On 1 April 2016 at 08:14, Sam Morris <s...@robots.org.uk> wrote: > On Tue, 2016-03-29 at 07:58 -0300, Felipe Sateler wrote: >> >> On 29 Mar 2016 04:03, "Felix Geyer" <fge...@debian.org> wrote: >> > >> > Hi Felipe, >> > >> > On Mon, 28 Mar 2016 20:56:48 -0300 Felipe Sateler wrote: >> > > I have uploaded an nmu. I have made the unit call out to the init >> > > script, because it does more work than simply invoking ferm. >> > > >> > > Please find attached the debdiff >> > >> > I see two problems with your systemd service: >> > >> > 1) By default (CACHE=yes) the init script writes to /var/cache/ferm/ and >> > the systemd service is ordered Before=network-pre.target. >> > If /var is on a remote filesystem you have created a dependency cycle. >> Hmm, correct. Ferm will have to start after the network. Sorry about that. > > It would be much better to drop the use of the caching feature. It's > not even used by upstream's systemd unit. And I really don't want a > firewall service that starts before the network is up.
You should open a new bug about that, as it is only tangentially related to the issue of this bug (missing systemd counterpart for a runlevel S script). -- Saludos, Felipe Sateler