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

Re: Cyclic dependencies in octave2.1 packages?



On Sat, Dec 04, 2004 at 09:37:32PM +0100, Richard B. Kreckel wrote:
> On Sat, 4 Dec 2004, Dirk Eddelbuettel wrote:
> > On Sat, Dec 04, 2004 at 11:03:34AM +0100, Andreas Barth wrote:
> > > * Dirk Eddelbuettel (edd@debian.org) [041203 22:45]:
> > > > On Fri, Dec 03, 2004 at 11:19:18AM +0100, Andreas Barth wrote:
> > > > > * Dirk Eddelbuettel (edd@debian.org) [041203 06:15]:
> > > > > > I think Richard is basically correct in his analysis. Bjorn's page lists
> > > > > >
> > > > > >   octave2.1
> > > > > >   octave-forge
> > > > > >   octave-sp     [ source package semidef-oct ]
> > > > > >
> > > > > > as mutually blocking themselves on Alpha -- but buildd.debian.org shows that
> > > > > > all packages have built correctly.
> > > > >
> > > > > I added an easy hint. Thanks for drawing our attention on it.
> > >
> > > > Any idea when the "hint" would result in an actual transfer to testing?
> > >
> > > I forget to add also ginac to that hint; should be working tonight, but
> >
> > Can you explain to me where the ginac issue arose, i.e. what create the
> > circle?
> 
> My theory is this: there's an upgrade from libginac1.2 to libginac1.3 and
> you've built against the newer one (thanks!), which isn't in testing yet.
> So your packages can't go in yet.  Upgrading GiNaC would render your old
> packages unusable, because they depend not only on the shared library but
> also on the -dev package.  Such circular dependencies have occured and
> will continue occuring.

Yup. Rafael stated something similar w.r.t. to qhull / plplot. I think that
was exactly this -- an upgrade to an underlying lib.

> >          Is there anything I can do better as Octave, octave-forge,
> > octave-sp maintainer?
> 
> Uhm, maybe yes: keep calm.  I really did love the original plan you had a
> day earlier:
> 
> On Fri, 3 Dec 2004, Dirk Eddelbuettel wrote:
> > A new Octave 2.1.64 was just released, but I'd like to get 2.1.63 into
> > testing first.
> 
> Now we are (almost) back at square one though it looks like it would've
> taken only one day to get everything in.  :-(

Such is life. It also took us 5 days to wake to this. Had we alerted
debian-release earlier, or with complete info at the time we did it, we'd
have 2.1.63 in now.

But what we need in there is 2.1.64 which will be the new stable. So let's
watch over how the pieces build.

Dirk


> I would not complain about this if we were able to clearly see such
> problems ahead during the ten day period.  But experience has shown that
> we routinely fail to recognize them from the output of Bj?rn's scripts.
> 
> The reason of this, one could conject, is that the output during that ten
> day period is just a rather barren "...is too young".  :-(
> 
> Then, after the ten day have passed, people (including myself) start
> panicking.
> 
> Regards
>   -richy.
> -- 
>   .''`.  Richard B. Kreckel
>  : :' :  <kreckel@debian.org>
>  `. `'   <kreckel@ginac.de>
>    `-    <http://www.ginac.de/~kreckel/>

-- 
If you don't go with R now, you will someday.
  -- David Kane on r-sig-finance, 30 Nov 2004



Reply to: