GCC build failed with your patch on 2001-01-09T11:35:00Z.

[Geoff Keating]

> Cc: rth@redhat.com, gcc@gcc.gnu.org, gcc-regression@gcc.gnu.org,
>         grahams@redhat.com, jsm28@cam.ac.uk, mhayes@redhat.com,
>         neil@daikokuya.demon.co.uk, osk@hem.passagen.se
> From: Alexandre Oliva <aoliva@redhat.com>
> Organization: GCC Team, Red Hat
> Date: 09 Jan 2001 17:14:21 -0200

> On Jan  9, 2001, Geoff Keating <geoffk@geoffk.org> wrote:
> 
> > powerpc-eabi is a 32-bit target.  I don't want this patch.  I've
> > carefully not made this change for several years, because it will make
> > code generation worse in the common case.  How can a backend override
> > the change?
> 
> Also note that you should be able to generate identical code for a
> powerpc-eabi target on a platform that has 64-bit longs (and they'd be
> used as HOST_WIDE_INT regardless of LONG_TYPE_SIZE), so it's probably
> a good idea to work on having powerpc-eabi work in such a
> configuration, instead of just punting at it by requiring a
> HOST_WIDE_INT that is 32-bits wide.

Last time I tried this, I stopped after working at it for two days,
when it was clear that the resulting patch would cause so much
instability in the tree that it would need weeks of testing.

For example, I believe one of the things it requires is the patch that
Jeff Law says breaks PA.

Now, perhaps things have changed since then, and I'm happy to try
again, if you're willing to do the testing on non-powerpc platforms
that it will require.  We'll probably get a much more stable compiler
out of it.

I'm not sure, though, that it's a good idea to do this just days
before we branch for a release.

-- 
- Geoffrey Keating <geoffk@geoffk.org>