You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I do not really know where it comes from. It became an issue running the Matplotlib CI on FreeBSD (and is readily solved by restricting to numpy<1.24).
Edit: this is a runtime issue. It compiles fine, but then, upon loading(?) it errors out as it cannot find csinl.
Error message:
No response
NumPy/Python version information:
1.24
Context for the issue:
I understand that this may not be a highly prioritized issue, but thought it was worth reporting in case it can be solved in an easy way. (And to some extent it should probably be solved in FreeBSD...)
It isn't immediately clear to me which of the cleanup npy_math cleanup PR removed the relevant fallback, seems likely that fallback was just a fallback to double-precision anyway.
Activity
rgommers commentedon Dec 21, 2022
@oscargus can you please add the full traceback you are seeing?
seberg commentedon Dec 22, 2022
It isn't immediately clear to me which of the cleanup npy_math cleanup PR removed the relevant fallback, seems likely that fallback was just a fallback to double-precision anyway.
rgommers commentedon Dec 22, 2022
FreeBSD is missing a longer list of
complex long doubleC99 functions, documented in https://wiki.freebsd.org/Numerics. Quote "A major focus of ongoing work is on implementing the remaining long double and complex functions specified by the C11 standard but not yet present in libm."So we should revive all the missing ones .... sigh
python-numpy: Remove dep on libandroid-complex-math
python-numpy: Remove dep on libandroid-complex-math