I'm not sure how to handle the failing check in pylint/config/__init__.py on line 129.
Seems to me that the checker has identified a problem in the current code, as globals() is not interpolated into the string. Do we want to do this? Or should we add the checker to the ignore list of this file?

Coverage increased (+0.003%) to 92.59% when pulling 0d0b145 on DanielNoord:format-string into a71cfe1 on PyCQA:main.

Regarding config init : Very good news we discovered a problem in pylint with this new checker ? 😄 I think we can remove globals() from this string (to keep the same behavior at least). I think this would help during debug but clearly no one used this in a long time or we'd have a bug report at this point.

This is probably me not understanding how old_names work, but now all tests are failing as all too-many-format-args seem to have become format-string-without-interpolation. Probably my mistake, but not sure what happened here..

Ho no sorry my bad. It's not possible to rename only a part of it. We'd have to rename both. It means we'd need to rename [too-many-format-args too. I don't know about that. this seems hacky. Maybe if we find a better name we can kill two bird with one stone ?

Hmm too-many-format-args seems pretty good as is. Is there no way to keep it the same?
Perhaps additional-format-args? Although I think too-many-format-args is still better, especially considering too-few-format-args is also a thing..

Another way would be to refactor the MessageIdStore so it's possible to partially rename a message. I think the consequences is that you'd need to instantly rename all the messages that need to be renamed when partially renaming a message or you'd get the old name (kinda defeating the purpose of it in the first place). Beside if you have a too-many-format-args this is most likely an error so you won't disable it very often. I think reverting the old_names I asked is the sane decision here. Sorry for wasting your time with this comment.

No problem, I can imagine why you wanted to add old_message initially!

Not that I care too much about it, but it seems like the Copyright lines at the top of a document are not updated automatically. Should I create a PR to do so?

It's done automatically but during release (check doc/release.md). It's a pretty heavy calculation using copyrite, also it changes the order of line randomly which is not ideal but not a real problem either. This lib could be optimized or changed but it never bothered me enough to actually do anything. I you have a better solution on the shelf I'm open to suggestion :)

Ah okay! Don't really have a better solution, but will keep my eyes open!