kdepim-libs-6:4.1.4-1.fc10.x86_64 has error in F10 x86_64 updates-testing

[stan]

José Matos wrote:
> On Monday 09 February 2009 22:12:24 Adam Williamson wrote:
>> Dumb question alert, but - wouldn't the correct way to test be simply to
>> run 'yum update --enablerepo=updates-testing' ? That should attempt to
>> update all installed packages from the updates-testing repo, which is
>> roughly what the 'official update' process will do once all these
>> packages are in the official update repo, yes? That seems to be the most
>> important thing to test.
> 
> Sometimes users want to test just a small set of packages. We have seen lots 
> of examples like this with the recent kde-4.2 update not only on the lists but 
> also on real life (I was asked this same question several times).
> 
> In most cases if what you want to test is just a given component this should 
> not matter. I know that there are exceptions FWIW.
> 
>> --
>> adamw
> 
Adam is right.  If you just do the full update no dependency issues will occur because they will be masked.  That 
doesn't mean they aren't there.  Testing means *looking* for errors, not finding a way to make things work.  If we 
require group updates to take care of dependencies, the individual package dependency system is broken.  Maybe there 
should be whole subsystems that are only group installed with no individual package dependencies because they behave as 
units.  KDE, Gnome, X, etc.

Jose is also right that people *do* want to work with individual packages sometimes.  Maybe that has to be disallowed in 
order to simplify the system.

Finally, this is a very hard problem, the problem of dependencies.  It is comparable to looking for Hamiltonian 
subcycles in a graph with bidirectional edges.  As near as I can guess from watching the output, yum does an exhaustive 
search, paring branches that get to the leaf nodes and cutting cycles as they occur.  But it depends on all the edges 
being in place so the search can find all the nodes.

I suppose my complaining is becoming tiresome because there *does* exist a workaround, so I'll stop now.  :-)