Allow multiple calls to loop.stop

Calling loop.stop and then immediately proceeding to exit the
`open_loop`'s context (without hitting any checkpoints) causes the
program to hang forever

Fix by reusing the same `waiter` event across multiple calls to `stop`

Fixes #58

Author: tjstum

tests/test_misc.py

@@ -26,6 +26,15 @@ def close_no_stop():
                 await trio.sleep(0.1)
                 await loop.wait_closed()
 
+    @pytest.mark.trio
+    async def test_too_many_stops(self):
+        with trio.move_on_after(1) as scope:
+            async with trio_asyncio.open_loop() as loop:
+                await trio.hazmat.checkpoint()
+                loop.stop()
+        assert not scope.cancelled_caught, \
+            "Possible deadlock after manual call to loop.stop"
+
     @pytest.mark.trio
     async def test_err1(self, loop):
         async def raise_err():

trio_asyncio/async_.py

@@ -60,7 +60,9 @@ def stop(self, waiter=None):
 
         """
         if waiter is None:
-            waiter = trio.Event()
+            if self._stop_wait is not None:
+                return self._stop_wait
+            waiter = self._stop_wait = trio.Event()
         else:
             waiter.clear()
 

trio_asyncio/base.py

@@ -172,6 +172,8 @@ def __init__(self, queue_len=None):
         self._stopped = trio.Event()
         self._stopped.set()
 
+        self._stop_wait = None
+
     def __repr__(self):
         try:
             return "<%s running=%s at 0x%x>" % (