Multilib Middle-Ground

[seth vidal]

On Fri, 2008-05-02 at 02:01 -0500, Callum Lerwick wrote:
> On Thu, 2008-05-01 at 13:18 -0500, Chris Adams wrote:
> > Once upon a time, Callum Lerwick <seg at haxxed.com> said:
> > > WINE, which has legitimate open source purposes. Sorry, some of us pay
> > > our rent by developing open source that runs on Windows. I'm not going
> > > to live in my car for the sake of ideological purity.
> > 
> > Is there a bug that prevents "yum install wine" from pulling in the 32
> > bit libraries as needed?
> 
> That wasn't your question. I was answering your question and countering
> your underlying assertion that i386 compatibility is ideologically
> impure and only useful for closed source software. A MinGW
> cross-compiler toolchain and WINE provide a complete Windows development
> environment with absolutely no closed source involved, providing tools
> with which to wean people off closed source software one application at
> a time.
> 
> Want another reason? Being able to develop and test i386 builds on my
> nice fast uber-RAM x86_64 box is incredibly useful. Not all of my
> machines are x86_64, and i386 is still the dominant architecture
> overall, so any sane developer HAS to make sure it works. Seriously,
> Fedora's ease of i386 compatibility is a *major* selling point for us.
> Go look around at the twisted flakey chroot hacks Debian/Ubuntu have to
> go through to get what we get with ease.

You realize that:
 1. we just changed the default from 'install all/both arches' to
'install only the 'best' arch for this system', right
 2. You can set this back by setting multilib_policy=all
 3. This does not impact depsolving at all. Any package which requires
an i386 dependency will get it pulled in.

The concern that started this thread is that since we changed the
default behavior the sky will fall down and the earth will tremble.
That's the bit I don't understand.

-sv