Re: [Libreoffice] Help on fixing set_soenv so that it uses configure --build options

Hi Caolán!

On Tue, Apr 19, 2011 at 4:35 AM, Caolán McNamara <caolanm@redhat.com> wrote:
> On Mon, 2011-04-18 at 20:57 -0400, Pablo Duboue wrote:
> > Moreover, it makes it impossible to build in a 32-bit chroot running in a
> > x86-64 machine.
>
> Well,
>
> a) you can run linux32 before running configure, that's sort of my
> typical route

AHHH!!!! (Pablo gets blinded with the light of knowledge)

So it turns out I was setting up my chroot wrongly and using linux32
is all I needed.
Thanks for pointing such an obvious fact (which I didn't know).
From my point of view, this solves the issue, I can summarize this in
the bug and close it
(I'll do this in a week time if nobody objects).

> b) for the other multilibs platforms, powerpc64 and s390x, we do an
> additional @SIZEOF_LONG@ test in set_soenv.in to see if the compiler is
> in 32bit or 64bit mode. As a quick fix you could merge Linux
> if ($platform =~ m/^i[3456]86/) and elsif ($platform =~ m/^x86_64/)
> together and then switch off ('@SIZEOF_LONG@' eq '8') to take the
> LinuxX86-64Env.Set.sh branch and LinuxX86Env.Set.sh otherwise.

I somehow feel the current LibO build scripts are doing things in two
places. The parameters to the configure script plus the set_soenv
script. From a conceptual perspective, it bothers me but if I'm the
only one feeling like that, then there's no issue. This proposed
solution is still patching rather than addressing the underlining
problem (and for that, I'd rather go with a well set-up chroot
environment per your point (a)).

All in all, this will only become a problem if (and that's a huge
"if", given the size of LibO) anybody attempts to do cross-compilation
of LibO. Maybe better wait for that case to arrive rather than me
patching blindly. (In summary, I won't attempt to improve set_soenv at
this time.)

Thanks everybody for their time!

P.