One option smart needs to conquer the world

[Axel Thimm]

On Thu, Feb 23, 2006 at 02:07:05PM -0300, Mauricio Teixeira wrote:
> Axel Thimm wrote:
> 
> > It could just be the opposite, too. The downgrading of package foo in
> > smart happens not because it's a rainy sunday, but because you asked
> > smart to perform an operation like perhaps upgrading another package,
> > bar, that *does* has a security issue.
> 
> I'm just reading your comments because I think I have nothing else
> interesting to add to that discussion better than you all have been
> saying. Very productive talk indeed.
> 
> I decided to break up my (less important) silence just to say that (if
> that matters anyone) I would use the above quoted argument as one of the
> reasons why Smart works out better than others. I completely agree with
> Gustavo when he says that Smart should do what the user *asks to* even
> if it needs to downgrade something, and Axel's comment is the perfect
> complement for that argument.
> 
> > Another depsolver would say: No, I won't upgrade bar to version 2
> > because foo requires bar = 1. So as long as the repo is broken that
> 
> It's very possible to have that erroneous behaviour when you have many
> different 3rd-party repositories that handles the same package (with
> different versions). Some day someone will screw up the "Requires" or
> "Provides" tags and you'll get a broken dependency solving. In most
> cases enforcing priorities on channels and/or packages would solve that
> problem.

The sad thing is that this is not only very possible, but it *already
happens* on a regular basis on *vendor provided* channels. No need to
add more variables like 3rd party repos to see the effect.

And it happens more often when there is a new distribution released
which is followed by the flurry of updates as seen on Fedora Core. And
the original poster with the request to disable the downgrading comes
from Fedora Core, so in fact Fedora Core does have a very large need
for this feature, and not for disabling it. If you now think that this
is paradox, you're right.

For some examples see two of them which I also posted to a response to
Seth Vidal:

https://www.redhat.com/archives/fedora-list/2005-August/msg03536.html
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=145415

-- 
Axel.Thimm at ATrpms.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: not available
URL: <http://lists.labix.org/pipermail/smart-labix.org/attachments/20060223/e27373ac/attachment-0003.pgp>