Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 27 Aug 2006 09:10:42 +1000
From:      Peter Jeremy <peterjeremy@optushome.com.au>
To:        Mike Meyer <mwm-keyword-freebsdhackers2.e313df@mired.org>
Cc:        hackers@freebsd.org, Steve Kargl <sgk@troutmask.apl.washington.edu>
Subject:   Re: amd64 questions
Message-ID:  <20060826231042.GI16768@turion.vk2pj.dyndns.org>
In-Reply-To: <17648.42078.268722.152591@bhuda.mired.org>
References:  <17648.35923.366716.65517@bhuda.mired.org> <20060826180900.GA81762@troutmask.apl.washington.edu> <17648.38296.39807.492937@bhuda.mired.org> <20060826192418.GA82155@troutmask.apl.washington.edu> <17648.42078.268722.152591@bhuda.mired.org>

next in thread | previous in thread | raw e-mail | index | archive | help

--rCb8EA+9TsBVtA92
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Sat, 2006-Aug-26 15:43:26 -0400, Mike Meyer wrote:
>In <20060826192418.GA82155@troutmask.apl.washington.edu>, Steve Kargl <sgk=
@troutmask.apl.washington.edu> typed:
>> On Sat, Aug 26, 2006 at 02:40:24PM -0400, Mike Meyer wrote:
>> > In <20060826180900.GA81762@troutmask.apl.washington.edu>, Steve Kargl =
<sgk@troutmask.apl.washington.edu> typed:
>> > > On Sat, Aug 26, 2006 at 02:00:51PM -0400, Mike Meyer wrote:
>> > > > 1) The compiler can build i386 binaries, but the toolchain in gene=
ral
>> > > >    doesn't do the right thing with the -m32 flag.
=2E..
>Yup. If you tell it -L/usr/lib32 (which gets installed if you build the
>world with WITH_LIB32 defined), it'll find that. Then complain because
>/lib/crt1.o is the 64 bit one. If I use the command line arguments:
>
>-m32 -nostartfiles /usr/lib32/crt1.o /usr/lib32/crti.o /usr/lib32/crtbegin=
=2Eo /usr/lib32/crtend.o /usr/lib32/crtn.o -L/usr/lib32
>
>simple programs build and run properly.

The incorrect linking path is fixable by doctoring the "specs" file
used by gcc (see the output of 'gcc -dumpspecs').  There have been
occasional threads in freebsd-amd64 about getting i386 mode fully
working but I don't think any of them have proceeded beyond agreeing
that there is still some work to be done in this area.  If you feel
that you have the time/skills to address some of these problems, your
input would be valued.

>> AFAIK, you can't rebuild the base system compiler with multilib
>> because it is integrated into the FreeBSD tree without the full
>> gcc configury.

I'm not sure that this is totally true because we're only talking
about i386 and amd64 - both of which are in the FreeBSD tree and
the default amd64 buildworld does build the i386 bits.

>> > > > 2) The system can run i386 binaries, but the pkg system doesn't
>> > > >    support installing packages from other architectures.

This is a known deficiency.  Again, check the -amd64 archives.  Note
that there are still problems with the emulation system: You can run
things like 'lame' successfully, but I've never managed to get (eg)
java to work.

>> > > > 3) openoffice doesn't build on amd64, and the i386 build doesn't r=
un
>> > > >    on amd64, so the recommended way to run openoffice on amd64 is =
to
>> > > >    run the Linux build.

OOo2.0 should (and generally does) build.  The entire OOo port seems
very fragile and occasionally breaks for no obvious reason.  I don't
recall ever seeing the recommendation to use the Linux build, though.
If you have problems with building OOo on a reasonably up-to-date
-stable or -current amd64 system, with an up-to-date ports tree, I
suggest you take it up on freebsd-openoffice.

--=20
Peter Jeremy

--rCb8EA+9TsBVtA92
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (FreeBSD)

iD8DBQFE8NTy/opHv/APuIcRAoCCAJ91ehIJChe7gdRMUA13FbLXWYLFMwCbBH/k
VAqUmLjgDDnNF47VVan4y9k=
=B7U1
-----END PGP SIGNATURE-----

--rCb8EA+9TsBVtA92--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060826231042.GI16768>