[25728] trunk/dports/python/py-numpy

[Mark Moll]

On May 30, 2007, at 7:44 AM, Jochen Küpper wrote:
> On 30.05.2007, at 05:06, source_changes at macosforge.org wrote:
>
>> updated to py-numpy version 1.0.3 swapped dependency from gcc41/ 
>> gcc42 to g95, much to the joy of everyone
>
> Not for me!
> I very much prefer gfortran over g95 for various reasons --  
> performance and political ones.
>
> Why don't you create variants for either fortran compiler, or use  
> the one that is already installed?

Also, gcc42 seems to be the de facto choice for a fortran compiler  
for almost every other port:

 > grep depends `grep -rl fortran */*/Portfile`|grep gcc42
graphics/pgplot/Portfile:       depends_lib-append      port:gcc42
lang/gcc42/Portfile:depends_lib         port:gmp port:mpfr port:libiconv
lang/gcc42/Portfile:    depends_lib-append      bin:odas:odcctools
math/R/Portfile:    depends_lib-append      port:gcc42
math/arpack/Portfile:depends_build port:gcc42
math/octave-forge/Portfile:             depends_lib-append       
port:gcc42
math/octave/Portfile:           depends_lib-append      port:gcc42
science/libnc-dap/Portfile:        depends_lib-append      port:gcc42
science/openmpi/Portfile:        depends_build port:gcc42

 > grep depends `grep -rl fortran */*/Portfile`|grep g95
math/fftw-3/Portfile:   depends_lib-append port:g95
math/octave-forge/Portfile:     depends_lib-append port:g95
math/octave/Portfile:   depends_lib-append      port:g95
science/libnc-dap/Portfile:    depends_lib-append      port:g95

In the ports that use g95 it is used as a variant. Perhaps one  
fortran compiler should be the "*officially* recommended fortran  
compiler," while the others can be made available in variants. I'd  
vote for gcc42 to provide the official fortran compiler, since it's  
already the de facto standard, plus---as Jochen writes---there are  
performance and political reasons to favor gcc42.

-- 
Mark