[SciPy-user] GCC 3.3 requirement

mfmorss@aep.com mfmorss at aep.com
Fri Mar 10 14:15:59 CST 2006

Exactly, but even for compiling C code, gcc-4.0 is less trustworthy.

The last verson of gcc that was produced before gcc-4.0 was, if I'm not
mistaken, gcc-3.4.5.  Not gcc-3.3.  I would recommend gcc-3.4.5 for
building anything, not just numpy and scipy.  I was surprised to learn here
that gcc-4.0 is shipped with SuSE.

Also unless I'm mistaken, there is no need to use a Fortran compiler more
advanced than g77 to compile the math libs, which are written in Fortran

I don't really see any problem with not being able or willing to use more
recent versions of gcc, since for the time being, the older version is more
stable.  Gcc-3.4.5 (including g77) isn't going to go away, and if you don't
have it, it's not a huge problem to build it.  My belief is that the
compiled distributions of Linux, for example, still all use gcc-3.  I
notice, for example, that Gentoo does not mark gcc-4 "stable" for a build
on any system (http://packages.gentoo.org/search/?sstring=gcc).  And Lunar
is still based on 3.4.5.

Mark F. Morss
Principal Analyst, Market Risk
American Electric Power

             <humufr at yahoo.fr>                                             
             Sent by:                                                   To 
             scipy-user-bounce         SciPy Users List                    
             s at scipy.net               <scipy-user at scipy.net>              
             03/10/2006 01:53                                      Subject 
             PM                        Re: [SciPy-user] GCC 3.3            
             Please respond to                                             
             SciPy Users List                                              
             <scipy-user at scipy                                             

One big problem with gcc4, especially 4.0 is that gfortran is buggy. A
lot of improvement has been done but in compariason with commercial
compiler or g95, it's not the same things. gfortran is an alpha version
of a true fortran 90 compilers. So many things in scipy are using
fortran so this is a critical issue for the moment.


Robert Hetland wrote:
> Apple shipped a crippled version of GCC  3.3 with the new intel Macs.
> (It seems like it is there -- there is a package, and you can
> gcc_select it -- but it does not work properly.)
> Many on this list have discouraged using 4.0 for building scipy.  I
> have also found that 3.3 works on the *PPC* Mac, where 4.0 does not.
> However, it seems inevitable that mac users will need to compile with
> 4.0 as more users get new Intel based machines.  I am sure that
> eventually someone will provide an unofficial gcc 3.3 package (like
> Guarav Khanna who maintains hpc.sf.net), but I don't think this is a
> good long-term solution.
> What are the issues involved?  Why do we need gcc 3.3?  Is there some
> sort of timescale for moving the code base up to GCC 4.0 (or 4.x)
> compliance?
> -Rob.
> -----
> Rob Hetland, Assistant Professor
> Dept of Oceanography, Texas A&M University
> p: 979-458-0096, f: 979-845-6331
> e: hetland at tamu.edu, w: http://pong.tamu.edu
> ------------------------------------------------------------------------
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.net
> http://www.scipy.net/mailman/listinfo/scipy-user

SciPy-user mailing list
SciPy-user at scipy.net

More information about the SciPy-user mailing list