add simple pre-commit config

[finswimmer]

This PR adds a basic pre-commit config. More pre-commit checks and configuration should be discussed.

To run the checks on a PR you have to give https://pre-commit.ci access to the repository. Checks will then run automatically on the next push

Closes: #197

[cooperlees]

Looks like a good start! Thanks. If you're planning to do more (I see this is draft) I'll leave it or we can merge and work on it.

[finswimmer]

Hey @cooperlees,

I'v updated the DEVELOPMENT.md to show how to run the linter, removed black from the dependencies in the setup.py and also removed from the black check in the github workflow, because it's running now with the pre-commit checks.

I'm not sure why the tests on python3.9 fail. Locally there is no problem. Any ideas?

fin swimmer

[cooperlees]

Yeah this is a weird fuzx test one - I checked on main to see if I could reproduce to I was unable to:

python3 -V
Python 3.9.7
python3 -m venv --upgrade-deps /tmp/tf
/tmp/tf/bin/pip install -e .[dev]
[cooper@cooper-fedora-MJ0BWXDF flake8-bugbear]$ /tmp/tf/bin/coverage run tests/test_bugbear.py
.............................
----------------------------------------------------------------------
Ran 29 tests in 3.307s

OK
[cooper@cooper-fedora-MJ0BWXDF flake8-bugbear]$ /tmp/tf/bin/coverage report -m
Name         Stmts   Miss  Cover   Missing
------------------------------------------
bugbear.py     403     13    97%   65-66, 76, 100-110, 322, 349-352, 360
------------------------------------------
TOTAL          403     13    97%

Fail is from test_does_not_crash_on_any_valid_code:
SystemError: Negative size passed to PyUnicode_New

Applying it to main + trying to reproduce I get an error as your code has changed things - Will try with your branch if I get time today.

[cooper@cooper-fedora-MJ0BWXDF flake8-bugbear]$ git diff
diff --git a/tests/test_bugbear.py b/tests/test_bugbear.py
index 0573e11..3dd0273 100644
--- a/tests/test_bugbear.py
+++ b/tests/test_bugbear.py
@@ -6,7 +6,7 @@ import subprocess
 import sys
 import unittest
 
-from hypothesis import HealthCheck, given, settings
+from hypothesis import HealthCheck, given, settings, reproduce_failure
 from hypothesmith import from_grammar
 
 from bugbear import BugBearChecker, BugBearVisitor
@@ -328,6 +328,7 @@ class BugbearTestCase(unittest.TestCase):
 
 
 class TestFuzz(unittest.TestCase):
+    @reproduce_failure('6.27.0', b'AXicY2RkZGZgZGBgYGIBkXDAzAjmQsQAAhkAFQ==')
     @settings(suppress_health_check=[HealthCheck.too_slow])
     @given(from_grammar().map(ast.parse))
     def test_does_not_crash_on_any_valid_code(self, syntax_tree):
[cooper@cooper-fedora-MJ0BWXDF flake8-bugbear]$ /tmp/tf/bin/coverage run tests/test_bugbear.py
.........................E...
======================================================================
ERROR: test_does_not_crash_on_any_valid_code (__main__.TestFuzz)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/cooper/repos/flake8-bugbear/tests/test_bugbear.py", line 332, in test_does_not_crash_on_any_valid_code
    @settings(suppress_health_check=[HealthCheck.too_slow])
  File "/tmp/tf/lib64/python3.9/site-packages/hypothesis/core.py", line 1113, in wrapped_test
    raise DidNotReproduce(
hypothesis.errors.DidNotReproduce: The shape of the test data has changed in some way from where this blob was defined. Are you sure you're running the same test?

----------------------------------------------------------------------
Ran 29 tests in 1.127s

FAILED (errors=1)

[finswimmer]

Hey @cooperlees,

could you find out what happens with this fuzzy test? I'm lost :(

fin swimmer

[cooperlees]

Yeah, me too. I think we can merge this tho and try work this out before releasing. This has been happening on other PRs too.

I'll try and get a local repro the next few days and nudge @Zac-HD for pointers on how to fix here ...

[Zac-HD]

Ah, that's actually a CPython bug! https://bugs.python.org/issue45738

I just haven't shipped a workaround in hypothesmith yet 😅

[Zac-HD]

Scratch that, Hypothesmith 0.2.0 is now out and works around this upstream bug 👍

[cooperlees]

Thanks @Zac-HD !