[Pkg-octave-devel] HDF5 transition is finished

[Rafael Laboissiere]

* Thomas Weber <tweber at debian.org> [2012-02-24 11:29]:

> On Thu, Feb 23, 2012 at 07:26:22PM +0100, Jörg-Volker Peetz wrote:
> > Sébastien Villemot wrote, on 02/23/12 18:52:
> > It seems still not possible to use the package "libhdf5-openmpi-dev" instead
> > of "libhdf5-serial-dev" together with octave 3.2 (please see my previous
> > e-mail
> > http://lists.alioth.debian.org/pipermail/pkg-octave-devel/2012-February/008326.html
> > ). With version 1.8.4 of "libhdf5-openmpi-dev" it worked.
> 
> Yes, and that's (probably) intentional. 
> http://bugs.debian.org/601511

Would this mean that the octave package will not be able to be installed
together with any package depending on libhdf5-openmpi-7?  They are
currently, in testing:

$ apt-cache rdepends libhdf5-openmpi-7
libhdf5-openmpi-7
Reverse Depends:
  libslepc3.2
  libpetsc3.2
  gmsh
  libxdmf2
  libsiloh5-0
  meep-openmpi
  libmeep-openmpi6
  libmedimport0
  libmedc1
  libmed1
  libmed-tools
  libhdf5-openmpi-dev
  libhdf5-openmpi-7-dbg
  code-saturne-bin

If you go further one level of reverse dependencies, then the number of
packages becomes disturbingly high:

$ apt-cache rdepends $(apt-cache rdepends libhdf5-openmpi-7 | grep "^  ") | grep "^  " | sort | uniq | wc -l
32

Even worse, octave-msh appears in the list!  This is due to the
dependency on gmsh.

That will be potentially a problem for the Debian users.  Is it really
the case that the octave package could not depend alternatively on
libhdf5-7 or libhdf5-openmpi-7?

Rafael