Re: -= PROPOSAL =- Release sarge with amd64
> Raul Miller <moth@debian.org> writes:
>
> > On Sat, Jul 17, 2004 at 05:49:41PM +0200, Goswin von Brederlow wrote:
> >> [d] multiarch systems where each arch has advantages over the others
> >
> > This should be outside the scope of package management. It's a build
> > time decision, not a system configuration decision.
On Sat, Jul 17, 2004 at 09:26:37PM +0200, Goswin von Brederlow wrote:
> Handling 64bit support for 32/64bit archs just like the cmove or mmx
> feature of i386 is just wrong. They don't compare. You can't replace
> the 32bit libc with the better 64bit libc, you need both.
So you replace it with both.
> You are talking about completly different ld.so, opcodes, abis,
> syscalls, parameter passing conventions, cpu features, size of types,
> .... Not just a few extra opcodes.
Your current system has both anyways, so this isn't a huge difference.
> > [That said, if we want to support "build packages from source with custom
> > options, that makes it viable to support this kind of fine control.
> > Though there's likely to be some cost in user confusion.]
>
> multiarch could be used to create (very partial) i586, i686,
> k7,... ports and have them transparently be used depending on the cpu
> type. The biarch discussions had this in mind but it wasn't used for
> the multiarch proposal. With multiarch you wouldn't be able to use the
> i686 libc in place of the i386 libc. dpkg wouldn't undertsand that
> they are the same ABI.
So what does it mean to install an amd64 gcc on an m68k system?
--
Raul
Reply to: