Nicolas Mailhot wrote:
Le lundi 31 juillet 2006 à 22:14 +1000, John Pye a écrit :
Hi Nicholas,
I'm with you on the frowning on this practise. But until a segfault in
gfortran.so.1 (gcc 4.1.x) is fixed, I can't build an important package
that I need on Fedora Core 5. Just not possible.
BTW will it work petter with FC6 ?
I don't expect so. It's a bug with the gfortran compiler, and
gcc-gfortran 4.1.x is pretty much bleeding edge already, as I understand it.
To add a little more background: the troublesome code here is an
optional closed source (but nevertheless very desirable) plugin from an
outside developer.
Filing under "proprietary crap" then?
No way dude :-) The proprietary crap in question is a really, really
well respected solver component used by a whole stack of commercial
mathematical modelling packages. Unfortunately its gfortran that's in
danger of getting a slamming here. gfortran has been progressing really
fast though, and I reckon it will be up to scratch pretty soon. But even
the developers of gfortran still urge people to stick with g77 if they
can, so it's not a broken promise. There have been commercial Fortran-90
and Fortran-95 compilers for, um I guess 15 years or so, and
unfortunately a lot of good math software depends them at present.
The alternative is to package up the FC4 libfortran.so.0 file in a new
libgfortran package (eg libgfortran0-4.0.2-0.i386.rpm) that is named
differently such that updates to the official libgfortran don't
conflict/replace the special FC4 libgfortran package.
That still sucks but is a tad better
Well that might be the best option then.
--
John Pye
Department of Mechanical and Manufacturing Engineering
University of New South Wales, Sydney, Australia
http://pye.dyndns.org/
_______________________________________________
Rpm-list mailing list
Rpm-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/rpm-list
