syncing from testing? [was: Quantal open for development]

[Colin Watson]

On Mon, Apr 30, 2012 at 05:32:14PM -0400, Scott Kitterman wrote:
> On Monday, April 30, 2012 04:07:08 PM Iain Lane wrote:
> > Seems this sentence was a mistake (at least for now). Quoting #-release:
> > 
> > 30/04 14:24:00 <Laney> "syncs from unstable": it changed then?
> > 30/04 14:25:59 <cjwatson> Laney: Hm, communication glitch
> > 30/04 14:26:05 <cjwatson> I was intending to run syncs from testing
> > until UDS
> 
> When you all that will be at UDS discuss this, would you please decide once 
> and for all that Ubuntu will sync from Testing for LTS and Unstable for non-
> LTS by default (there may, someday be a reason to diverge from this for a 
> cycle) so that we don't have to keep revisiting the "which release do we sync 
> from" question.  I think the answer is generally clear and it's a waste of 
> energy to keep asking the same question again and again when we ~always get 
> the same answer.

Sorry I'm late in following up on this.

The general consensus at UDS was indeed largely as you say: we should
sync from unstable for non-LTS by default, and we agreed to switch to
unstable for this cycle.  I will try to remember to not have this debate
in future non-LTS cycles. :-)

Actually implementing this for quantal was blocked on some toolchain
work on ARM which we wanted to take the opportunity to complete before
doing another pile of syncs, but Adam told me this morning that this was
complete.  I've dug myself into a little bit of a hole by telling
Launchpad that quantal's parent series is wheezy, but I'm in the process
of hacking my way around that in the auto-sync script.  I expect to do
an initial auto-sync from unstable later today.

There remains some question about what to do with future LTS cycles.
Depending on the progress of work on running our own equivalent of
Debian's testing migration (which I seriously hope will be complete by
T, given that it's currently scheduled for quantal!), it's quite
possible that we'll reach the point where syncing from testing merely
introduces an extra source of delay and risk, rather than being a useful
protection.  However, I don't think we can answer that question until we
have some practical experience with running our own migration scripts.

-- 
Colin Watson                                       [cjwatson at ubuntu.com]