Re: multiarch and interpreters/runtimes
On Sat, 2013-05-04 at 06:36:31 +0200, Matthias Klose wrote:
> Am 19.04.2013 00:33, schrieb Guillem Jover:
> I think the full-multiarch support for python in
> > experimental should really be reverted.
> No. This is backward, and the wrong way to go forward.
Sorry, but the way to go forward is not to subvert the dependency
system with metadata that can allow breaking installed systems.
> I do acknowledge that
> there are issues with the current state of dpkg, but I'm not seeing how you are
> planning to address these for jessie, and if this can be used for the jessie
> release, and not only in jessie+1. There are other ways to address the issue
> you did raise, but they could be considered as ugly, but in the end I would
> prefer doing with an ugly plan B, if we don't have a plan A.
I'd be interested to hear about those ugly solutions, from where I
sit though, this is simply not a matter of ugly, but wrong.
> You always can remove the Multi-Arch attribute from packages, but developers
> should continue to prepare interpreters for a multiarch environment.
Switching interpreters that allow them to be fully multiarch should be
done and played with, uploading those that can break the system to
Debian unstable should not be considered acceptable.
Also I did say, “full-multiarch support” should be reverted, which
implies partial multiarch support could still be deployed w/o subverting
the dependency system.