On Mon, Sep 13, 2010 at 22:10:51 +0200, Ola Lundqvist wrote: > > diff -u vzctl-3.0.24/debian/vzctl.postinst > > vzctl-3.0.24/debian/vzctl.postinst > > --- vzctl-3.0.24/debian/vzctl.postinst > > +++ vzctl-3.0.24/debian/vzctl.postinst > > @@ -52,6 +41,25 @@ > > rm $NAMECFG > > fi > > done > > + fi > > + # This is not an upgrade, check if we should start it as well. > > + if [ -d /proc/vz ] ; then > > + if [ -x /usr/sbin/invoke-rc.d ] ; then > > + invoke-rc.d vz start || true > > + else > > + /etc/init.d/vz start || true > > + fi > > + else > > + echo "The kernel do not support openvz, do not start vz." > > + fi > > + if [ -e /sys/module/vzevent/parameters/reboot_event ] ; then > > + if [ -x /usr/sbin/invoke-rc.d ] ; then > > + invoke-rc.d vzeventd start || true > > + else > > + /etc/init.d/vzeventd start || true > > + fi > > + else > > + echo "The kernel do not support vzevent, do not start vzeventd." > > fi > > ;; > > > > > > Why are you ignoring errors from the init script? > > Good question. It is a practice of mine to make sure that upgrade > do not fail just because of some issue during the startup. Especially > important as this one checks for kernel modules which may be provided > by custom kernels. Do you have a problem with that? > Well, kind of, yeah. If a service can't start for whatever reason then I don't think it should be ignored and swept under the carpet.
> > It looks like you've done the autotools run with older versions > > (automake 1.10 instead of 1.11)? That makes the diff quite a bit larger > > than it would otherwise be. > > Yes. I run stable here, except for some few backported tools. So, yes > it is larger than necessary. > > > Why does src/vzctl-actions.c now ignore malloc() failure? > > It was part of 'Revert "Check for memory granted by get_dist_name"'. > See > http://git.openvz.org/?p=vzctl;a=commit;h=9588ecbb8c8e3b628a1187e88d664662b83c97d5 > There were issues with the change in 3.0.24. > > > The upstart fix doesn't seem necessary, although probably harmless.. > > I'm not sure what upstart fix you are referring to. > > If you are referring to the change in vzeventd, then yes it should be > harmless. > Not strictly necessary but could be useful > for people. I thought it would be useful to add as this was the first > time this script was introduced anyway. > > If you are referring to the solution for #411902, I solved that one as > vzeventd should be started on install. I made sure vz was also > started at the same time. > > Or did you refer to something else? > I'm referring to the first and second hunks in etc/dists/scripts/debian-add_ip.sh. Cheers, Julien
signature.asc
Description: Digital signature