_codex.py
index 6c0963cf3..4bd7aa16b 100644
--- a/tests/test_agent_evaluation_codex.py
+++ b/tests/test_agent_evaluation_codex.py
@@ -71,8 +71,16 @@ async def test_execute_sends_exact_prompt_and_preserves_nonzero_exit_output(tmp_
 
 
 @pytest.mark.asyncio
-async def test_execute_app_server_preserves_streamed_terminal_evidence(tmp_path: Path) -> None:
+async def test_execute_app_server_preserves_streamed_terminal_evidence(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
     """app-server 的终端增量和同一进程输入事件应归一化为可评分的原生轨迹。"""
+    async def hold_stderr_reader(stream: asyncio.StreamReader, result: bytearray) -> None:
+        """让测试覆盖正常收尾时主动取消 stderr 读取任务的路径。"""
+        del stream, result
+        await asyncio.Future()
+
+    monkeypatch.setattr(codex, "_read_output", hold_stderr_reader)
     command = _program(tmp_path, """
 import json
 import sys