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

Re: Upload of GNOME 2.6 to unstable

On Sun, Apr 18, 2004 at 09:15:18AM +1000, Anthony Towns wrote:
> On Sat, Apr 17, 2004 at 11:57:14AM -0500, Steve Langasek wrote:
> > On Fri, Apr 16, 2004 at 12:10:00AM +0200, Jordi Mallach wrote:
> > > As you know, the Debian GNOME team has been working on packaging
> > > GNOME 2.6 during the last weeks. While many of us didn't count on having
> > > them ready to opt for their inclussion in Sarge when we started, the
> > > situation has changed now that packages are judged to be at least
> > > unstable quality.
> At the moment, afaics, packages have only been uploaded to experimental
> for i386 and powerpc. Please make sure they're building on all
> architectures before even considering a major change like this.

  Sorry, but this is not easy... Currently my s/390 machine is busy
  making coffee ;-)

  <joke off>IMO this is quite difficult without experimental having
  autobuilder support.

> Second, the quality level you need to be thinking about for uploading
> to unstable isn't "unstable quality", it's "release quality". Are these
> packages suitable to form a basis for other developers' work? Are they
> reliable for everyday use? Have they been run through any automated
> test suites we have available (at a minimum building them in a
> pbuilder/autobuild environment) and are they reliably passing? If not,
> they're not release quality, so they're not "unstable" quality.

  Perhaps "unstable" should be renamed to "devel" then.

> Our goal is to cope with transitions like this on fairly short notice. So
> that if Gnome X.Y releases in June, we can have "release quality" Debian
> packages in July, and include them in a new Debian stable release that
> comes out in August.
> There are obvious concerns at each point in the above:
> 	* Gnome X.Y might not be stable enough for release in a mainstream
> 	  distribution; we might need to wait for X.Y.1, or X.Y.2. We
> 	  might need more than a month to make this decision.
> 	* We might not be able to prepare release quality packages in a
> 	  month -- porting upstream stuff across all our architectures,
> 	  or adapting it to Debian policy, or providing some reasonable
> 	  testing, could easily take more than a month's work too.
> 	* Integrating the new set of packages into the distribution --
> 	  both satisfying dependencies so the testing scripts are happy,
> 	  and checking for bugs and misfeatures and incompatabilities
> 	  and fixing them so that _users_ are happy -- can take more than
> 	  a month too.
> The best way of meeting the first two concerns is to prepare and maintain
> the packages in experimental; that means finding people to build it
> on various architectures, and pestering them when you make uploads,
> and getting people to install the stuff you upload to experimental and
> report problems.
> You need to do some more of that before thinking about uploading to
> unstable yet.

  But having packages in experimental means things are not so widely
  tested, and as I said before there are also problems with having the
  packages available in all arches. So somehow this is a chicken - egg
  problem. We want to make such changes painless and quick, but at the
  same time we need to use experimental release, which is much more 
  marginal than unstable.

  Leaving apart that I think that for Sarge+1 we should somehow change
  the way we relese Debian (I'd like to have a comprehensive *discussion*,
  not flamewar,  about this when it's the time), I think that we should
  start to change something about the way we think on releases. Why
  couldn't we move GNOME 2.6 to unstable and then use
  sarge-proposed-updates for fixing GNOME 2.4 problems that could appear
  until Sarge is released? This will give us a lot of advantages, and as
  I see it, not a lot of problems with Sarge release.

  Believe me that I understand your way of think, but also try to see
  our point. We will have GNOME 2.8 in October. Not being even able to
  put 2.6 in sarge means that a lot of work made to have 2.6 in a good
  shape will be throw away when 2.8 comes out, and later when 3.0 is
  ready and even perhaps when 3.2 is born. And this also will force a
  lot of people to move to unstable (the name is really scary) or to
  move to other distro. Well, I'm talking again about release process
  so I'll shut up right now. But I hope you see what I try to express.


Jose Carlos Garcia Sogo

Attachment: signature.asc
Description: Digital signature

Reply to: