On Tue, 11 Apr 2017 at 17:25:00 +0000, Niels Thykier wrote:
> On Thu, 2 Mar 2017 17:59:02 +0100 Louis Bouchard
> <louis.bouch...@canonical.com> wrote:
> > It now tests correctly and I am preparing an upload to our development
> > release.
>
> Any news on this? I cannot see an upload to unstable yet, did something
> hold you up?
The "us" referred to in "our development release" appears to have been
Ubuntu, not Debian.
However, the systemd unit proposed by Scott Leggett is not actually the
same as the one now shipped in Ubuntu zesty. Is this deliberate? Here
is a diff:
--- ubuntu/debian/unattended-upgrades.service
+++ scott/debian/unattended-upgrades.service
@@ -1,7 +1,6 @@
[Unit]
Description=Unattended Upgrades Shutdown
-DefaultDependencies=no
-Before=shutdown.target reboot.target halt.target network.target local-fs.target
+After=network.target
Documentation=man:unattended-upgrade(8)
[Service]
@@ -11,4 +10,4 @@
ExecStop=/usr/share/unattended-upgrades/unattended-upgrade-shutdown
TimeoutStopSec=900
[Install]
-WantedBy=shutdown.target
+WantedBy=multi-user.target
I'm concerned that the version now shipped in Ubuntu might in fact be shut
down *after* the network is taken down, because Before/After dependencies
are about the order of startup: shutdown happens in the reverse order.
So if unattended-upgrades.service has After=network.target, as in Scott's
proposed unit, it will be started (which now does nothing) after
network.target is started, but stopped (which is where the real work
happens) before network.target is stopped. That seems like the right
thing, and the version in Ubuntu zesty seems like only a partial fix.
S
