Hi, On Sat, Jun 16, 2012 at 02:24:56PM +0200, Andreas Beckmann wrote: > the problem now also happens on a regular update from testing.
[...] > Installing dialog does not help. > You should ask a debconf guru as that seems to be a debconf > related problem, /usr/share/debconf/frontend seems to be failing. This is not a debconf problem. Here is what happens: * first install The postinst creates a zope instance: if dzhandle -z $ZVER make-instance -u "$user:$password" $INSTANCE \ --addon-mode=all --restart=end -t tree-linked \ --service-port=$port --service-user=zope:zope $ZEOSERVER This creates a directory /var/lib/zope2.12/instance/sandbox/ with a number of files. * remove package The directory /var/lib/zope2.12/instance/sandbox/ is still there (with files). * second install /var/lib/dpkg/info/zope2.12-sandbox.config runs. It checks if there is an instance without datafs: if [ -d /var/lib/zope$ZVER/instance/$INSTANCE ] && \ [ "$DATABASE" != "ZEO" ] && \ [ ! -f /var/lib/zope$ZVER/instance/$INSTANCE/var/Data.fs ]; then FIRST_STATE=1 else FIRST_STATE=2 fi This is the case, so FIRST_STATE=1 In this case, the script checks what to do, by asking the user: db_get $PKG/remove-instance-without-data This template defaults to abort: Template: zope-common/remove-instance-without-data Type: select Choices: abort, remove and continue Default: abort So this script aborts (by exiting with return code 10). In other words: the package creates an (empty) zope instance in the postinst, but fails to install if such an instance exists. This makes a reinstall fail. It seems the relevant parts of these scripts come from zope-debhelper, so this bug probably originates there. Cheers, Ivo -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org