-
-
Notifications
You must be signed in to change notification settings - Fork 30.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix an outdated test in asyncio #24477
Conversation
The test was added in f3e2e092 before `_StopError` was removed in 41f69f4, after which this test will not fail without the fix it covers. This PR fixes the test to resume its coverage.
This PR is stale because it has been open for 30 days with no activity. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ran 2287 tests in 83.131s
OK (skipped=18) All windows tests
Also hit base_events.
Looks good.
✅ Deploy Preview for python-cpython-preview canceled.
|
self.loop.run_forever() | ||
except KeyboardInterrupt: | ||
pass | ||
self.loop.call_later(0.01, func) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A bit unfortunate that we had to add a 10 msec delay here -- in most cases that will waste 10 msec and occasionally it will not wait long enough and the test will flake-fail. Why won't call_soon work?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Background: this test covers a bug that base exceptions like KeyboardInterrupt
will cause the loop to be closed twice, leading to an early exit in the next loop run.
This test was added before _StopError
was dropped, at that time the loop will stop as soon as _StopError
is raised, even if the _ready
queue still has a handle (which is the next run - the func()
here), therefore this was a good test. But now the loop will exhaust the _ready
queue before stopping, so this test couldn't cover the bug anymore without this PR. That's also why we cannot simply add func()
into the _ready
queue by call_soon()
.
You're right about the flakyness of using call_later()
tho. Now I think a better fix would be to use chained call_soon()
to skip the first iteration where the buggy stop occured:
self.loop.call_later(0.01, func) | |
# If the issue persists, the loop will exit early after the first iteration, | |
# in that case our func() in the second iteration won't be called. | |
self.loop.call_soon(self.loop.call_soon, func) |
I verified and this error fails if the fix in f3e2e09 is reverted.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Clever solution, go ahead and make it a PR, assign to me or @kumaraditya303 for review.
The test was added in f3e2e092 before
_StopError
was removed in 41f69f4, after which this test will not fail without the fix it covers. This PR fixes the test to resume its coverage.Refs: MagicStack/uvloop#337