cannot handle b-d [!linux-any]

[Matthias Klose]

Binary package hint: germinate

see gvfs, b-d on hal:

   libhal-dev (>= 0.5.10) [!linux-any]

but hal shows up in component-mismatches

[Colin Watson (cjwatson)]

apt (0.7.26~exp8) experimental; urgency=low

[...]
  [ Julian Andres Klode ]
[...]
  * apt-pkg/deb/deblistparser.cc:
    - Handle architecture wildcards (Closes: #547724).
[...]

 -- Michael Vogt <email address hidden> Fri, 09 Jul 2010 19:16:20 +0200

Backporting apt to cocoplum is probably a non-starter. I'm not sure exactly what to do here. Perhaps we need to work around it in germinate somehow.

[Colin Watson (cjwatson)]

>>> import apt_pkg
>>> apt_pkg.init_config()
>>> apt_pkg.config.set("APT::Architecture", "i386")
>>> apt_pkg.init_system()
>>> apt_pkg.parse_src_depends("hal [!linux-any]")
[[('hal', '', '')]]

(Succeeds on Natty.)

[Colin Watson (cjwatson)]

Actually, the apt patch isn't that enormous. Maybe it would be best to backport it.

[Colin Watson (cjwatson)]

The attached patch backports what look like the relevant commits from current apt:

        revno: 1327.24.434
        fixes bug(s): http://bugs.debian.org/547724
        committer: Julian Andres Klode <email address hidden>
        branch nick: debian-experimental-ma
        timestamp: Sat 2010-06-26 20:56:44 +0200
        message:
          * apt-pkg/deb/deblistparser.cc:
            - Handle architecture wildcards (Closes: #547724).

        revno: 1327.24.436
        committer: Julian Andres Klode <email address hidden>
        branch nick: debian-experimental-ma
        timestamp: Sat 2010-06-26 23:01:49 +0200
        message:
          apt-pkg/deb/deblistparser.cc: Fix bug in architecture wildcard support.

          Previously, linux-any was always matched, because the code simply
          appended linux- to the APT::Architecture value. Now, it does this
          only if the APT::Architecture value does not contain "-".

        revno: 1327.24.437
        committer: Julian Andres Klode <email address hidden>
        branch nick: debian-experimental-ma
        timestamp: Sun 2010-06-27 21:04:53 +0200
        message:
          deblistparser: Special-case *-armel, lpia and powerpcspe architectures.

        revno: 1327.76.1
        author: Sebastian Andrzej Siewior <email address hidden>
        committer: David Kalnischkies <email address hidden>
        branch nick: apt-sid
        timestamp: Thu 2010-10-07 14:50:25 +0200
        message:
          * apt-pkg/deb/deblistparser.cc:
            - support ArmHardFloat port in CompleteArch, thanks to Sebastian
              Andrzej Siewior for the patch!

Michael, does this seem reasonable? Do you think we might be able to backport this to lucid-proposed?

[Colin Watson (cjwatson)]

With this patch, in a python session:

>>> import apt_pkg
>>> apt_pkg.init_config()
>>> apt_pkg.config.set("APT::Architecture", "i386")
>>> apt_pkg.init_system()
>>> apt_pkg.parse_src_depends("hal [!linux-any]")
[]

Of course, apt's general behaviour would need to be regression-tested.

[Martin Pitt (pitti)]

Accepted apt into lucid-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Martin Pitt (pitti)]

I have used the proposed apt in my lucid chroot for 1.5 weeks now, and haven't noticed any regressions. apt-cache, apt-get install/purge etc. work fine.

[Martin Pitt (pitti)]

I also confirm that the test case now succeeds with the proposed version.

[Colin Watson (cjwatson)]

With apt 0.7.25.3ubuntu9.5:

>>> import apt_pkg
>>> apt_pkg.init_config()
>>> apt_pkg.config.set("APT::Architecture", "i386")
>>> apt_pkg.init_system()
>>> apt_pkg.parse_src_depends("hal [!linux-any]")
[]
>>> apt_pkg.parse_src_depends("hal [i386]")
[[('hal', '', '')]]
>>> apt_pkg.parse_src_depends("hal [powerpc]")
[]
>>> apt_pkg.parse_src_depends("hal [linux-i386]")
[[('hal', '', '')]]
>>> apt_pkg.parse_src_depends("hal [linux-any]")
[[('hal', '', '')]]
>>> apt_pkg.parse_src_depends("hal [any-i386]")
[[('hal', '', '')]]
>>> apt_pkg.parse_src_depends("hal [linux-any], foo [any-powerpc], bar [hurd-any]")
[[('hal', '', '')]]
>>> apt_pkg.parse_src_depends("hal [linux-any], foo [any-powerpc], bar [!hurd-any]")
[[('hal', '', '')], [('bar', '', '')]]

This all looks correct. I also did an upgrade of my lucid chroot to a bunch of batched-up stuff from lucid-updates:

The following packages will be upgraded:
  dhcp3-client dhcp3-common initscripts libdbus-1-3 libgssapi-krb5-2
  libk5crypto3 libkrb5-3 libkrb5support0 libldap-2.4-2 libpam-modules
  libpam-runtime libpam0g libtiff4 libtiff4-dev libtiffxx0c2 linux-libc-dev
  logrotate ntpdate openssh-client perl perl-base perl-modules sysv-rc
  sysvinit-utils tzdata x11-xserver-utils
26 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.

... and that succeeded. I'm happy to call this good.

[Launchpad Janitor (janitor)]

This bug was fixed in the package apt - 0.7.25.3ubuntu9.5

---------------
apt (0.7.25.3ubuntu9.5) lucid-proposed; urgency=low

  [ Michael Vogt ]
  * debian/apt.cron.daily:
    - source /etc/default/locale (if available) so that the
      apt-get update cron job fetches the right translated package
      descriptions (LP: #652951)

  [ Loïc Minier ]
  * Backport r2014.1.2 (see below); fixes bug in APT when running the
    linaro-image-tools testsuite; LP: #709895.

    [ David Kalnischkies ]
    * apt-pkg/acquire-item.cc:
      - use also unsigned Release files again (Closes: #596189)

  [ Colin Watson ]
  * Backport r1327.24.434, 1327.24.437, 1327.76.1 (LP: #787500)
 -- Michael Vogt <email address hidden> Tue, 24 May 2011 18:23:29 +0200