From 526a0015fbe1fe8cc9870263909435ce3b59534b Mon Sep 17 00:00:00 2001
From: chaochen <orangercc@gmail.com>
Date: Fri, 11 Sep 2026 00:19:21 +0800
Subject: [PATCH] fix(decode): tolerate non-UTF-8 child output in two remaining
 capture paths

Same class as the cron script runner: text=True decodes with the locale codec under strict error handling, so one undecodable byte in a child's output raises UnicodeDecodeError inside subprocess.run. That is a ValueError - neither except (subprocess.SubprocessError, OSError) nor except TimeoutExpired catches it - so it escaped into user-visible paths instead of the intended error handling.

tools/bot_mode_dm._run_local_turn: a transport that exits 0 but prints a byte the locale cannot decode crashed the delivery instead of re-emitting the transport's streams (stdout is the reply text the completion notification carries back).

agent.command_token_source._mint: a key_cmd printing a non-UTF-8 byte killed token minting with a traceback instead of the documented CommandTokenError path, so provider auth failed in a way the caller could not report.

Both decode with errors=replace now: the exit code and the surrounding checks still decide what happens, and damaged bytes appear as U+FFFD in the text we carry.
---
 agent/command_token_source.py            |  2 +-
 tests/agent/test_command_token_source.py | 12 ++++++++++++
 tests/tools/test_bot_mode_dm.py          | 15 +++++++++++++++
 tools/bot_mode_dm.py                     |  2 +-
 4 files changed, 29 insertions(+), 2 deletions(-)

diff --git a/agent/command_token_source.py b/agent/command_token_source.py
index 38ace24324..08c6d04763 100644
--- a/agent/command_token_source.py
+++ b/agent/command_token_source.py
@@ -51,7 +51,7 @@ def _mint(command: str, label: str) -> tuple[str, Optional[float]]:
 
     try:
         completed = subprocess.run(
-            command, shell=True, capture_output=True, text=True, timeout=_MINT_TIMEOUT_SECONDS,
+            command, shell=True, capture_output=True, text=True, errors="replace", timeout=_MINT_TIMEOUT_SECONDS,
             env=served_profile_child_env(inherit_credentials=True),
         )
     except subprocess.TimeoutExpired as exc:
diff --git a/tests/agent/test_command_token_source.py b/tests/agent/test_command_token_source.py
index 71251e9963..78bd86ec1f 100644
--- a/tests/agent/test_command_token_source.py
+++ b/tests/agent/test_command_token_source.py
@@ -77,6 +77,18 @@ class TestMinting:
         assert "dbx" in message          # names the provider to fix
         assert "exited" in message       # states what happened
 
+    def test_undecodable_output_does_not_raise(self):
+        """One non-UTF-8 byte from the helper must not crash token minting.
+
+        A strict decode raised UnicodeDecodeError inside subprocess.run — a ValueError,
+        so neither the TimeoutExpired nor the OSError handler above caught it: provider
+        auth died with a traceback instead of the documented CommandTokenError path.
+        """
+        token, ttl = _mint("printf 'tok\\377'", "dbx")
+
+        assert ttl is None
+        assert token.startswith("tok")
+
 
 class TestNoCredentialLeak:
     def test_failure_message_excludes_command_output(self):
diff --git a/tests/tools/test_bot_mode_dm.py b/tests/tools/test_bot_mode_dm.py
index d068c4f12e..2632cede2f 100644
--- a/tests/tools/test_bot_mode_dm.py
+++ b/tests/tools/test_bot_mode_dm.py
@@ -1102,3 +1102,18 @@ def test_poll_reply_is_persisted_as_a_delivery_row_when_the_runner_exits(tmp_pat
     assert kw["display_kind"] == "process_complete"
     assert "PAYLOAD_SENTINEL_42" in kw["content"]
     assert procs[0].id in kw["content"]
+
+
+def test_local_turn_survives_undecodable_transport_output(tmp_path, capsys):
+    """A transport that exits 0 while printing a non-UTF-8 byte must still deliver.
+
+    A strict decode raised UnicodeDecodeError inside subprocess.run — a ValueError, so no
+    handler caught it and the delivery crashed instead of re-emitting the transport's
+    streams (stdout is the reply text the completion notification carries back).
+    """
+    dm_file = tmp_path / "dm.txt"
+    dm_file.write_text("hello", encoding="utf-8")
+    argv = [sys.executable, "-c", "import sys; sys.stdout.buffer.write(b'reply \\377')"]
+
+    assert bot_mode_dm._run_local_turn(argv, str(dm_file)) == 0
+    assert "reply" in capsys.readouterr().out
diff --git a/tools/bot_mode_dm.py b/tools/bot_mode_dm.py
index c6aebeb965..090061d3ba 100644
--- a/tools/bot_mode_dm.py
+++ b/tools/bot_mode_dm.py
@@ -416,7 +416,7 @@ def _run_local_turn(argv: list[str], dm_file: str, *, env: Optional[dict[str, st
 
     def _turn(turn_env=env):
         return subprocess.run([*argv, "--query-file", dm_file], check=False, stdin=subprocess.DEVNULL,
-                              capture_output=True, text=True, env=turn_env)
+                              capture_output=True, text=True, errors="replace", env=turn_env)
 
     proc = _turn()
     if proc.returncode != 0: