Etienne Dysli-Metref <etienne.dysli-met...@switch.ch> writes: > On 19/01/17 23:46, Ferenc Wágner wrote: > >> I couldn't reproduce this on a real jessie system, only in a piuparts >> chroot. I think the reason is that piuparts removes init-system-helpers >> (the home of deb-systemd-helper) before the shibboleth-sp2-utils postrm >> could instruct it to clean up. I'm not sure what we could do about >> this. > > Indeed piuparts does remove init-system-helpers before > shibboleth-sp2-utils. I hadn't noticed before but it's in the log: > [...] > Why would puiparts do it in this order? shibboleth-sp2-utils depends on > init-system-helpers!
https://www.debian.org/doc/debian-policy/ch-maintainerscripts.html says: The postrm script is called after the package's files have been removed or replaced. The package whose postrm is being called may have previously been deconfigured and only be "Unpacked", at which point subsequent package changes do not consider its dependencies. Therefore, all postrm actions may only rely on essential packages and must gracefully skip any actions that require the package's dependencies if those dependencies are unavailable. This is exactly what happens. Shibboleth-sp2-utils is removed, then init-system-helpers is removed, then shibboleth-sp2-utils is purged, but it can't use init-system-helpers to fully clean up after itself. >>> So I bumped the build-dep up a bit to: dh-systemd (>= 9.20160709). >> >> Why? I mean, where did this version number come from? > > The version comes from debhelper's changelog in jessie-backports . > It's when dh-systemd was moved to debhelper. Got it, thanks. > Since adding this version constraint was motivated by piuparts's report, > it may not be necessary in the end... But we'd still need the functionality of dh-systemd in our backport. I'll look through #822670 and #837585 for hints. -- Regards, Feri