[Pkg-ia32-libs-maintainers] Bug#540303: Bug#540303: ia32-libs: seems like I figured out why this happened

[Goswin von Brederlow]

reassign 540303 fakeroot
thanks

Dmitri Gribenko <gribozavr at gmail.com> writes:

> Package: ia32-libs
> Severity: normal
>
>
> Hi,
>
> I'm on testing and after libc update I have only fakeroot's files left in
> /emul.  I've got fakeroot 1.12.4 installed and it Conflicts with libc6-dev-i386
> <= 2.9-13.  But I don't have libc6-dev-i386 installed, only libc6-i386.  So,
> when fakeroot update has migrated to testing (in June), I installed it and
> didn't notice anything, until now.  So I think that the problem is that
> fakeroot 'Conflicts' on the wrong package.  Does this sound reasonable?

At least it is missing a "Conflicts: libc6-i386 (<= 2.9-18)". Not sure
about the libc6-dev-i386. From the version that seems unrelated.

To give some background:

When libc6-i386 transitioned /usr/lib32 it became neccessary that the
preinst of libc6-i386 is run before any other lib32* package
is unpacked. As solution I suggested "Pre-Depends: libc6-i386 (<=
2.9-18)". But in fakeroot that would mean that all of a sudden
fakeroot would pull in libc6-i386.

As it turns out the Pre-Depends actually triggers a bug in apt/aptitude
on upgrades where apt-aptitude would refuse to upgrade without
running "apt-get install libc6-i386" first. Confering with the dpkg
maintainer revealed that a "Conflicts: libc-i386 (<= 2.9-18)" has the
same effect (run preinst of libc6-i386 before unpacking any lib32*
package) without triggering the apt/aptitude bug.

> If my thoughts a correct, then fakeroot is now partially broken on a number of
> systems (that is, it is not my local problem) and something should be done with
> this.  Although I can erinstall it manually, it would be better if it would be
> done automatically for all affected users.
>
> Best regards,
> Dmitri Gribenko

Yes.

MfG
        Goswin