Re: changing CHOST in stage3 (was : [gentoo-user] default stage3)

[Richard Fish]

On 11/22/05, Matthias Langer <mlangc@×××.at> wrote:
> On Tue, 2005-11-22 at 17:11 -0500, Matthew Cline wrote:
> > On 11/22/05, Alexander Skwar <listen@×××××××××××××××.name> wrote:
> > >
> > > Why? What do you expect to gain?
> > >
> >
> > The computer I am installing this on is an old Compaq laptop with a
> > Cyrix MediaGX processor. Everything I have read suggests that this is
> > equivalent to an i586.
> >
> > Am I wrong in thinking that the CHOST variable should reflect the kind
> > of processor in the machine? Wouldn't leaving the CHOST at
> > "i386-pc-linux-gnu" build unoptimized binaries?
> >
> >
> > Matt
> >
>
> I'm not an expert, but this is just copied and pasted from the gentoo
> handbook:
>
> The CHOST variable declares the target build host for your system. This
> variable should already be set to the correct value. Do not edit it as
> that might break your system. If the CHOST variable does not look
> correct to you, you might be using the wrong stage3 tarball.
>
> Cheers, Matthias

AFAICT, CHOST is (mostly) used by portage for the "--host" argument to
the autoconf scripts, which in turn uses it to determine the path for
the tool chain.  So having CHOST=i386-blah-blah means you are using
the tool chain /usr/bin/i386-blah-blah-*.

This is also the _default_ host that gcc will build code for.  So
without any -march/-mcpu/-mtune settings in CFLAGS, you will get i386
code.  However, you can override that in CFLAGS, so that gcc produces
i586 code by default (if that is what you want).

AFAIK, there should be no difference in code produced by
"i586-pc-linux-gnu-gcc" and "i386-pc-linux-gnu-gcc -march=i586".  If
there is, then things like distcc should be horribly broken.

-Richard

-- 
gentoo-user@g.o mailing list