Bug#630399: DynaLoader issue with perl 5.12 on armhf when using fakechroot

[Niko Tyni]

On Mon, Jun 20, 2011 at 04:05:49PM -0500, Jonathan Nieder wrote:
> Niko Tyni wrote:
> > On Mon, Jun 20, 2011 at 04:33:59AM -0500, Jonathan Nieder wrote:
> 
> >> Should libc6 add a Breaks for this (to help with the upgrade path
> >> from squeeze)?
> >
> > Eventually yes, but I've been delaying that request because we don't
> > have the full list of broken versions on all architectures yet.
> 
> I asked because it could have helped avoid breakage in the meantime.

Yes, I probably should have given this a higher priority.

> As long as perl-base only pre-depends on libc6 from squeeze, not
> wheezy, it should be fine.

Thanks for the analysis. It looks like the low version constraint of
the libc pre-dependency helps here.

> 	unpack new perl-base
> 	unpack multiarchified libc
> 	configure multiarchified libc
> 	configure new perl-base

I wonder if there's a danger of an unrelated future change (in either
perl or eglibc) bumping the version constraint and causing problems.
However, I don't think we should worry about that overmuch at this point.

> To make the versions consistent between architectures, it should be
> enough to make a sourceful upload that Build-Depends on
> multiarch-support.

Thanks for this idea!

It isn't really obvious from the eglibc changelog, but browsing the
SVN repository confirms that the multiarch-support package was added in
2.13-5 which seems to be enough on !sparc.

 http://anonscm.debian.org/viewvc/pkg-glibc?view=revision&revision=4674

I see this isn't quite the intended use of the multiarch-support package.
Would it really work? There's no guarantee that multiarch-support
pulls in a recent enough libc6, so we'd just be relying on the general
availability of multiarch-support itself. I think arch:all packages may
well be available before the corresponding arch:any packages get built
and uploaded.

Of course, all this is probably theoretical at this point on release
architectures (unlike a week ago when the binNMUs were built) but I see
armhf is still lagging behind.

Not sure if we need to add something explicit about sparc. Probably not.
-- 
Niko Tyni   ntyni at debian.org