This is the mail archive of the
mailing list for the GCC project.
Re: SH Linux: remove big endian multilib
On Mon, Aug 06, 2001 at 02:16:32PM +1000, Fergus Henderson wrote:
> On 05-Aug-2001, M. R. Brown <firstname.lastname@example.org> wrote:
> > * NIIBE Yutaka <email@example.com> on Sun, Aug 05, 2001:
> > >
> > > I agree with this more or less. Using specified one is better and
> > > simple. But the definition of sh-unknown-linux-gnu is supporting four
> > > targets.
> > According to the people here, sh-unknown-linux-gnu specifies sh3,
> > little-endian. You're the only person I've seen that says it represents
> > something else.
> Doesn't sh3-linux-gnu also specify sh3 little-endian?
> If so, what's the point in having two configuration names
> for the same configuration?
> Even if sh-unknown-linux-gnu previously meant sh3 little-endian,
> wouldn't it be more useful to do as NIIBE Yutaka has suggested,
> and make sh-unknown-linux-gnu be the multilibbed configuration,
> even if that requires installing libc multiple times?
Something worth keeping in mind here is that most things do not need to
know what kind of processor they are being configured for. For
powerpc, for instance, we can just configure gcc with --with-cpu (when
it's working, see PR target/3722... please do, in fact). GCC and glibc
may also need --without-fp. Then everything else builds for
powerpc-unknown-linux-gnu, and does not need to know more. That's the
way we're currently managing our SH tools here. We configure apps for
sh-unknown-linux-gnu (because that's the only one that most config.*
scripts recognize!) and use patches to tell the tools which particular
SH to target.
(Of course, i?86 does it exactly the opposite way...)
Daniel Jacobowitz Carnegie Mellon University
MontaVista Software Debian GNU/Linux Developer