Experienced package builders: please test my first RPM

[Paul Howarth]

On Tue, 2005-09-06 at 14:01 -0400, Robert P. J. Day wrote:
> On Tue, 6 Sep 2005, Tony Nelson wrote:
> 
> > At 7:53 AM +0100 9/6/05, Paul Howarth wrote:
> > >On Mon, 2005-09-05 at 14:14 -0400, Tony Nelson wrote:
> > >> I've written a tiny utility and packaged it in my first RPM.  I'd
> > >> appreciate a couple of experienced packagers taking a look at it and
> > >> telling me about any problems. ...
> >  ...
> > >> <http://georgeanelson.com/apropos2.htm>
> > >
> > >Summary conventionally should start with a capital letter and not end
> > >with a period.
> >  ...[other comments]...
> >
> > >Package seems to work OK :-)
> > >
> > >I've attached a spec file incorporating these suggestions.
> >
> > Wow.  Thanks, Paul!  I'll get to work on it.  (I'll do it per your
> > suggestions, and then compare with what you sent, to learn better.)
> 
> i've just recently run across a couple tarballs that incorporate the
> python "distutils" utility for building their own RPM.
> 
> http://www.python.org/sigs/distutils-sig/
> 
> all i know about this is that it seems to work.  any comments on its
> packaging efficacy?

This generates workable RPM packages, which are certainly better than
installing direct from tarballs (upgrading and uninstalling being two of
RPM's big advantages). Care needs to be taken to ensure that proper
dependency information is included in the resulting package, but for
many python modules and small projects, distutils provides a good
solution. For more complex projects, e.g. where
post-install/post-uninstall scripts need to be run (e.g. to update the
MIME database or the GTK icon cache), building fully-functional RPMs
using distutils may not be possible. I'd say there's no substitute for a
carefully hand-crafted RPM spec file in general.

Paul.
-- 
Paul Howarth <paul at city-fan.org>