I'm using dbus for IPC. In order to have exactly one bus throughout the livetime of my program I'm using a singleton here. For the sake of demonstration I'm connecting to NetworkManager but that could be exchanged. Furthermore, I'm using asyncio
for the entire project.
This is a minimalistic working example of the module that will highlight the problem described below:
import asyncio # noqa
from dbus_next.aio import MessageBus
from dbus_next import BusType
BUS = None
async def get_bus():
# Returns a BUS singleton
global BUS
if not BUS:
BUS = await MessageBus(bus_type=BusType(2)).connect()
return BUS
async def introspect():
# Get the dbus singleton and call a method on that singleton
bus = await get_bus()
return await bus.introspect(
'org.freedesktop.NetworkManager',
'/org/freedesktop/NetworkManager',
)
I'm using pytest
with the pytest-asyncio
plugin for testing which works like charm except for this case.
This is a minimalistic work test module:
import pytest
from example import introspect
@pytest.mark.asyncio
async def test_example_first():
# With only this first call the test passes
await introspect()
@pytest.mark.asyncio
async def test_example_second():
# This second call will lead to the exception below.
await introspect()
When I execute that test, I'm getting the following exception indicating that the event loop changed:
example.py:22: in introspect
'/org/freedesktop/NetworkManager',
../.local/lib/python3.7/site-packages/dbus_next/aio/message_bus.py:133: in introspect
return await asyncio.wait_for(future, timeout=timeout)
/usr/lib/python3.7/asyncio/tasks.py:403: in wait_for
fut = ensure_future(fut, loop=loop)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
coro_or_future = <Future pending>
def ensure_future(coro_or_future, *, loop=None):
"""Wrap a coroutine or an awaitable in a future.
If the argument is a Future, it is returned directly.
"""
if coroutines.iscoroutine(coro_or_future):
if loop is None:
loop = events.get_event_loop()
task = loop.create_task(coro_or_future)
if task._source_traceback:
del task._source_traceback[-1]
return task
elif futures.isfuture(coro_or_future):
if loop is not None and loop is not futures._get_loop(coro_or_future):
> raise ValueError('loop argument must agree with Future')
E ValueError: loop argument must agree with Future
I'm guessing that pytest started an event loop and during module import another event loop was started but I'm not sure. I tried enforcing using the pytest
or the module event loop using asyncio.set_event_loop()
but I had no success. The result stayed the same.
Is my assumption correct? How can I either enforce using a global event loop? Alternatively, how should I define the singleton to make this work with pytest
?
Probably worth noting is that this singleton construct works perfectly fine in the context of the program. It's just the test which I can't figure out how to make it work.
From looking at this, looks like you're using a global variable which connects to something. Persistent connections are bound to an event loop. So if you want to continue using a global variable, the scope of the event loop fixture needs to match the scope of this variable.
I would redefine the event loop fixture to be session scoped, and create a session scoped fixture, depending on the event loop fixture, to initialize this global variable (by just calling
get_bus
, I suppose). The second fixture is required to ensure proper initialization ordering - i.e. that the event loop is set up properly first.Edit: The documentation of
pytest-asyncio
says that event loops by default are test function scoped. Thus, they are being recreated in for each test. This default behavior can be simply overridden: