tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: math/py-numpy + MacOS 11.1 / Darwin 20.2.0

Brook Milligan <> writes:

> I had to use gcc10 to compile math/py-numpy on a MacOS 11.1 / Darwin
> 20.2.0 machine.  None of gcc7, gcc8, or gcc9 worked; they all failed
> to find an appropriate fortran.
> I forced gcc10 by setting GFORTRAN_VERSION=10 in mk.conf, but presumably this should be encoded in the package Makefile.
> Should this be done with GCC_REQD=10 conditional on the Darwin version?  Is there a better idea?

That's pretty intrusive.   It seems that there is something wrong that
hasn't been understood yet, and I think it's better to hold off on
kludges until at least a bit more figuring out has happened.

If anything, I would hoist this to the code that proceses
USE_LANGUAGES=fortran to do something on Darwin.

There is prior art for a different issue about 64-bit atomics, for
dealing with i386.   This isn't really the same thing, but I would much
rather see some common subexpression elimination.

or maybe this is really nust numpy.

What happens if you try to build numpy outside of pkgsrc using the
upstream build instructions?  Are there open bugs upstream?

Attachment: signature.asc
Description: PGP signature

Home | Main Index | Thread Index | Old Index