[Pkg-octave-devel] [RFU] octave3.0 3.0.0-10

[Ólafur Jens Sigurðsson]

On Sat, Mar 29, 2008 at 06:29:48PM +0100, Rafael Laboissiere wrote:

> This would reduce the quarantine time the package must stay in unstable
> before being considered for testing, which is ten days by default.  Whether
> this would make them go faster through the buildd queue, I do not know.

And this is important because ... ?
I am sorry, I allways like to know why we do things.
You think that we wont be able to get 3.0 into Lenny unless we put the
important flag on it? If so then yes, this is important.

> > Yeh, this suitsparse transition thing. I have spent some time looking
> > through the release.debian.org and other stuff for this matter and it
> > seems to me that all these packages are waiting for each other, I
> > can't see any one (or more) package that is hindering any of them.
> > What is the main problem there?
> 
> No, there are no real circular dependencies.  The fact that they are all
> depending on each other to enter testing [0] is normal and will be fixed by
> the release team through a hint.  However, all the package should be ready
> before being hinted (i.e., should build from source on all releasable arches
> and should have completed the quarantine period in sid). 
> 
> [0] http://release.debian.org/migration/testing.pl?package=octave3.0
> 
> There are, at least, three unrelated problems blocking the transition:
> 
> 1) petsc FTBFS on ia64, powerpc, and sparc [1].  The bug is fully mysterious
>    and is caused by some Python configuration failure.  On armel, hppa,
>    mips, mipsel, and s390, petsc is waiting for spooles [2].  The spooles buildd
>    on s390 is waiting for openmpi, which also FTBFS on s390 [3].
>  
>    [1] http://buildd.debian.org/~jeroen/status/package.php?suite=unstable&p=petsc
>    [2] http://buildd.debian.org/~jeroen/status/package.php?suite=unstable&p=spooles   
>    [3] http://buildd.debian.org/~jeroen/status/package.php?suite=unstable&p=openmpi
>    
> 2) openoffice.org needs buildds arm, hppa, mips, mpisel, powerpc, s390, and
>    sparc [4].  This should happen naturally, as soon as the autobuilders
>    pick the package for build.
>  
>    [4] http://buildd.debian.org/~jeroen/status/package.php?suite=unstable&p=openoffice.org
>    
> 3) octave3.0 has to be upgraded on all arches to hdf5 1.6.6, otherwise it
>    will be uninstallable in sid (this is making all packages that
>    build-depend on octave3.0-headers to FTBFS, like plplot [5]).  The hdf5
>    maintainers introduced a serious problem in version 1.6.6-1 by adding the
>    libopenmpi-dev build-dependency.  openmpi FTBFS on several arches.  I
>    already contacted the DebianGis Group [6] about this.  I am wondering
>    whether I wold be better to file bug report about the problem.
>    
>    [5] http://buildd.debian.org/~jeroen/status/package.php?suite=unstable&p=plplot
>    [6] http://lists.alioth.debian.org/pipermail/pkg-grass-devel/2008-March/003579.html

Really nice summery, thanks for that!

So the main culprit seems to be openmpi. There is a bug report out
there that is related to this [1], we should keep an eye on that one
perhaps.

[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=376833

Oli