Sergey B Kirpichev

FYI, @oscarbenjamin: this is valid for the SymPy as well. I think it's an issue.

> setup.py is programmatically choosing the right path separator depending on the OS Exactly. > if you use static strings in setup.py I believe that you are going to observe...

> It looks like Python 3.13 will provide public C API Nit: 3.14. Maybe.

Maybe it's wise to wait for a fate of above CPython PR. Note that special rules for real arguments can't save all identities, e.g. for atan: https://github.com/python/cpython/pull/124829#issuecomment-2387555557 Proper solution would...

Ok, the CPython PR was merged and I'm working on this issue. Probably, imaginary type (https://github.com/skirpichev/cpython/pull/1) will not be supported by Python. Anyway, the MPC has mixed-mode primitives only for...

> I suspect regenerating just parts of the build will be complicated. Well, regenerating zip files should be trivial. Anyway, (1)-st option still relevant. I see several requests to remove...

After removing `*.zip` files I got a lot of warnings from sphinx-build while rebuilding. This will break CI.

CC @serhiy-storchaka

Yes, I expected that argument, though I wasn't sure if it was the major reason to not have closure property for arithmetic in subclasses (other candidates: (a) legacy, (b) speed)....

> Yes, this is our long-standing rule. I don't think it should be changed. With this, subclassing any numeric types in the stdlib is useless, isn't?