Debian Bug report logs - #761006
dpkg: provide a way to install MultiArch:same packages with different versions

Reply or subscribe to this bug.

Display info messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Lucas Nussbaum <lucas@debian.org>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: dpkg: provide a way to install MultiArch:same packages with different versions

Date: Tue, 09 Sep 2014 21:42:13 +0200

Package: dpkg
Version: 1.17.13
Severity: wishlist

Hi,

I was trying to cross-grade an amd64 system to x32.
the libpcre3 package on debian-ports has a patch to work-around #760327.
As a result, the debian-ports package has a different version, so I ran
into:

Preparing to unpack libpcre3_1%3a8.35-3+x32.1_x32.deb ...
Unpacking libpcre3:x32 (1:8.35-3+x32.1) over (1:8.35-3+x32.1) ...
dpkg: error processing package libpcre3:x32 (--install): package libpcre3:x32 1:8.35-3+x32.1 cannot be configured because libpcre3:amd64 is at a different version (1:8.35-3)

It would be great if dpkg had a --force option to override this error.

Lucas


-- System Information:
Debian Release: jessie/sid
  APT prefers testing
  APT policy: (990, 'testing'), (800, 'stable'), (300, 'unstable'), (150, 'experimental')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 3.14-2-amd64 (SMP w/4 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

Versions of packages dpkg depends on:
ii  libbz2-1.0   1.0.6-7
ii  libc6        2.19-9
ii  liblzma5     5.1.1alpha+20120614-2
ii  libselinux1  2.3-1
ii  tar          1.27.1-2
ii  zlib1g       1:1.2.8.dfsg-1

dpkg recommends no packages.

Versions of packages dpkg suggests:
ii  apt  1.0.6

-- no debconf information

Message #10 received at 761006@bugs.debian.org (full text, mbox, reply):

From: Guillem Jover <guillem@debian.org>

To: Lucas Nussbaum <lucas@debian.org>, 761006@bugs.debian.org

Subject: Re: Bug#761006: dpkg: provide a way to install MultiArch:same packages with different versions

Date: Wed, 25 Feb 2015 12:29:27 +0100

Hi!

On Tue, 2014-09-09 at 21:42:13 +0200, Lucas Nussbaum wrote:
> Package: dpkg
> Version: 1.17.13
> Severity: wishlist

> I was trying to cross-grade an amd64 system to x32.
> the libpcre3 package on debian-ports has a patch to work-around #760327.
> As a result, the debian-ports package has a different version, so I ran
> into:
> 
> Preparing to unpack libpcre3_1%3a8.35-3+x32.1_x32.deb ...
> Unpacking libpcre3:x32 (1:8.35-3+x32.1) over (1:8.35-3+x32.1) ...
> dpkg: error processing package libpcre3:x32 (--install): package libpcre3:x32 1:8.35-3+x32.1 cannot be configured because libpcre3:amd64 is at a different version (1:8.35-3)
> 
> It would be great if dpkg had a --force option to override this error.

I'm afraid what you want is not possible, or put it another way, it's
not possible to get what you want and get sane results at the same
time. The reference counting code uses the package version of each
Multi-Arch:same instance to know when it can replace files or when the
packages are getting in sync and as such need to be exactly the same.

If this was a matter of different file contents that needed to be
overwritten, that would be fine, because the configure check that
guarantees that all instances have the same version would somehow
mitigate such disasters.

But if we had such force option (to disable the version sync check at
configure time), then you'd need to use it all the time, as long as
you have such packages with diverging versions, and then dpkg would be
unable to tell when it should replace or not the reference counted
files when unpacking multiple Multi-Arch:same instances on
upgrades/downgrades for example. I assume apt would also need an
option to override that, or it might get really unhappy.

I will probably add it anyway, because dpkg --force options have
historically pretty much allowed to destroy your own system, but just
wanted to mention the (possibly) unexpected consequences.

Thanks,
Guillem

Send a report that this bug log contains spam.