Bug#746578: More systemd fallout :-/ [and 1 more messages]
On Fri, 19 Sep 2014 13:31:51 +0100 Ian Jackson <ijackson@chiark.greenend.org.uk> wrote:
> Josh Triplett writes ("Bug#746578: More systemd fallout :-/"):
> > On Wed, 17 Sep 2014 15:34:48 +0100 Ian Jackson <ijackson@chiark.greenend.org.uk> wrote:
> > > As I understand it from reading the threads in the bug and on
> > > debian-devel, the effect of this would be:
> ...
> > The latter two points are not actually accurate. I just tested [...]
>
> Thanks very much for that information.
>
> Do you agree with Cameron's summary:
>
> Cameron Norman writes ("Bug#746578: More systemd fallout :-/"):
> > Also, although the squeeze/wheezy -> jessie bit Ian wrote seems to be
> > incorrect, his last point still stands: on a jessie minbase (with init
> > shifted to !systemd-sysv), if you install libpam-systemd, your init is
> > changed back to systemd.
> >
> > So the "systemd-sysv | systemd-shim" bit is either pointless and
> > redundant (upgrading to Jessie) or actively disruptive (installation
> > of libpam-systemd on jessie+ systems).
This is the same point I agreed with from Steve's mail: if you install
jessie, *explicitly* switch to sysvinit-core (or upstart for that
matter), and then install libpam-systemd, that shouldn't switch your
init system. If we can fix that scenario without breaking others, we
should.
> That is:
>
> * At the moment existing systems are switched from sysvinit to
> systemd (unless the administrator takes some action to prevent it).
> This is done by other means (involving the init packages), even if
> libpam-systemd is not pulled in. Having the init system switch
> done by a peripheral package like libpam-systemd is not sensible.
>
> Therefore the existing dependency ordering in libpam-systemd is
> unnecessary (and strange) in this situation (supposing that you
> think that people upgrading should be switched by default) or
> actively harmful (if you think that they should not).
>
> * On a jessie system which has been deliberately set up not to use
> systemd, installing a package which pulls in libpam-systemd causes
> the system to switch to systemd.
>
> This is not desirable. It is a consequence of the existing
> dependency ordering in libpam-systemd.
>
> * There are no other relevant consequences of the dependency
> ordering.
>
> From which we can conclude that the dependency ordering should be
> changed.
Assuming that apt does the right thing with the dependencies reversed,
yes. I outlined several specific scenarios in my response to Steve's
mail, which someone ought to test with a modified libpam-systemd
package: new installs of jessie, installs of wheezy upgraded to jessie
(with or without a package installed whose upgraded version depends on
libpam-systemd), d-i (with and without selecting task-desktop), and
debootstrap (with and without simultaneously installing libpam-systemd).
If all of those scenarios work identically with the libpam-systemd
dependency reversed, and reversing that dependency improves the
case Steve raised where the admin has explicitly installed an alternate
init that libpam-systemd should not replace with systemd-sysv, then I
see no harm in reversing the dependency.
- Josh Triplett
Reply to: