On Wed, 2007-07-18 at 16:01 +0200, Thorsten Leemhuis wrote:
> On 18.07.2007 15:37, Jeremy Katz wrote:
> > On Wed, 2007-07-18 at 14:27 +0300, Manuel Wolfshant wrote:
> >> Peter Vrabec wrote:
> >>>> I'd say that since sysklogd is erased, rsylogd should be automatically
> >>>> enabled and started. Yes, I know the packaging recommendations, but I
> >>>> think that in this case an exception should be accepted.
> >>> I agree, but I don't know any nice and secure way we can do it.
> >> shipping an /etc/init.d/rsyslog with default options set to enabled +
> >> service rsyslog start in %post is not OK ?
> > Starting a service in %post is not in at all okay. It can cause all
> > sorts of havoc when installing systems, creating live images, init'ing
> > chroots for building in, ...
>
> But shouldn't it be safe to start if the daemon (or the predecessor in
> this case) was running already? That's what other scripts do iirc.
Daemons do a condrestart in their %post, not a start.
I guess you could conceivably do something kind of painful like
if service sysklogd status == running
service sysklogd stop
service rsyslog start
But I'm not 100% certain that the ordering of package install/remove on
an obsolete makes this even feasible
Jeremy
--
fedora-devel-list mailing list
fedora-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-devel-list
