On Thu, Aug 12, 2010 at 10:38 AM, Michael Schwendt <mschwendt(a)gmail.com> wrote:
I've filed lots of -1 in bodhi for EPEL 5, and that doesn't
cover all issues.
Apparently, there are packagers, who mark their updates "stable" without
even having tried to install the packages while they were being offered
in epel-testing. This affects new packages as well as updates. It gives
the smell of treating EPEL as just another build target and repo to dump
builds at. In one case, the packager has admitted he doesn't have an
installation to test with. At least using a compatible CentOS
installation on a multi-boot machine ought to be mandatory.
The RubyGems stack:
There are several rubygem* (and ruby*) packages, which suffer from
unresolvable dependencies. Dunno how complete repoclosure is on EL 5
(e.g. with regard to Obsoletes), but it's available:
$ sudo yum -y install yum-utils
$ repoclosure | tee el5-broken-deps.log
At least "libevent" has had a SONAME bump in the base dist.
Corresponding rebuilds of packages in EPEL 5 are missing.
epel-devel-list mailing list
Michael, thanks for the filings. We're aware of many of the rubygem
issues, and have bugs opened to fix them.
I was playing with repoclosure a couple nights ago on EPEL and found
quite a few dep issues, and was concerned that I was doing something
wrong with it. Perhaps I wasn't.
EPEL Team -- Should we start planning another bug day? It certainly
is a goal to at least have the stable repo not have dep issues.