twisted
txaws.s3.tests.test_client.LiveS3TestCase.test_objects can fail with DirtyReactorAggregateError
On a local test run:
[ERROR]
Traceback (most recent call last):
Failure: twisted.trial.util.DirtyReactorAggregateError: Reactor was unclean.
Selectables:
<<class 'twisted.internet.tcp.Client'> to ('s3.amazonaws.com', 443) at 7fd3e54bb550>
txaws.s3.tests.test_client.LiveS3TestCase.test_objects
This came from the improved-s3-put_object-testing branch which may be relevant but my hunch is that it's caused by a race condition between disconnection and the last Deferred in the test and happens randomly, even on master, with some frequency.
The deferred returned by Agent.request will fire immediately when the request completes, whereas the underlying connection may still be busy with TCP/TLS shutdown, or idle in a connection pool. The way I've usually solved this is to create a connection pool explicitly in my tests, then you can invoke https://twistedmatrix.com/documents/current/api/twisted.web.client.HTTPConnectionPool.html#closeCachedConnections at the end of the test to wait for the connection(s) to be closed.
