On Wed, Sep 21, 2016 at 02:55:38AM +0200, Matthias Klumpp wrote:
> 2016-09-21 0:48 GMT+02:00 Josh Triplett <j...@joshtriplett.org>:
> > While it makes sense that gnome-packagekit should depend on packagekit
> > if it can't function without packagekit installed, that then suggests
> > that gnome should recommend gnome-packagekit rather than depending on
> > it.
> > Having packagekit installed has (or had, it might not anymore) some
> > system-wide side-effects,
> Concrete examples please, none of these "effects" should happen unless
> triggered by something else in the system.
> > such as letting gnome-packagekit invoke the
> > systemd offline-upgrade functionality
> gnome-packagekit never does that. GNOME Software however, can do that,
> but that's hardly the fault of PackageKit.
> If anything, this should be solved in GNOME Software, which is the
> sole thing triggering offline-upgrades at time.
Good to know, thank you. I didn't realize that behavior was limited to
GNOME Software. That completely addresses my concern.
> > which isn't well-established or
> > tested in Debian.
> I tested that on my machine and in a couple of VMs and couldn't spot
> any issues. The integration with Plymouth could be better, but aside
> from that, I saw no problems.
I don't just mean testing the mechanism itself, but testing its use for
many different variations on system upgrades and software installations.
It isn't the common path, so it doesn't get nearly as much testing.
Changing that would require a significant fraction of unstable and
testing users to actually use the offline update mechanism as their
primary updater for an extended period of time.
- Josh Triplett