Hi all,

I'm filing this bug for transparency, I'm not actually expecting us to
fix it considering the upcoming new solver in apt.

In the gnome-shell transition, we were debugging failures in the
autopkgtest of gnome-shell-pomodoro which shows that during the build
(due to the needs-build restriction) autopkgtest doesn't apply the
fallback of using everything from the source suite that it does while
running the test. This is inconsistent.

Paul

https://ci.debian.net/data/autopkgtest/testing/amd64/g/gnome-shell-pomodoro/52897268/log.gz

  35s autopkgtest [18:07:59]: testing package gnome-shell-pomodoro
version 0.26.0-1
  35s autopkgtest [18:07:59]: build needed for tests
  35s autopkgtest [18:07:59]: ERROR: Failed to resolve build-deps for
/tmp/autopkgtest-lxc.x_ev68qg/downtmp/build.UQA/src; apt-get --simulate
output:
  35s autopkgtest [18:07:59]: ERROR: | Note, using directory
'/tmp/autopkgtest-lxc.x_ev68qg/downtmp/build.UQA/src' to get the build
dependencies
  35s autopkgtest [18:07:59]: ERROR: | Reading package lists...
  35s autopkgtest [18:07:59]: ERROR: | Building dependency tree...
  35s autopkgtest [18:07:59]: ERROR: | Reading state information...
  35s autopkgtest [18:07:59]: ERROR: | Some packages could not be
installed. This may mean that you have
  35s autopkgtest [18:07:59]: ERROR: | requested an impossible situation
or if you are using the unstable
  35s autopkgtest [18:07:59]: ERROR: | distribution that some required
packages have not yet been created
  35s autopkgtest [18:07:59]: ERROR: | or been moved out of Incoming.
  35s autopkgtest [18:07:59]: ERROR: | The following information may
help to resolve the situation:
  35s autopkgtest [18:07:59]: ERROR: |
  35s autopkgtest [18:07:59]: ERROR: | The following packages have unmet
dependencies:
  35s autopkgtest [18:07:59]: ERROR: |  libmutter-14-0 : Depends:
mutter-common-bin (= 46.4-1) but 47.0-3 is to be installed
  35s autopkgtest [18:07:59]: ERROR: | E: Unable to correct problems,
you have held broken packages.
  35s blame: gnome-shell-pomodoro
  35s badpkg: Can't resolve build dependencies on testbed
  35s autopkgtest [18:07:59]: ERROR: erroneous package: Can't resolve
build dependencies on testbed

On Thu, 10 Oct 2024 Paul Gevers <elbrus@debian.org> wrote:> I'm filing
this bug for transparency, I'm not actually expecting us to

Hi, pinning is also inconsistent when doing the initial dist-upgrade of
the testbed, as can be seen in this case (after all the noise due to an
openstack failure and retry):

https://objectstorage.prodstack5.canonical.com/swift/v1/AUTH_0f9aae918d5b4744bf7b827671c86842/autopkgtest-plucky/plucky/amd64/liba/libalgorithm-merge-perl/20241022_010901_b3072@/log.gz

I agree the way forward is using the new solver and --no-string-pinning,
but details are to be worked out.

Also, the fallback logic (should we end up keeping any) should be moved
in some common function, like _run_apt_command().

Paride