Re: python version confusion with cython (Re: math/py-scipy)

[Greg Troxel <gdt%lexort.com@localhost>]

"Dr. Thomas Orgis" <thomas.orgis%uni-hamburg.de@localhost> writes:

> Indeed this seems to be what upstream strongly suggests: Cython being
> implemented in Python itself is an implementation detail and we should
> not bother versioning it. Just like meson. Seems reasonable to me. Well
> … if there's a chance to get rid of multiversion packages, shouldn't it
> be grabbed?

If it isn't versioned, then only one version can be installed.  It then
follows that it is an unacceptable bug if any package cares which python
version of cython is installed, or doesn't work with any of them.

Do we believe that if the only option was building it with/for the
default python version, that nothing bad would come of it?

My view without knowing the details is that history tells us that the
python world claims that things work with any version, but that it
doesn't work that way.  We certainly need to have scons with 2.7 and
with 3.x, for various users of it, as an example.