bpo-39010: Improve test shutdown (#22066)
Simply closing the event loop isn't enough to avoid warnings. If we don't also shut down the event loop's default executor, it sometimes logs a "dangling thread" warning. Follow-up to GH-22017
This commit is contained in:
parent
0770ad948c
commit
be435ae2b0
|
@ -225,10 +225,18 @@ class ProactorTests(test_utils.TestCase):
|
|||
self.loop.run_forever()
|
||||
self.loop.stop()
|
||||
self.loop.run_forever()
|
||||
# If we don't wait for f to complete here, we may get another
|
||||
# warning logged about a thread that didn't shut down cleanly.
|
||||
|
||||
# Shut everything down cleanly. This is an important part of the
|
||||
# test - in issue 39010, the error occurred during loop.close(),
|
||||
# so we want to close the loop during the test instead of leaving
|
||||
# it for tearDown.
|
||||
#
|
||||
# First wait for f to complete to avoid a "future's result was never
|
||||
# retrieved" error.
|
||||
self.loop.run_until_complete(f)
|
||||
self.loop.close()
|
||||
# Now shut down the loop itself (self.close_loop also shuts down the
|
||||
# loop's default executor).
|
||||
self.close_loop(self.loop)
|
||||
self.assertFalse(self.loop.call_exception_handler.called)
|
||||
|
||||
|
||||
|
|
Loading…
Reference in New Issue