New fedora-rpmdevtools: cleanup old stuff

[Toshio Kuratomi]

On Sun, Mar 20, 2005 at 09:05:26PM +0200, Ville Skyttä wrote:
> 
> How about if I just remove the "full featured" spec template from
> rpmdevtools, and we provide one with a better coverage, docs and
> annotations somewhere else, eg. Wiki?
> 
> The template is still hopelessly incomplete for various common packaging
> tasks.  Updating it to cover everything doesn't make sense, it'd be too
> bloated to be really useful as a spec template for anything, really.
> 
I like this idea.  I was looking through my scriptlets sections and
contemplating how to make spectemplate scale to include scrollkeeper
information.  A wiki page with sample scriptlets would be much better.

> [...]
> > It also adds killall -HUP gconfd-2 which is known to work in FC3+
> > versions of GConf2.  I haven't evaluated whether it's harmful for pre-
> > FC3 gconf (I don't have pre-FC2 anymore.)
> 
> That killall as it was in the patch will cause upgrade failures and
> duplicate leftover packages if there's no gconfd-2 running.  Quick
> "fix": killall -HUP gconfd-2 || :
> 
This is longer but more "fixed":
  GCONF_PIDS=`pidof gconfd-2`
  if [ x"$GCONF_PIDS" != "x" ]; then 
    kill -HUP $GCONF_PIDS
  fi

Is the cleanliness worth the verbosity?

> > This patch could be improved if someone can figure out how to deal with
> > previous packages built with the broken %preun scriptlets.  (As it is,
> > the problem expressed in the cited email will continue when upgrading
> > the set of currently broken packages.)
> 
> %triggerpostun could help.  See /usr/share/doc/rpm-*/triggers
> 
Seat of my pants, untested, just before bedtime try::
  %triggerpostun -- PACKAGENAME
  if [ $1 -gt 0 ]; then
    gconftool-2 --makefile-install-rule \
      %{_sysconfigdir}/gconf/schemas/SCHEMAFILE
    kill -HUP `pidof gconfd-2` || :
  fi

I've always heard that triggers were dangerous and to be avoided if
possible.  This would seem to be the only way to work around packages that
had previously broken GConf uninstall scriptlets.  Does that justify its use
to fix known broken packages?  

-Toshio