No output when running from nosetests and jobs != 1
Dear Flake8ers, I am trying to run flake8 programmatically in a Test class as follows: import flake8.main import nose.tools as nt class TestCodeFormat(object): def setUp(self): self.files = find_py_files(os.path.dirname(mmfutils.__file__)) def test_flake8_conformance(self): # Pass in the files here so that config setup.cfg file gets found style_guide = flake8.main.get_style_guide(paths=self.files) result = style_guide.check_files() nt.eq_(result.total_errors, 0, "Found {0} code style errors (and warnings)." .format(result.total_errors)) The problem I am having is that there is no output when jobs > 1. This is probably something to do with stdout flushing and multiprocessing when running under nosetests. Is there any solution that will allow parallel processing with displayed output. (I could change the error message to use result.messages or something, but don't see an easy way of properly formatting output with filenames and lines numbers etc.) Thanks, Michael. P.S. 1) Please let me know if I should file an issue. I thought this might be resolved by #21 but even using the latest sources does not help. 2) flake8 and python setup.py flake8 both work fine, so another solution would be if I can somehow trigger this from "python setup.py nosetests" or "python setup.py test" but it seems to be something of a challenge to customize setup.py commands. 3) Setting nose's nocapture=1 also works, but the errors then are not deferred until the end of the tests and so can get lost.
On Sat, Mar 7, 2015 at 10:23 PM, Michael McNeil Forbes <michael.forbes+python@gmail.com> wrote:
Dear Flake8ers,
I am trying to run flake8 programmatically in a Test class as follows:
import flake8.main import nose.tools as nt
class TestCodeFormat(object): def setUp(self): self.files = find_py_files(os.path.dirname(mmfutils.__file__))
def test_flake8_conformance(self): # Pass in the files here so that config setup.cfg file gets found style_guide = flake8.main.get_style_guide(paths=self.files) result = style_guide.check_files() nt.eq_(result.total_errors, 0, "Found {0} code style errors (and warnings)." .format(result.total_errors))
The code looks vaguely correct. I don't remember the behaviour of using get_style_guide without providing a list of default config files though (I doubt it affects this though).
The problem I am having is that there is no output when jobs > 1. This is probably something to do with stdout flushing and multiprocessing when running under nosetests.
Is there any solution that will allow parallel processing with displayed output. (I could change the error message to use result.messages or something, but don't see an easy way of properly formatting output with filenames and lines numbers etc.)
nose may not support this usecase.
Thanks, Michael.
P.S.
1) Please let me know if I should file an issue. I thought this might be resolved by #21 but even using the latest sources does not help.
2) flake8 and python setup.py flake8 both work fine, so another solution would be if I can somehow trigger this from "python setup.py nosetests" or "python setup.py test" but it seems to be something of a challenge to customize setup.py commands.
3) Setting nose's nocapture=1 also works, but the errors then are not deferred until the end of the tests and so can get lost.
So if you do `python setup.py tests -s` then the errors will all be printed, correct? I suspect this is something to do with how flake8 defers printing and how nosetests expects to receive data on stdout. I'm not sure flake8 wants to support nose like this.
participants (2)
-
Ian Cordasco
-
Michael McNeil Forbes