Hi! On Thu, 2014-11-20 at 18:35:51 +0100, Sven Joachim wrote: > On 2014-11-20 16:51 +0100, Guillem Jover wrote: > > On Thu, 2014-11-20 at 11:43:15 +0100, Andreas Beckmann wrote: > >> Package: dpkg > >> Version: 1.17.21 > >> Severity: important > >> User: multiarch-de...@lists.alioth.debian.org > >> Usertags: multiarch > > > >> dpkg fails to configure a second architecture of this package if a > >> first architecture is already installed: > >> > >> Package: libbabl-dev > >> Multi-Arch: same > >> Source: babl > >> Version: 0.1.10-2 > >> Replaces: libbabl-0.0-0-dev > >> Provides: libbabl-0.0-0-dev > >> Breaks: libbabl-0.0-0-dev > >> > >> Both apt and aptitude consider the package to be installable, but dpkg > >> fails to configure it: > >> > >> # dpkg --configure --pending > >> dpkg: dependency problems prevent configuration of libbabl-dev:amd64: > >> libbabl-dev:i386 (0.1.10-2) breaks libbabl-0.0-0-dev and is installed. > >> libbabl-dev:amd64 (0.1.10-2) provides libbabl-0.0-0-dev. > >> > >> dpkg: error processing package libbabl-dev:amd64 (--configure): > >> dependency problems - leaving unconfigured > >> Errors were encountered while processing: > >> libbabl-dev:amd64 > > > > This is expected behavior in dpkg. Conflics/Breaks/Replaces get an > > implicit "any" arch qualifier, so those make the dependency > > unsatisfiable due to the Provides. > > Are you sure about that? How come I have libncurses5-dev:amd64 and > libncurses5-dev:armhf coinstalled then, when they both > Provides/Conflicts/Replaces libncurses-dev? It seems to me that dpkg > actually treats Breaks different from Conflicts here.
Oh, so it does. :( I've started looking into fixing this, but I'll have to ponder about it to maybe change the behavior the other way around :/, because as it stands, and as pointed out by Johannes off-BTS there are at least 114 M-A:same possibly affected packages, so this is really way off to even consider breaking during the freeze. So this is something for 1.18.x. Apt probably should also hold off fixing anything until dpkg behaves in a consisten way, in either direction. Thanks, Guillem -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
