Re: enable n32 and n64, and move o32 into mips/mips32

[Alexandre Oliva <aoliva at redhat dot com>]

On Mar 20, 2003, Roland McGrath <roland at redhat dot com> wrote:

> Tough beans.  The same is true of sparc or x86 with -m32 and -m64, etc.

'cept glibc doesn't support them as such, since there's a different
machine name that implies the -m32 ABI.

> The config.sub canonical tuple is used as a key to many things where it
> must correspond exactly to an ABI--that's what it's for.

Hmm...  Maybe I could arrange to have the ABI detected by config.sub
and added to the canonical triplet, even though it's not in the --host
command line.  This would be ideal from my perspective.  I'll look
into this possibility.

> It is by far more reasonable to add compiler flags based on the
> tuple than to use compiler settings instead of just the tuple to
> choose the configuration.

Point is mips64-linux-gnu isn't enough to determine the
configuration.  I'm proposing that we get aid from the compiler to
determine which ABI it defaults to.

> The former I object to on the grounds of consistent GNU configure
> conventions.  Those require that $CC as given/defaulted be a
> compiler that produces correct code for the host tuple
> given/defaulted.

Right.  But when the tuple covers multiple ABIs, we rely on the
compiler, like all other tools do.

> If you want to get AC_PROG_CC changed in Autoconf so that it does
> some mips hooey universally, I would not object to that.

Doing it in AC_PROG_CC would be silly.  config.guess/config.sub feels
like the right place for this.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva at {redhat dot com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva at {lsd dot ic dot unicamp dot br, gnu.org}
Free Software Evangelist                Professional serial bug killer