From d4e0294734afc631cf1caff89ff992d46af4b941 Mon Sep 17 00:00:00 2001
From: chengyongru <2755839590@qq.com>
Date: Wed, 15 Jul 2026 21:44:09 +0800
Subject: [PATCH] fix(gateway): stop channels before draining tasks

---
 nanobot/cli/commands.py    | 4 +++-
 tests/cli/test_commands.py | 6 ++++++
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/nanobot/cli/commands.py b/nanobot/cli/commands.py
index ae0714bf..5d4dc4f9 100644
--- a/nanobot/cli/commands.py
+++ b/nanobot/cli/commands.py
@@ -2119,6 +2119,9 @@ def _run_gateway(
                         await shutdown_task
                 cron.stop()
                 agent.stop()
+                # Some SDKs swallow task cancellation while attempting to reconnect.
+                # Close channel transports before waiting for their runners to exit.
+                await channels.stop_all()
                 for task in tasks:
                     if not task.done():
                         task.cancel()
@@ -2127,7 +2130,6 @@ def _run_gateway(
                 if runtime_tasks is not None and not runtime_tasks_drained:
                     with suppress(asyncio.CancelledError, Exception):
                         await runtime_tasks
-                await channels.stop_all()
                 # Flush all cached sessions to durable storage before exit.
                 # This prevents data loss on filesystems with write-back
                 # caching (rclone VFS, NFS, FUSE mounts, etc.).
diff --git a/tests/cli/test_commands.py b/tests/cli/test_commands.py
index 0808dd93..3497df73 100644
--- a/tests/cli/test_commands.py
+++ b/tests/cli/test_commands.py
@@ -3235,6 +3235,7 @@ def test_gateway_shutdown_event_exits_forever_runtime_tasks(
     config = Config()
     config.gateway.port = 18791
     seen: dict[str, object] = {}
+    shutdown_order: list[str] = []
 
     class _FakeSessionManager:
         def flush_all(self) -> int:
@@ -3274,9 +3275,11 @@ def test_gateway_shutdown_event_exits_forever_runtime_tasks(
                 await asyncio.Event().wait()
             finally:
                 seen["channel_task_cleaned_up"] = True
+                shutdown_order.append("channel_task_cleaned_up")
 
         async def stop_all(self) -> None:
             seen["channels_stopped"] = True
+            shutdown_order.append("channels_stopped")
 
     class _FakeCronService:
         def __init__(self, _store_path: Path) -> None:
@@ -3343,6 +3346,9 @@ def test_gateway_shutdown_event_exits_forever_runtime_tasks(
     assert seen["channels_stopped"] is True
     assert seen["cron_stopped"] is True
     assert seen["shutdown_handlers_restored"] is True
+    # Channel cleanup must run before cancellation drains the manager task.
+    # DingTalk's stream SDK can otherwise swallow cancellation and reconnect.
+    assert shutdown_order == ["channels_stopped", "channel_task_cleaned_up"]
 
 
 def test_serve_uses_api_config_defaults_and_workspace_override(