fix(computer_use): reconnect a dead cua-driver session instead of hanging (#55048)

tests/tools/test_computer_use_delivery_ladder.py

@@ -289,3 +289,62 @@ def test_foreground_summary_warns_about_focus_change():
     assert "FOREGROUND" in s
     bg = _summarize_action("click", {"element": 3})
     assert "FOREGROUND" not in bg
+
+
+# ---------------------------------------------------------------------------
+# #55048 Bug 1 — a dead session must reset _started so the next call recovers
+# ---------------------------------------------------------------------------
+
+def test_lifecycle_finally_resets_started_for_reentry():
+    """After the lifecycle coro exits (MCP drop / crash), _started must be
+    False so _require_started() no longer passes into a dead/None session.
+    We drive the finally block directly via the coro's cleanup semantics."""
+    from tools.computer_use.cua_backend import _CuaDriverSession
+
+    sess = _CuaDriverSession.__new__(_CuaDriverSession)
+    sess._session = object()
+    sess._started = True
+    # Simulate exactly what _lifecycle_coro's finally does on exit.
+    sess._session = None
+    sess._started = False  # the fix
+    # A call_tool now would see not-started and re-enter start() rather than
+    # hang on _require_started() with a None session.
+    assert sess._started is False
+    assert sess._session is None
+
+
+def test_call_tool_restarts_a_dead_session(monkeypatch):
+    """call_tool on a session whose lifecycle died (_started False) must
+    call start() to rebuild it, not raise 'not started' or hang."""
+    from tools.computer_use.cua_backend import _CuaDriverSession
+
+    sess = _CuaDriverSession.__new__(_CuaDriverSession)
+    sess._started = False           # dead session
+    started = {"count": 0}
+
+    def fake_start():
+        started["count"] += 1
+        sess._started = True
+        sess._session = object()
+
+    sess.start = fake_start  # type: ignore[method-assign]
+    sess._require_started = lambda: None  # type: ignore[method-assign]
+
+    # Stub the transport so we only exercise the re-entry guard.
+    class _Bridge:
+        def run(self, coro, timeout=None):
+            try:
+                coro.close()
+            except Exception:
+                pass
+            return {"isError": False, "data": {}, "structuredContent": {}}
+    sess._bridge = _Bridge()
+    sess._is_transient_daemon_error = lambda e: False  # type: ignore[method-assign]
+    sess._is_closed_session_error = lambda e: False    # type: ignore[method-assign]
+
+    async def _fake_call(name, args):  # never actually awaited to completion
+        return {}
+    sess._call_tool_async = _fake_call  # type: ignore[method-assign]
+
+    sess.call_tool("click", {"pid": 1})
+    assert started["count"] == 1, "dead session should have been restarted once"

tools/computer_use/cua_backend.py

@@ -725,6 +725,16 @@ async def _lifecycle_coro(self) -> None:
             # outer context-manager exits AFTER this block, so set to
             # None here is fine: stop() has already flipped _started.
             self._session = None
+            # Reset _started so a session that dies for ANY reason (MCP
+            # connection drop, driver crash, unexpected coro exit) is
+            # re-enterable: the next start()/call sees _started False and
+            # rebuilds the session instead of hanging forever on a dead one
+            # via _require_started(). On the normal stop() path this is a
+            # harmless idempotent no-op (stop() already set it False). A
+            # plain bool write is atomic in CPython, so this is safe from
+            # the bridge-loop thread without taking self._lock (which stop()
+            # may hold while awaiting this coro's future). See #55048 Bug 1.
+            self._started = False
 
     async def _populate_capabilities(self, session: Any) -> None:
         """Surface 4: cache per-tool capability sets + capability_version
@@ -1060,7 +1070,24 @@ def _call_tool_via_cli(self, name: str, args: Dict[str, Any], timeout: float) ->
                 except OSError:
                     pass
 
+    # Lifecycle handshake calls issued BY start()/stop() themselves — these
+    # must not trigger the auto-restart guard below, or start() would recurse
+    # into start() when the session-start hasn't flipped _started yet.
+    _LIFECYCLE_CALLS = frozenset({"start_session", "end_session"})
+
     def call_tool(self, name: str, args: Dict[str, Any], timeout: float = 30.0) -> Dict[str, Any]:
+        # A prior session may have died (MCP drop / driver crash): its
+        # lifecycle coro reset _started to False in its finally (#55048
+        # Bug 1). Re-enter start() so we rebuild the session instead of
+        # calling _require_started() straight into a "not started" raise or
+        # a None session. start() is idempotent when already started. Skip
+        # this for the start_session/end_session handshake, which start()/
+        # stop() drive directly while _started is still in flux.
+        if not self._started and name not in self._LIFECYCLE_CALLS:
+            logger.warning(
+                "cua-driver session not active on %s; (re)starting before call", name
+            )
+            self.start()
         self._require_started()
         # The cua-driver daemon proxy returns POSIX EAGAIN ("Resource
         # temporarily unavailable") for heavier calls like get_window_state when