Thanos

`site_data_path` is explicitly documented as returning only one path, even with multipath passed in as True. Have you considered using `site_data_dir(multipath=True)` for your use case?

As a bystander, I can't speak for the developers of this library, but if I had to guess? It might be an established convention in the library for `*_path` methods...

I would personally consider this a known API feature, but I’m not you.

Would you prefer fixing the upstream package or redoing the logic internally? I ask because I’ve written a [proof of concept](https://github.com/Sachaa-Thanasius/experimental/blob/master/src/__experimental__/features/_inline_import.py) for a shorter version of the same functionality. It...

Regarding the remaining points: - I have a bot, but depending on the meaning of "production", it might not fill that criteria? It's a small one just for running jishaku,...

On a tangential note: it was mentioned in the related issue that the inline import feature should pass "the litmus test" on 3.8-3.12, but in the overall tracking issue for...

Since ioistired is back, I'll close this PR once upstream has been fixed.

I just assumed my attempt at implementation was wrong and gave up. Should this be reopened, then?

Understood. Should slice be changed in the runtime cpython as well to support `__class_getitem__` via `GenericAlias`, or is this purely meant to be a typeshed thing? Not something I considered...

Also, sidenote: The pyright errors specifically for `slice.stop` and and the first overload of `slice.__new__` (with text like the following) are the result of a bug that was fixed [earlier...