GCC Invokation, Specifically the -m switch with regard to ix86

[James <james at thesoftwaresmith dot com dot au>]

Hi GCC Developers.

I'm using egcs-2.91.66 on a Linux 2.1.12 kernel (RedHat 6.1)

Recently I have been rebuilding source RPM's for GNU/Linux. I've been
trying to optimize the packages for the i686, as this is the
architecture I expect to run my packages on. Many packages choke when
given an architecture higher than a 486. The problem shows up in the
configure script as the gcc compiler not working.

What I believe happens is the script pulls the architecture from uname
or something similar. On my build system this returns i686. It uses this
value and modifies it to an -m switch, and the compiler (gcc) is called
by configure as gcc (flags)-m686 (mumble) and the configure script
fails. When I force the architecture to below 486, the build is
successful, configure continues, because -m486 is recognized by gcc.

Because the build must be automatic, I can't change the config.guess
each time a package is rebuilt, and changing the configure script in
494+ SRPM's is a bit daunting (disregarding updates etc.)

What would be really useful is if options like -m586 and -m686 were
automatically mapped to -mpentium and -mpentiumpro. I am guessing that
this would involve adding some kind of OR in the recognition statements.
This would make it much easier for me to build the SRPM's in an
optimized state.

Thank you
James Vandenberg