Don't run pylint by default even if installed

[ptrcnull]

tests for 0.28.4 fail because mockobject.py was not linted correctly:

..............................................s.................************* Module dbusmock.mockobject
dbusmock/mockobject.py:355:0: R0022: Useless option value for 'disable', 'no-self-use' was moved to an optional extension, see https://pylint.pycqa.org/en/latest/whatsnew/2/2.14/summary.html#removed-checkers. (useless-option-value)
dbusmock/mockobject.py:455:17: C3001: Lambda expression assigned to a variable. Define a function using the "def" keyword instead. (unnecessary-lambda-assignment)

------------------------------------------------------------------
Your code has been rated at 9.96/10

Es...............................ssssssssss.sssssssssssssssssss....ssssss.......ssss..sssssssssssss.....
======================================================================
ERROR: test_pylint (tests.test_code.StaticCodeTests)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "tests/test_code.py", line 39, in test_pylint
    subprocess.check_call([sys.executable, '-m', 'pylint'] + glob.glob('dbusmock/*.py'))
  File "/usr/lib/python3.10/subprocess.py", line 369, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['/usr/bin/python3', '-m', 'pylint', 'dbusmock/__main__.py', 'dbusmock/__init__.py', 'dbusmock/mockobject.py', 'dbusmock/testcase.py']' returned non-zero exit status 24.

----------------------------------------------------------------------

i'm not really sure why pylint has to be launched from tests, but if you include lints in your testsuite, please check them before releasing a new version :p

[martinpitt]

I do check them -- they run in CI. But only on Fedora stable, not on Rawhide any more. Recently pylint changed behaviour wrt. that no-self-use thing, and there just isn't a way to make the code happy with both 2.13 and 2.14.

Where do you see this failure exactly? It's not in any of the distro builds (pylint should not run anywhere there, it changes too often) or upstream CI.

[martinpitt]

In other words, this can never be truly "fixed" -- if pylint changes to a newer version and changes rules, this will break all released versions. That's not really pylint's fault, it just means that it does not make much sense to run pylint on released versions.

Right now the unit tests merely check if pylint and friends are installed. I should probably change that to explicitly opt-in, so that my CI can do this in a way that doesn't break test runs for users?

[ptrcnull]

thanks for the quick response!

there just isn't a way to make the code happy with both 2.13 and 2.14

ohh, okay, that makes way more sense now

Where do you see this failure exactly?

running python3 setup.py test on Alpine Linux edge (pylint 2.14.4)

pylint should not run anywhere there

well, it's being picked up by the test runner and while it's not a huge problem while packaging for distributions (unless pytest is pulled in by something), it would be cool if lints were not included in the test suite by default

[martinpitt]

Fixed in PR #156