[Debian-med-packaging] Mgltools are escaping our sentinel (Was: Bug#592417: marked as done)

[Andreas Tille]

On Mon, Dec 06, 2010 at 11:15:40AM +0100, Steffen Möller wrote:
> On 12/06/2010 10:58 AM, Andreas Tille wrote:
> > 
> > I wonder how we can prevent such problems in the future.  IMHO one
> > problem is that all the mgltools packages are hidden from all our
> > sentinels because they are not mentioned in the tasks files.  I once
> > talked with Steffen about this but he thinks they should not be
> > mentioned there.  (I do not remember the reasons any more but I
> > remember that I was not convinced but trusted him as the maintainer
> > and user of these packages.)
> 
> The reason is that those packages are perceived as one inseparable
> thing by the user. Only for programmers the distinction makes some
> sense and some fraction of the mgltools, like -vision, only use
> some fraction of the modules.

I admit I continue in failing to understand the reasoning because we
have also means for this.  If users should see only one inseparable
thing, just build a mgltools metapackage (not a Blend-ish one) which
depends from all mgltools packages and this metapackage could be
included into the biology task.  For developers simply add those
packages to med-dev and we are safe that the packages are in our focus.
 
> Another issue is that the mgltools-X packages in part should just
> be renamed to python-X, since those are just swig wrappers around
> some C libraries. So I am not completely convinced that everthing
> should stay as it is for long but I do not have the time for
> everything.

Well, this might be an issue but does not change the principle.  If a
package has to be renamed it is fine - wee have to rename it in the
tasks files as well.
 
> I personally started to doubt that we really need the mgltools with us.
> If it was used more frequently, then we would have seen more complaints.

Taht's finally your decision.  We actually have some low popcon packages
in Debian Med and I do not consider this as a big problem as long as
these are properly maintained.  The bugs reportet against
mgltools-utpackages are not the problems usually reported by plain users
of the package but rather from a distribution QA perspective - so the
missing complains from users side is not an argument here.  This is no
real vote from my side to keep the package.  I just think you had some
reason to introduce it (and using it in your personal setup is a
perfectly valid reason IMHO).  But if we keep it here this should be
done properly.

> My local setup worked for me, otherwise I would have spotted the
> issue myself earlier.

Well, having an unusual dir does not necessarily affect the functionality,
but I hope you have spotted the lintian error:

  http://lintian.debian.org/full/debian-med-packaging@lists.alioth.debian.org.html#mgltools-utpackages

BTW, I have a lintian sentinel for Blends packages somewhere on my long
term todo list as well.   

> > I would like to discuss means to enable us effectively watching our
> > packages and I wonder what method you would like to see implemented.  I
> > I have a clear preference for 3. because it fits the content and uses
> > uses an available (but rarely used and known) method.
> 
> We can discuss this in January and start experimenting with it.

While we can surely do this I would love to see the mgltools packages in
a better shape until then (renaming, metapackage for all of them and
including packages into bio-dev task).

Kind regards

       Andreas.

-- 
http://fam-tille.de