On Sun, 2005-04-03 at 13:17 +0100, Alan J. Flavell wrote: > On Sat, 2 Apr 2005, John Franks wrote: > > > Here is what I am > > doing -- it's not pretty, but it seems to work. > > > > rpm -q librsvg2 (check you really have two versions installed) > > > > rpm -e librsvg2-2.2.3-2 > > (remove the old one -- this generates errors, but works) > > Confirmed for 3.0.3 > > At this point, $ rpm -V librsvg2 > missing /usr/share/doc/librsvg2-2.2.3 > missing d /usr/share/doc/librsvg2-2.2.3/AUTHORS > missing d /usr/share/doc/librsvg2-2.2.3/COPYING > missing d /usr/share/doc/librsvg2-2.2.3/COPYING.LIB > missing d /usr/share/doc/librsvg2-2.2.3/ChangeLog > missing d /usr/share/doc/librsvg2-2.2.3/NEWS > missing d /usr/share/doc/librsvg2-2.2.3/README > > > rpm -Uvh --force librsvg2-2.2.3-6 > > That would be something like > > rpm -Uvh --force librsvg2-2.2.3-6.i386.rpm Yes, that's correct. > > (Actually we removed some docs we want to keep so re-install > > the new version. This generates errors too, but seems ok.) > > That all seems to fit, in 3.0.3 also. > > > rpm -V librsvg2 (looks clean) > > Confirmed > > However, I spotted a rather curious item left around after the > manual procedure (which was performed at 12:48 local time): > > -rw-r--r-- 1 root root 0 Apr 3 12:48 > /etc/gtk-2.0/gdk-pixbuf.loaders > > whereas on another machine that had run the defective update but > hasn't had the manual procedure applied, this file was a duplicate of > > /etc/gtk-2.0/i686-redhat-linux-gnu/gdk-pixbuf.loaders > My guess is that this is again a consequence of the failing install script for librsvg2-2.2.3-6.i386.rpm. I suspect that script which runs at the end of installation is supposed to do cleanup and remove the older version. It is failing because the pixbuf.loaders script has been moved to a subdirectory named i686-redhat-linux-gnu instead of i386-redhat-linux-gnu. There is a good chance that changing the spec file (or the script it refers to?) will fix the problem.