[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#109577: version 1.23 available



On Tue, Aug 28, 2001 at 04:41:23PM +0100, Philip Blundell wrote:
> >2. if 1.23-x is in sid, and a fix for 1.22 is required for woody, then
> >it cannot be done correctly, as the new package should go through sid,
> >and this wouldn't be possible.  Even if we bump the epoch to get
> >1:1.22-5 >> 1.23-1, then the people running sid would get a problem
> >because their 1.23 would get downgraded
> >
> >This looks like a major problem with the current implementation of
> >testing - one more that would advocate direct uploads into testing, or
> >at least through a testing-fixes.
> 
> Well, no, it calls for direct uploads into frozen.  The problem only exists 
> once woody freezes and it's no longer realistic to get 1.23-x into that 
> distribution.  Most of the time there's no real issue.

There is the issue when major packaging changes need to be done, and
the maintainer is not sure to have enough time to finalize with a
release-ready quality.  In that, the current process either encourages
people to delay their work (with the risk that they don't care any
more when the time has come), or to start things without being sure to
be able to finish, thus leaving more work at pre-release time.

Even without the "not sure to have enough time" case, when a redesign
is in progress, or a package is under heavy development (who told the
name of any of joeyh's package ;), bugfixes never get it into testing
unless the maintainer pauses in his release rates, and/or uses
"urgency=high" to get things done - which is nothing but a workaround
of an inadequate process.  IIRC joeyh had to do that at least once,
for debconf - I was the one who forgot a ">= 0.5" on a debconf
dependency, erroneously allowing console stuff to enter testing, and
there was nothing to do to fix testing.  And this was far from the
freeze.

Testing in its current state has potential to be more unstable than
sid, has already been, and will be again if the process does not
evolve.  My idea to work around it is expressed above, but I'm quite
sure I overlooked some things - as the problems I describe were
overlooked originally.

Regards,
-- 
Yann Dirson <Yann.Dirson@fr.alcove.com>                 http://www.alcove.com/
Free-Software Engineer				      Ingénieur Logiciel-Libre
Free-Software time manager    	       Responsable du temps Informatique-Libre



Reply to: