Test failure on aarch64

[dotlambda]

When I run the test suite (python setup.py test) on a 64-bit ARM machine running NixOS and Python 3.6, I get the following error:

======================================================================
FAIL: test_idle_start__exits_queueget_without_timeout_error (aioimaplib.tests.test_aioimaplib.TestAioimaplibClocked)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/nix/store/q4mxm1959nr9bh6vmbml6f5vr0phfgh8-python3.6-asynctest-0.11.1/lib/python3.6/site-packages/asynctest/case.py", line 294, in run
    self._run_test_method(testMethod)
  File "/nix/store/q4mxm1959nr9bh6vmbml6f5vr0phfgh8-python3.6-asynctest-0.11.1/lib/python3.6/site-packages/asynctest/case.py", line 351, in _run_test_method
    self.loop.run_until_complete(result)
  File "/nix/store/q4mxm1959nr9bh6vmbml6f5vr0phfgh8-python3.6-asynctest-0.11.1/lib/python3.6/site-packages/asynctest/case.py", line 221, in wrapper
    return method(*args, **kwargs)
  File "/nix/store/4yfkb61l14x37pkqdmg4ck61xh5wg49z-python3-3.6.4/lib/python3.6/asyncio/base_events.py", line 467, in run_until_complete
    return future.result()
  File "/build/source/aioimaplib/tests/test_aioimaplib.py", line 852, in test_idle_start__exits_queueget_without_timeout_error
    self.assertEqual(STOP_WAIT_SERVER_PUSH, r)
AssertionError: 'stop_wait_server_push' != ['OK Still here', 'OK Still here', 'OK St[1263 chars]ere']
-------------------- >> begin captured logging << --------------------
asyncio: DEBUG: Using selector: EpollSelector
--------------------- >> end captured logging << ---------------------

Do you know the reason for this?

See also https://hydra.nixos.org/build/68772894/log.

[dotlambda]

Btw, there are even more failures on Darwin: https://hydra.nixos.org/build/68775014/log Nevertheless, tests are running without a problem on x86_64.

[bamthomas]

Sorry for this very late response, I've been a bit overwhelm with my job. No I don't know the reason. At first sight, I'm wondering about charset management, that was often the case for this literal length failures.

[aliceh75]

The same test is failing here with Python 3.5 on an x86_64 Ubuntu 16.04 box. It doesn't always fail - there's a timing issue - but it does most of the time.

I've had a look at it, and I can say the following is happening:

• IMAP4.wait_server_push returns the last thing present on IMAP4ClientProtocol.idle_queue
• The test in question expects that value to be STOP_WAIT_SERVER_PUSH, which is added on the queue by IMAP4.stop_wait_server_push, which is called 29 minutes (the default) after IMAP4.idle_start was invoked. The purpose of this is that, even though we're waiting for server to push things to us, we should talk to it every 29 minutes as it might time us out otherwise;
• What happens when the test fails is that IdleCommand.flush is invoked after IMAP4.stop_wait_server_push, and that adds on top of IMAP4ClientProtocal.idle_queue a number of 'OK Still here' messages that were received from the server and buffered on IdleCommand.buffer. So the last entry on that queue is not STOP_WAIT_SERVER_PUSH as the test was expecting.

I don't know yet whether it is expected that IdleCommand.flush can be invoked after IMAP4.stop_wait_server_push. Depending on that, there is either a problem in the code, or a problem in the test.

I won't have time to look at it more for a few days, so if anyone has some further insights feel free to continue working on this one :-)

[bamthomas]

the test aioimaplib.tests.test_aioimaplib.TestAioimaplibClocked is always failing on travis with ubuntu 14.04 / python 3.5.5 installed from tarball. It has passed 10 times on my laptop xubuntu 16.04 with python 3.6.3 installed from ppa jonathonf/python-3.6 . It has passed also with python 3.5.2

I'm wondering if it is related to https://github.com/bamthomas/aioimaplib/issues/30