rednose icon indicating copy to clipboard operation
rednose copied to clipboard

Rednose and processes do not work togeather

Open bartlomiej-kurek opened this issue 8 years ago • 9 comments

Hi, I've run into a bizarre issue with rednose, python 3.5.2, Ubuntu 16.04. Problem: when multiple inheritance is used rednose discards the failure. It depends on the order of defining test classes/methods.

The code to reproduce the problem: $ cat test_rednose.py

import unittest


class FrontendCase(unittest.TestCase):
    pass

class BackendCase(unittest.TestCase):
    pass

class BackendUserCase(BackendCase):
    pass

class ExistingBackendUserCase(BackendUserCase):
    pass

class FrontendUserCase(FrontendCase, BackendUserCase):
    pass

class ExistingFrontendUserCase(FrontendUserCase, ExistingBackendUserCase):
    pass

class Test_0(FrontendUserCase):
    def test_dummy(self):
        pass


class Test_1(ExistingFrontendUserCase):
    def test_assertion(self):
        print("Should fail")
        self.assertTrue(False)

$ cat nosetests.cfg

[nosetests]
verbosity = 1
with-doctest = 0
logging-level = DEBUG
logging-format = "%(asctime)s %(name)-12s %(levelname)-8s %(module)s:%(lineno)s %(message)s"
nocapture = 1
nologcapture = 1
rednose = 1
processes = 1

Run it:


$ nosetests -c nosetests.cfg test_rednose.py 
Should fail
.
----------------------------------------------------------------------
Ran 1 test in 0.031s

OK

Now, comment out rednose in nosetests.cfg and run again:

$ grep -v rednose nosetests.cfg > nosetests.without_rednose.cfg
$ nosetests -c nosetests.without_rednose.cfg test_rednose.py 
.F
======================================================================
FAIL: test_assertion (test_rednose.Test_1)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "./test_rednose.py", line 30, in test_assertion
    self.assertTrue(False)
AssertionError: False is not true

----------------------------------------------------------------------
Ran 2 tests in 0.020s

FAILED (failures=1)

Voila, finally assertTrue(False) fails as it should.

This issue doesn't seem to be related to "--processes" in nosetests. What is really bizarre is that: a) if we just remove Test_0::test_dummy method (leaving Test_0 class empty), the assertion will properly fail. b) The test method in Test_1 class is actually executed up to that assertion point.

I didn't debug rednose itself.

$ python -V Python 3.5.2 $ uname -rmpov 4.4.0-62-generic #83-Ubuntu SMP Wed Jan 18 14:10:15 UTC 2017 x86_64 x86_64 GNU/Linux

bartlomiej-kurek avatar Feb 15 '17 00:02 bartlomiej-kurek

I'll try to reproduce this later this week using your code and dig into it! Thanks for the bug report, what an odd bug.

JBKahn avatar Feb 15 '17 01:02 JBKahn

hmmm it does seem like it works with everything but that library, the test output isn't even colored with processes = 1

JBKahn avatar Feb 15 '17 22:02 JBKahn

[nosetests]
rednose = 1
processes = 1

alone breaks it.

JBKahn avatar Feb 15 '17 22:02 JBKahn

I will look into it.

JBKahn avatar Feb 15 '17 22:02 JBKahn

for example,

(rednose) josephkahn (2017-02-15 18:00:18)>~/dev/rednose (master)$ nosetests --rednose test_files/new_tests.py  --processes=1
.
----------------------------------------------------------------------
Ran 1 test in 0.235s

OK
(rednose) josephkahn (2017-02-15 18:00:31)>~/dev/rednose (master)$ nosetests --rednose test_files/new_tests.py
......

-----------------------------------------------------------------------------
6 tests run in 0.178 seconds (6 tests passed)

(rednose) josephkahn (2017-02-15 18:00:35)>~/dev/rednose (master)$ nosetests test_files/new_tests.py  --processes=1
......
----------------------------------------------------------------------
Ran 6 tests in 0.236s

OK

JBKahn avatar Feb 15 '17 23:02 JBKahn

I started taking a look and this is non-trivial so I'm not sure I'll have time to do a crazy deep dive on it, it's an artifact of a really awful plugin api for nose. Colorizing the output shouldn't affect this at all, but it does, in ways I don't entirely understand just yet.

JBKahn avatar Feb 16 '17 15:02 JBKahn

I can comment all of rednose, except for

    def prepareTestResult(self, result):  # noqa
        """Required to prevent others from monkey patching the add methods."""
        return result

and it still breaks the multiprocess package but I can't see why that is.

while this works


    def prepareTestResult(self, result):  # noqa
        """Required to prevent others from monkey patching the add methods."""
        return None

JBKahn avatar Feb 16 '17 15:02 JBKahn

Fixed in here: https://github.com/JBKahn/rednose/pull/22

JBKahn avatar Feb 11 '18 18:02 JBKahn

I'm not sure it actually works, I think it might just be you cannot use them together and it's running only one of the two now.

JBKahn avatar Feb 11 '18 18:02 JBKahn