[Bug 268348] math/py-numpy: introduce FORTRAN / NOBLAS options

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=268348

Charlie Li <vishwin@freebsd.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|0mp@FreeBSD.org             |python@FreeBSD.org
              Flags|maintainer-feedback?(python |
                   |@FreeBSD.org)               |
             Status|In Progress                 |Closed
         Resolution|---                         |Not Accepted

--- Comment #20 from Charlie Li <vishwin@freebsd.org> ---
(In reply to Mateusz Piotrowski from comment #19)
> I've reworked the original patch and I wonder if you could take another look.
Absolutely not.

While we can add however many warnings and notes about how going without
Fortran/BLAS works but is not performant, the general expectation amongst this
userbase, ports or not, assumes that some Fortran/BLAS implementation is always
available. This is simply one of those situations that can garner
negativity/pessimism on FreeBSD as a scientific computing platform, far beyond
any reduced dependency footprint, even if it is a pilot error from not reading
warnings (and nobody in this case will actually read and heed them). The
significant use of numpy and ecosystem amongst academia, particularly those who
do not have the cycles to work through pilot errors like this, is another
factor.

Please refer to comment 15 as to why GCC is a runtime dependency. It does not
have to be with an alternative Fortran compiler.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.