postinst calls start(8) instead of invoke-rc.d(8), ignores policy-rc.d(8).

[Trent W. Buck]

policy-rc.d(8) can be used to prevent sysvinit services from booting.
This is particularly important within chroots.

In Debian Sid, I used pbuilder(8) to create a lucid "base.tgz" chroot.
I then used pbuilder --login to get a shell, and ran "apt-get install
apport". (I had heard that apport didn't just fire up a web browser,
and so I was trying to use it to report a bug in lucid's kernel.)

To my surprise, I found the postinst was trying to start a service:

    Setting up apport (1.12-0ubuntu5) ...
    start: Unable to connect to Upstart: Failed to connect to socket /com/ubuntu/upstart: Connection refused

Luckily it failed, because I wasn't running upstart outside the
chroot. But it seems to me that either upstart's start(8) should
either honour policy-rc.d, or at least reinvent it.

I see five things that could happen here:

 - I've misunderstood the error message. Flag MOREINFO.

 - upstart already has a policy-rc.d equivalent, and apport needs to
   be told to honor it in apport's postinst. Assign to apport.

 - upstart already has a policy-rc.d equivalent, and pbuilder needs to
   be told to create the appropriate "deny all" policy when creating
   Ubuntu-based chroots. Assign to pbuilder.

 - upstart has no policy-rc.d equivalent, and needs to grow one.
   Assign to upstart.

 - upstart has no policy-rc.d equivalent, and the upstart developers
   think it doesn't need one. Flag WONTFIX and I get grumpier.

ProblemType: Bug
Architecture: amd64
Date: Thu Feb 18 08:25:52 2010
DistroRelease: Ubuntu 10.04
Package: apport 1.12-0ubuntu5
PackageArchitecture: all
ProcEnviron:
 LC_COLLATE=C
 PATH=(custom, no user)
 LANG=C
 SHELL=/bin/bash
SourcePackage: apport
Uname: Linux 2.6.24-23-openvz x86_64

[Trent W. Buck (twb)]

Acting upon information received, I have filed
http://bugs.debian.org/571054 and http://bugs.debian.org/571056

This malone ticket should probably be linked to those debbugs tickets,
and reassigned to pbuilder.

[Colin Watson (cjwatson)]

Fixed in live-helper 2.0~a9-1, per the Debian bug, and thus in maverick.

[Mattia Rizzolo (mapreri)]

Acting the same as https://bugs.debian.org/571054

Maybe the reasons can be considered different: the problem does not hold anymore, it's quite old and the affected releases are EOL.