scoder

Thanks for getting this working.

> I'm pretty suspicious of this in general. So am I. Let's not go down this path right now. We might allow annotation style typing in .pxd files in a...

Py2.7 doesn't have `ensurepip`, so _relying_ on setuptools in Cython 3.0 doesn't seem like a good idea. However, _preferring_ it over distutils, if available, seems worthwhile. For Cython 3.1, I...

> > I don't see a need to do so, though, until we're gaining better integration with setuptools than with distutils, i.e. until setuptools _does_ something for us > >...

To come back to this, I think we should switch to setuptools in Cython 3.0, with an import fallback to distutils (since Py2.7 still needs it). We should keep support...

Hmm. Interesting. No-one ever complained about this. It's how C enums work, and `cpdef enum` defines a C enum (`cdef enum`) and makes it _additionally_ available to Python. Whether that...

Given that we allow both ```cython cpdef enum: A, B, C ``` and ```cython cpdef enum ANamedEnum: X, Y, Z ``` it seems reasonable to distinguish between the two so...

Here's a PR. Let's discuss if we want this change in behaviour. https://github.com/cython/cython/pull/6382 It will probably break existing code and/or tests. It seems just too easy to use enum names...

> It may be easier to adjust what is written than to arrange for `__members__` to be directly accessible from `.pyx` Cython code. It wasn't difficult, though. Since we're generating...

> It has some cost in the non-free-threaded builds because: > > * it unchains cdef class attribute lookups so that they go via temps. > * It makes some...