Review: Syncing from testing a success?

[Kees Cook]

On Tue, Apr 20, 2010 at 02:02:37PM +0200, Martin Pitt wrote:
> Kees Cook [2010-04-08 13:38 -0700]:
> > >  (4) made library transitions easier or harder
> > 
> > I wasn't directly involved in any, but my gut said this was actually
> > harder, since it created some issues when suddenly everything for
> > a transition appeared in testing, instead of following the same
> > order/progression as they went into unstable.
> 
> How was this a problem in Ubuntu? At first sight this seems to be an
> advantage to me, since when that happens we know that all the reverse
> dependencies finally build with the new library and there are no
> urgent RC bugs left any more.
> 
> Did the new packages not have strict enough versioned build
> dependencies, so that they built too early and failed?

There were delays in the builders, since amd64 was running behind by about
2 days.  When gnome-panel went through a transition I accidentally ended up
with a system that could not run ubuntu-desktop for just under 2 days.  It
was my own fault to allow that upgrade, but it kind of showed to me that
doing a huge number of transitions at once runs the risk of breaking people
if builds get out of sync or delayed.  With a slower transition, I feel
this is minimized.

-- 
Kees Cook
Ubuntu Security Team