[Piuparts-devel] Bug#676694: Bug#676694: piuparts: [install-upgrade] should first install from distribution

Luca Capello luca at pca.it
Mon Jun 11 20:07:16 UTC 2012

Hi there!

On Sun, 10 Jun 2012 13:00:44 +0200, Holger Levsen wrote:
> On Freitag, 8. Juni 2012, Luca Capello wrote:
>> When doing an install-upgrade test, the apt-get package installation
>> should be forced from the first distribution set via -d (using the
>> apt-get -t option).  Otherwise, if the package being tested depends on
>> the same binary version of other packages, then installation from
>> official sources is not possible.  Git patch attached.
> Your patch modifies install_packages_by_name() to unconditionally use the 
> first distribution which makes a lot of other cases fail...

I thought about that after having sent the bug report (the first idea
being to add another argument to install_packages_by_name()...) and also
got caught later when trying to tests Bacula's -dbg packages
(install-purge failed because there is no such a package in sid).  But
you were quicker to disclose it ;-)

>> I encountered this bug while testing the new bacula-console package,
> have you tried latest piuparts from git branch develop?

No, because I thought that simply using piuparts.py would have not been
enough.  And I was/am a bit too lazy to rebuild it (I am trying to
upload Bacula ASAP, see #639466).

>> Piuparts logs are available upon request or simply grab the testing
>> package at <http://pkg-bacula.alioth.debian.org/tmp/> ;-)
> please send me the log, 
> http://pkg-bacula.alioth.debian.org/tmp/bacula_5.2.6+dfsg-1_all.deb.piuparts
> is not what I'm looking for :)

This is why I wrote "grab the testing package", what you want could be:x


On Sun, 10 Jun 2012 21:18:16 +0200, Andreas Beckmann wrote:
> On 2012-06-08 23:17, Luca Capello wrote:
>> When doing an install-upgrade test, the apt-get package installation
>> should be forced from the first distribution set via -d (using the
> What is the exact command line failing?
> There should be only one package source available during the test.

I do not see why there should be only one package source, especially for
those packages depending on another package from the same source package
and with a strict dependency (as in bacula-console <-> bacula-common).

Basically, my setup is the following:

a) pbuilder builds the packages in base-sid
b) dput/mini-dinstall adds the packages to my local APT repository
c) piuparts tests the packages in pbuilder's base-sid-local, which has
   two main differences WRT base-sid:

   1. /etc/apt/sources.list.d/local_luca.list
      deb file:/home/luca/public_html/debian gismo-unstable/
      #deb-src file:/home/luca/public_html/debian gismo-unstable/
   2. /etc/apt/preferences.d/local_luca
      Package: *
      Pin: release a=gismo-unstable
      Pin-Priority: 500

The second difference is a workaround, because I want the local
repository to be equal to unstable, otherwise the strict
same-source-package dependencies are not installed and piuparts'
install-upgrade fails.

BTW, please note that the latter is caused by how I configured
     mini-dinstall, i.e. not simply 'unstable' and with
     'experimental_release = 1': gismo-unstable should be considered
     like experimental (no automatic upgrades) and thus it gets by
     default a Pin-Priority of 1, see `man 5 apt_preferences` and:


> Can you provide a logfile with the "bad" behavior?

See above.

Thx, bye,
Gismo / Luca
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/piuparts-devel/attachments/20120611/cfb54e28/attachment.pgp>

More information about the Piuparts-devel mailing list