On May 7, 2013 7:32 AM, "Christoph Berg" <cb@df7cb.de> wrote:
>
> Re: Martin Pitt 2013-05-06 <20130507044020.GD2875@piware.de>
> > > Martin: What do you think about generating the logrotate config at
> > > install (configuration) time so we can drop the fixed logrotate
> > > depends/breaks? Using "ucf" for generated files in /etc/ is pretty
> > > easy to do, the open question would be how to detect that logrotate
> > > was just upgraded and that we need to re-generate the file. Maybe with
> > > a trigger on /usr/sbin/logrotate? Or is that too much magic? (Are we
> > > aware of other packages that have this problem?)
> >
> > It's too much magic for my taste. With conffiles there's so much that
> > can go wrong, and one often runs into situations where they cannot be
> > updated automatically.
>
> I'm less worried about that part, ucf is really good at that. It would
> be using a similar sed/awk rule like we have now, and the result of
> that is the new template that is presented to the admin. (It even
> supports 3-way merges.)
>
> > But in the end I don't think that this is such a big problem; the
> > non-LTS releases are really not a good thing to deploy PostgreSQL on,
> > especially now that from 13.04 on they are only supported for 9
> > months. So perhaps the right answer for now is just "don't do that
> > then"?
>
> People still like to do development on their laptops...
>
Yeah, that's an important consideration. And it's important to be able to run the same version as you're on in production, which might very well not be whatever happens to be in the latest desktop distribution - in most cases it's probably something older, given the refresh rate of desktop Ubuntu.
So if it's possible to get a proper fix for it without huge amounts of work that would be good. And I assume it's not wasted effort, since the next lts will build on the same thing and eventually require the same fix?
/Magnus