[Yum] yum's update behavior on broken repositories

[Eric S. Raymond]

Robert G. Brown <rgb at phy.duke.edu>:
> What do your /etc/yum.repos.d/*.repo's look like?

They point at the core FC3 repositories and livna.org.  The Dag Wieers
repository is described but disabled.

> In the past the reason for 1 at least that has been given that yum is
> designed to have "no unexpected side effects".  Either it works or it
> doesn't work atomically, and if it doesn't work you are responsible for
> figuring out how the fact that it doesn't work alters your plans.

I understand that philosophy.  It's appropriate when you're doing an
install of one specified set of packages, but much less so when you're
doing a general update and don't want to have over 200 package updates
blocked because one unrelated component that you're not interested in
is missing.  That's what's happening to me.

And the philosophy doesn't explain why the exclude option is useless.
I tried to use it to tell yum to not install any of the dozen or so packages
dependant on the missing xfce part.  It completely ignored that instruction
and treated the xfce dependencies as showstoppers. aborting the *entire*
update.

I'd like to fix at least thec exclude bug, but I can't because I can't figure
out how to test the CVS version.

esr at snark:/etc/yum.repos.d$ cd ~/cvs
esr at snark:~/cvs$ cd yum
/home/esr/cvs/yum
esr at snark:~/cvs/yum$ bin/yum.py
Traceback (most recent call last):
  File "bin/yum.py", line 6, in ?
    import yummain
  File "/usr/share/yum-cli/yummain.py", line 23, in ?
    import yum
  File "/home/esr/cvs/yum/bin/yum.py", line 7, in ?
    yummain.main(sys.argv[1:])
AttributeError: 'module' object has no attribute 'main'

Would someone please document a test procedure for this puppy?
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>