GeezerVison

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

This feels unfortunate and maybe should be fixed in CPython instead. Or perhaps the parameter should be made positional-only.

This feels unfortunate and maybe should be fixed in CPython instead. Or perhaps the parameter should be made positional-only.

Yeah, I think one of the two would be the better fix. Kinda the point of os.path is that the module is meant to be an abstraction that's fully cross-platform. So it makes things somewhat awkward when a function in os.path has a different signature on Windows than on other platforms. See #6812, where we had a similar issue, and tackled it by pretending the parameter was positional-only.

Here I think the best solution might be to raise this with CPython. Since the ntpath implementation is new, it seems unlikely to me that they intended to give it a different parameter name to the posix implementation.

Yeah I think if this was not intentional in CPython, there's still time to change it. I will open an issue now.

The fix was merged into CPython and is planned to go into 3.12 eventually, so I think we shouldn't change anything here. Thanks for uncovering this issue!