Bug#387699: exim4: Should create /var/run/exim4/ if it is missing

Andreas Metzler ametzler at downhill.at.eu.org
Sat Sep 16 11:22:44 UTC 2006


On 2006-09-16 Petter Reinholdtsen <pere at hungry.com> wrote:
> [Andreas Metzler]
[...]
> > You should somehow try to save this info, instead of changing every
> > package using its own directory in /var/run.

> I fail to see the advantage of adding that complexity when we can just
> add 'mkdir' calls to the scripts expecting some directories in
> /var/run/.

Hello,
There are two *possible* obvious advantages:
 - changes only to only one package.
 - Works for stuff that is not started by /etc/init.d/. - I do not
   know whether this is relevant currently. 

Another point is that the whole thing is not always _that_
simple to do in packages, as there are legitimate uses of *users* adding
such directories. e.g. I am running spamassassin under its own
designated user listening on /var/run/am-spamd/socket. - The packages
make this simple:
1 add the user
2 mkdir /var/run/am-spamd/
3 chown am-spamd:am-spamd /var/run/am-spamd
4 add
  --username=am-spamd --socketpath /var/run/am-spamd/socket"
  to /etc/default/spamassassin.

If directories under /var/run/ are no persistent, this breaks. (You
might tell me that I should have used /var/local/run. ;-)

There is also the partial-upgrades issue, upgrading to a version of
sysvinit that drops everything under /var/run/ on every reboot will
break every package currently relying on /var/run/* being persistent.
(adding a bunch of versioned conflicts to the new sysvinit versions
would fix this.)
thanks for listening, cu andreas
-- 
The 'Galactic Cleaning' policy undertaken by Emperor Zhark is a personal
vision of the emperor's, and its inclusion in this work does not constitute
tacit approval by the author or the publisher for any such projects,
howsoever undertaken.                                (c) Jasper Ffforde




More information about the Pkg-exim4-maintainers mailing list