From 00f9b4f0ce3122020de4f2db459187705243eb50 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E9=82=B9=E6=B0=B8=E8=B5=AB?= <1259085392@qq.com>
Date: Thu, 26 Mar 2026 02:10:43 +0900
Subject: [PATCH] fix: observe lsp reader task failures

---
 astrbot/_internal/protocols/lsp/client.py     | 19 +++++++++
 .../fixtures/hanging_lsp_server.py            | 13 ++++++
 tests/integration/test_lsp_integration.py     | 21 ++++++++++
 tests/unit/test_internal/test_lsp_client.py   | 41 +++++++++++++++++++
 4 files changed, 94 insertions(+)
 create mode 100644 tests/integration/fixtures/hanging_lsp_server.py
 create mode 100644 tests/unit/test_internal/test_lsp_client.py

diff --git a/astrbot/_internal/protocols/lsp/client.py b/astrbot/_internal/protocols/lsp/client.py
index 18d0dc265..b529233a2 100644
--- a/astrbot/_internal/protocols/lsp/client.py
+++ b/astrbot/_internal/protocols/lsp/client.py
@@ -43,6 +43,24 @@ class AstrbotLspClient(BaseAstrbotLspClient):
         """True if connected to an LSP server."""
         return self._connected
 
+    def _handle_reader_task_done(self, task: asyncio.Task[None]) -> None:
+        try:
+            exception = task.exception()
+        except asyncio.CancelledError:
+            return
+
+        if exception is None:
+            if self._connected:
+                self._connected = False
+                log.warning("LSP reader task exited unexpectedly")
+            return
+
+        self._connected = False
+        log.error(
+            "LSP reader task failed",
+            exc_info=(type(exception), exception, exception.__traceback__),
+        )
+
     async def connect(self) -> None:
         """
         Connect to configured LSP servers.
@@ -79,6 +97,7 @@ class AstrbotLspClient(BaseAstrbotLspClient):
 
         # Start reading responses in the background.
         self._reader_task = asyncio.create_task(self._read_responses())
+        self._reader_task.add_done_callback(self._handle_reader_task_done)
 
         # Send initialize request
         await self.send_request(
diff --git a/tests/integration/fixtures/hanging_lsp_server.py b/tests/integration/fixtures/hanging_lsp_server.py
new file mode 100644
index 000000000..46a862dca
--- /dev/null
+++ b/tests/integration/fixtures/hanging_lsp_server.py
@@ -0,0 +1,13 @@
+"""LSP fixture that never answers initialization."""
+
+from __future__ import annotations
+
+import time
+
+
+def main() -> None:
+    time.sleep(60)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/tests/integration/test_lsp_integration.py b/tests/integration/test_lsp_integration.py
index 8716227da..6010a1a55 100644
--- a/tests/integration/test_lsp_integration.py
+++ b/tests/integration/test_lsp_integration.py
@@ -11,11 +11,13 @@ from pathlib import Path
 
 import anyio
 import pytest
+from anyio.lowlevel import checkpoint
 
 from astrbot._internal.protocols.lsp.client import AstrbotLspClient
 
 TEST_DIR = Path(__file__).resolve().parent
 SERVER_PATH = TEST_DIR / "fixtures" / "echo_lsp_server.py"
+HANGING_SERVER_PATH = TEST_DIR / "fixtures" / "hanging_lsp_server.py"
 
 
 @pytest.mark.anyio
@@ -115,3 +117,22 @@ async def test_lsp_client_connect_does_not_corrupt_anyio_cancel_scope():
     finally:
         with anyio.fail_after(5):
             await client.shutdown()
+
+
+@pytest.mark.anyio
+async def test_lsp_client_connect_timeout_does_not_corrupt_anyio_cancel_scope():
+    """Test timeout-driven cancellation leaves later fail_after scopes usable."""
+    client = AstrbotLspClient()
+
+    with pytest.raises(TimeoutError):
+        with anyio.fail_after(0.1):
+            await client.connect_to_server(
+                command=[sys.executable, str(HANGING_SERVER_PATH)],
+                workspace_uri="file:///tmp",
+            )
+
+    with anyio.fail_after(5):
+        await client.shutdown()
+
+    with anyio.fail_after(1):
+        await checkpoint()
diff --git a/tests/unit/test_internal/test_lsp_client.py b/tests/unit/test_internal/test_lsp_client.py
new file mode 100644
index 000000000..4524a08a9
--- /dev/null
+++ b/tests/unit/test_internal/test_lsp_client.py
@@ -0,0 +1,41 @@
+from __future__ import annotations
+
+import asyncio
+from types import SimpleNamespace
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from astrbot._internal.protocols.lsp.client import AstrbotLspClient
+
+
+@pytest.mark.asyncio
+async def test_lsp_reader_task_failure_marks_client_disconnected_and_logs():
+    """Test reader task failures update connection state and are logged."""
+    client = AstrbotLspClient()
+    fake_process = SimpleNamespace(stdout=MagicMock(), stdin=MagicMock())
+
+    with (
+        patch(
+            "astrbot._internal.protocols.lsp.client.anyio.open_process",
+            AsyncMock(return_value=fake_process),
+        ),
+        patch.object(
+            client,
+            "_read_responses",
+            AsyncMock(side_effect=RuntimeError("reader crashed")),
+        ),
+        patch.object(client, "send_request", AsyncMock(return_value={})),
+        patch.object(client, "send_notification", AsyncMock()),
+        patch("astrbot._internal.protocols.lsp.client.log") as mock_log,
+    ):
+        await client.connect_to_server(["python", "fake_lsp.py"], "file:///tmp")
+        await asyncio.sleep(0)
+
+        reader_task = client._reader_task
+        assert reader_task is not None
+        _ = reader_task.exception()
+        await asyncio.sleep(0)
+
+        assert client.connected is False
+        mock_log.error.assert_called_once()