Coverage: 99.424% (-0.08%) from 99.505% when pulling 9c9df06 on eriknw:remove_scipy_matrix into b7e9485 on python-graphblas:main.

Oh, interesting test failures (randomized testing ftw?!). This is using python-suitesparse-graphblas from upstream, which supports complex on Windows.

It looks like asin(4 + 0j) gives the wrong results on Windows in SuiteSparse:GraphBLAS. This is a known limitation. From SuiteSparse:GraphBLAS documentation:

5. On Windows, the casin, casinf, casinh, and casinhf functions provided by Microsoft do not return the correct imaginary part. As a result, GxB_ASIN_FC32, GxB_ASIN_FC64 GxB_ASINH_FC32, and GxB_ASINH_FC64 do not work properly on Windows. This affects the GrB/asin, GrB/acsc, GrB/asinh, and GrB/acsch, functions in the MATLAB interface. See the MATLAB tests bypassed in gbtest76.m for details, in the GraphBLAS/GraphBLAS/test folder.

So, we'll need to update tests to ignore these on Windows.

I'm not sure what's up with the error with ldexp. Need to print more info.

Heh, so we may have more places where we use if not _supports_complex instead of if is_win. If so, we'll fine them eventually.

I'm going to go ahead and merge this since it contains a fix, and I don't think the removed deprecations are particularly significant.

Changes unknown
when pulling 0213c4b on eriknw:remove_scipy_matrix
into ** on python-graphblas:main**.