False positive: E1307 Argument 'builtins.NoneType' does not match format type 'd'

[tardyp]

Steps to reproduce

Lint following snippet on python3:

# pylint: disable=C
a = None
if a is not None:
    print("%d" % (a,))

Real life usecase is this code:

https://github.com/buildbot/buildbot/blob/master/master/buildbot/schedulers/forcesched.py#L172
https://github.com/buildbot/buildbot/blob/master/master/buildbot/steps/python_twisted.py#L444

Current behavior

pylint output wrong error:

testpylint.py:5:10: E1307: Argument 'builtins.NoneType' does not match format type 'd' (bad-string-format-type)
buildbot/schedulers/forcesched.py:172 Argument 'builtins.NoneType' does not match format type 'd' [bad-string-format-type]
buildbot/steps/python_twisted.py:444 Argument 'builtins.NoneType' does not match format type 'd' [bad-string-format-type]

Expected behavior

no error

pylint --version output

astroid 2.1.0
Python 3.5.2 (default, Nov 12 2018, 13:43:14)
[GCC 5.4.0 20160609]

This is a regression from pylint 2.1.1

[tardyp]

looks like E1307 was added in 2.2.0 in ba62048

[tardyp]

@yugr do you think it is feasible to add a check for None inspection in your checker?

[yugr]

Hm, I think the real problem here is not the checking of None per se (it's indeed incorrect to print("%d" % None)) but rather that utils.safe_infer returns NoneType for a. Is there some general approach to fix false positives that require context-specific handling?

[PCManticore]

This is a problem with our control flow inference, which cannot understand the if guard clause. We have a lot of these false positives for other checks as well, all stemming from the fact that our inference engine doesn't understand flow.

[yugr]

@PCManticore Right, I guess there is not much the checker can do then.

[mbyrnepr2]

This one no longer reproduces (pylint 2.17.0-dev0 and astroid 2.15.0-dev0) & should be fixed according to pylint-dev/astroid#1189.