MAINT: NumPy complex shim needed? about pythran

[rgommers]

This was already fixed in gh-2128. If you use a numpy nightly, you should temporarily install Pythran from latest master, just like the SciPy pre-release job does.

There's nothing else to do here I think, aside from waiting for a new Pythran release and bumping the minimum version needed in SciPy.

[ngoldbaum]

pythran/pythonic/types/complex.hpp:231:15: error: request for member ‘real’ in ‘val’, which is of non-class type ‘__complex__ long double’

ping @lysnikolaou

Could this be related to changes to numpy's wrappers for the C99 complex types?

See numpy/numpy#24085

[tylerjereddy]

To be fair, the pythran master branch testsuite failed for me locally with NumPy main so I just gave up on it--I guess I should have tried using it anyway.

111 failed, 3402 passed, 56 skipped, 65 warnings in 2603.89s (0:43:23)

Many of the test failures proceed through distutils.errors.CompileError, but I think the problem is that pythran can't dig around for my OpenBLAS without using a ~/.pythranrc.