Re: [gentoo-dev] Cross arch ACCEPT_KEYWORDS in existence ?

[Drake Wyrm]

On Thu, 2004-04-08, 17:30:45 +0200, in
<1081438245.407570253d839@××××××××××××.be>, Jochen Maes
<sejo@g.o> wrote:
> Quoting Drake Wyrm <wyrm@×××××.com>:
>
> > On Wed, 2004-04-07, 14:29:47 -0400, in
> > <200404071429.47124.vapier@g.o>, Mike Frysinger
> > <vapier@g.o> wrote:
> > > On Wednesday 07 April 2004 02:51 pm, Meir Kriheli wrote:
> > > > Question is: Can there be more than one arch in ACCEPT_KEYWORDS
> > > > in a single system (like: "~amd64 ~x86") ? Is such state legal
> > > > at all ?
> > >
> > > should it be done ? no ... does it work ? yes ... users do it ...
> > > should it be supported ? no ...
> >
> > I disagree. Hypothetical situation: one box serving /usr to several
> > hosts of different ARCHes. Maybe we should look improving support
> > for such a situation.
>
> Serving /usr for serveral hosts of different ARCHes, like in the
> example, this would imply that fe. X would be compiled for x86, but
> the amd64 ARCH using the /usr won't be able to use that X version (in
> 64 bit) imho.

I could be wrong, but I see little difficulty in cross-compiling for
amd64 from an x86 server. Any amd64 or gcc gurus available to clear this
up? How wild and crazy different can you be with '-mcpu=foo' flags?

Irregardless, the only section under /usr required to be
arch-independent is /usr/share. FHS doesn't specify this usage (probably
doesn't allow it, either), but perhaps we could constrain the
arch-specific stuff to /usr/CPU-VEND-KERNEL-OS/{bin,lib,sbin,whatever}.
Anybody know of a good precedent regarding how to make this work?

-- 
Batou: Hey, Major... You ever hear of "human rights"?
Kusanagi: I understand the concept, but I've never seen it in action.
  --Ghost in the Shell