yum error

Patrick fedora at puzzled.xs4all.nl
Tue May 17 12:06:03 UTC 2005


On Tue, 2005-05-17 at 06:26 -0400, Ivan Gyurdiev wrote: 
> I disagree 

I respectfully disagree with you.

> - the point of test releases is not development

With the exception of this one. That is, assuming you will install a
test release and test just that. Meaning you will not be installing
updates from rawhide. Once you are installing updates from rawhide you
are essentially running rawhide which is a fast moving, unstable
*development* tree with absolutely *no* guarantuee that it will install,
work or keep working. Stick with test releases if you want to stay clear
from possible dependency issues resulting from rawhide updates.

>  - it's
> stabilization. The point of rawhide may be development, but broken
> dependencies are unnecessary, and scare away the beta testers.

Broken dependencies can only exist in rawhide. If you stick with testing
just the official test release and stay away from rawhide then broken
depencies will not bite you. Find bugs in the test release, report them
if they have not been reported already, wait for the next test release,
install that one, rinse and repeat. This is a most welcome contribution
too.

The entire dependency tree of FC4Tx is quite complex making broken
dependencies a possibility. It's quite challenging to keep rawhide fully
installable 100% of the time when development is going at a furious
pace. I have never seen anyone complain about it until this thread
started. And how difficult is it to use yum update --exclude=$package?
It's rawhide, it's continuously changing, it can be highly unstable at
times, it can eat your pets for breakfast.

Frankly rawhide does not scare me away at all. I have been using it for
a long time without going into shock and I think there are many, many
others that have a similar experience.

> A package which introduces dependency damage will not be tested,
> until the dependency damage is fixed - so why is it even released
> until that point?

Because it is ready, because the developer wants to push it out and get
on to the next thing on his/her todo list so deadlines are met, items
can be signed off, PHB's are kept happy and Bill Gates continues having
nightmares. Or perhaps because it is a dependecy of something that was
pushed to rawhide earlier. The catch22 of rawhide moving forward.

Obviously a package with missing dependencies can not be tested because
it can not be installed *yet*. My experience is that in general it takes
only a few hours to a day or so for any missing dependencies to be
fixed. Works for me and afaict it works quite well for FC.

Regards,
Patrick




More information about the fedora-test-list mailing list