The test didn't fail on the original PR but now often do. That is:

test_ufunc_noncontiguous

fails for ldexp and frexp.

The failures are harmless w.r.t. to unrelated changes: They are not new, the test is just new.

@r-devulap do you have time to have a look at fixing ldexp and frexp? Not sure what is best, aligning to use npyv_loadable_stride_##sfx/npyv_storable_stride_##sfx, or just adding the alignment check that some of the other macros already have here.
The choice of the max stride is also slightly different, although I think compatible.

Originally posted by @seberg in #27027 (comment)