Capturing output causes test to fail.

[dh-nunes]

When running with capture disabled, test passed but with enabled test fails with no reason given. Not sure if relevant, but the test is failing on a c extension module built with cython.

Windows 64-bit, Python 3.4.5 32-bit, pytest 3.0.5

A screenshot with pip list and the results of running py.test -s and py.test respectively. bethesdalib is the package being tested.

The test code:

import os
from bethlib.tesfive import pex, esp

def test_pex():
    ...

def test_esp():
    data_folder = os.path.join(os.path.abspath(os.path.dirname(__file__)), "data", "esp")
    for plugin_file in next(os.walk(data_folder), (None, None, []))[2]:
        plugin = esp.parse(os.path.join(data_folder, plugin_file))
        with open(os.path.join(data_folder, plugin_file), 'rb') as open_plugin:
            assert open_plugin.read() == esp.to_string(plugin)

Thanks!

[nicoddemus]

Hi @GandaG,

pytest's exception capture in fd mode is probably causing the C extension module to fail somehow. It is probably the fault of the extension module not handling the fact that the out/err handles have been redirected, but is hard to know without debugging this somehow.

What happens if you run pytest with --capture=sys?

Could you try running pytest with a debugger attached and see where it crashes?

[dh-nunes]

🤦‍♂️ Sorry, I just went full stupid. I am redirecting the output from the extension module to handle errors. It functions perfectly if I remove that part.

Thank you! And sorry :/

[nicoddemus]

Thank you! And sorry :/

No worries! Thanks for closing the issue then.