Re: AMD64 for sarge [Re: <rant> Package: ftpmasters, Severity: serious, ...]

[Daniel Jacobowitz <dan@debian.org>]

On Wed, Jul 07, 2004 at 04:51:08PM +0200, Josselin Mouette wrote:
> Le mer, 07/07/2004 à 10:35 -0400, Daniel Jacobowitz a écrit :
> > > In summary: The pure64 amd64 port doesn't support ia32 in the same way 
> > > other vendors do, or all that well in general really.  It's not meant 
> > > to.  The /lib vs. /lib64 hack is bad and wrong and will be going away 
> > > in favor of multiarch, a much cleaner and more elegant solution.  
> > 
> > The rest of the universe appears to disagree with the Debian AMD64
> > porters on this issue, including existing Linux distributions and
> > commercial Unices.
> 
> The rest of the universe made a move towards lib64 while already
> existing setups (namely Debian ia64) didn't use it.
> 
> > The whole time you've been discussing this I've been wondering how you
> > can make a value judgement with such limited view.  There's a lot of
> > benefit to doing it the way other people do.  Has it actually been
> > discussed with any other distributors or any standards body?
> 
> Did the other distributors discuss with us before starting with the
> lib64 idea?

You have your history quite confused; do you think that the Debian ia64
port is "old" in this regard?

Debian's SPARC port used lib64 in 1999.  Irix was using it by 1994.
Solaris was using /usr/lib/64 by 1998.  All of these were probably
earlier also; I didn't do much research.

This convention has been around for a very long time.

> On a more practical note, lib64 being a symlink doesn't affect all
> programs that don't set a RPATH. Thus we can support quite a large
> subset of existing ia32 software, without claiming LSB compliance for
> ia32, which you can easily achieve anyway by installing a i386
> distribution (or a chroot).

I was objecting to multiarch, not to the lib64 symlink.

-- 
Daniel Jacobowitz