Ignores all errors when using noqa: X to only ignore X

[blueyed]

When using # noqa: E501 all errors are excluded.

This is due to the code in

flake8-bugbear/bugbear.py

Lines 39 to 40 in 6bf40ea

	if pycodestyle.noqa(self.lines[e.lineno - 1]):
	continue

.

pycodestyle.noqa searches for r'# no(?:qa|pep8)\b', which matches # noqa: E501 also.

Without knowing the internals it seems like flake8 itself should include the errors itself - it has logic to ignore only specific errors (https://github.com/PyCQA/flake8/blob/013b925624bd5e4c026e83eb2c6e9f45cafa0dc0/src/flake8/defaults.py#L39-L51).

Example:

s = 'some very, very, very, very, very, very, very, very, very, very, very, very, very, very, very long line'  # noqa: E123

flake8-bugbear should report error t-noqa-colon.py:1:110: B950 line too long (109 > 80 characters) for it, but is silent.

[plinss]

It's considered an anti-pattern for flake8 plugins to filter errors based on # noqa comments. Best practice is to simply send all errors and let flake8 filter them based on user settings, e.g. --disable-noqa should show these errors, but won't because they never get to flake8.

As an aside, this behavior also breaks a plugin of mine: https://github.com/plinss/flake8-noqa, see plinss/flake8-noqa#1