Review: Syncing from testing a success?

[Lucas Nussbaum]

On 20/04/10 at 15:15 -0400, Scott Kitterman wrote:
> "Kees Cook" <kees at ubuntu.com> wrote:
> >On Tue, Apr 20, 2010 at 05:14:45PM +0200, Martin Pitt wrote:
> >> The problem in the past was that there were quite a lot of
> >> transitions which were only half-done because DIF hit in the middle
> >> of them. It seems to me that an immediate transition of all
> >> affected packages handled this much better?
> >
> >This is probably true.  I don't have a strong opinion about it, as it
> >was only a single glitch I hit.
> >
> I think they are both true. 
> 
> Syncing from Testing we get only completed transitions, but the risks
> of archive skew are greater. 
> 
> We've still got some packages that are depwait because a sufficient
> version of a build depend didn't make it (and I was involved in a
> couple of forced downgrades already), so this isn't a panacea. 

It seems to me that what you want, in the end, is something similar to
the britney script that makes the transitions from debian unstable to
debian testing. You probably need to add more intelligence to the
process, so you don't sync packages that break other packages, or that
will fail to build/to install. You also need to have a way to sync
several packages together, to avoid deadlocks. And you need people
monitoring the process to see why the remaining packages don't get
synced.

I actually wonder if britney or britney2 couldn't directly be used for
your needs.
-- 
| Lucas Nussbaum
| lucas at lucas-nussbaum.net   http://www.lucas-nussbaum.net/ |
| jabber: lucas at nussbaum.fr             GPG: 1024D/023B3F4F |