`B950` ignores `max-doc-length` flake8 configuration option

[SRv6d]

flake8 has a configuration option max-doc-length, which configures the maximum line length for docstrings. flake8-bugbear currently ignores this setting.
It makes sense to limit line length of code to ~80-88 for smaller screen sizes, diffs etc. Especially within tests though, limiting docstring line length to the same value can cause very awkward docstring formatting, severely limiting readability.

[cooperlees]

PR would be accepted to make us respect it. We don't lint docstrings and I don't think we ever should. So happy to respect this.

[samueljsb]

pydocstyle doesn't use this setting for the E501 check (maximum line length) that B950 replaces. The setting is only used to configure the W505 check (maximum doc length).

How should flake8-bugbear respect max-doc-length? Should it apply different rules to docstring and non-docstring lines (like a combination E501/W505 check), or should multiline strings not be linted at all (by B950)? In either case I think the check would need to be changed to run against the AST rather than just the raw line; I'm not sure how else we'd know whether we were looking at a multiline string or not.

[cooperlees]

My thoughts here are:

1. We shouldn't upset black or rather not let black upset us
   • Which I think keeps similar lengths unless it's URLs (but I can't remember)
2. I think I am happy to support the different lengths as long as we uphold point 1 and don't upset behavior today people have become accustom to with max-length like behavior on non docstrings today and does the same behavior if max-doc-length is not set.