Comments regarding petsc_3.8.0+dfsg1-1exp1_amd64.changes

[Drew Parsons]

On Sat, 2018-02-24 at 16:07 +0000, Thorsten Alteholz wrote:
> Hi Drew,
> 
> one of our trainess had a look at your package and found the
> following 
> stuff. He is nice and doesn't want to reject the package but just
> have
> your comments :-).
> 
> Thanks!
>  Thorsten
> 

Thanks Thorsten.


> 
> Not lintian-clean:
>  * package-has-unnecessary-activation-of-ldconfig-trigger (1)

This seems to be a lintian bug (the detailed message for the warning
says so).  There's nothing in debian/rules that explicitly invoke
ldconfig.


>  * csh-considered-harmful (1)

Annoying to edit upstream scripts, so I haven't bothered much with
them. But in the long run it's probably good for them if we do.  These
are auxiliary utility scripts, not routinely used. Most programs would
use PETsc via the C++ API instead (or petsc4py for python)

>  * executable-not-elf-or-script (37)

matlab scripts.  Not sure why upstream marks them as executable.  I
guess it's not needed, I can remove the executable flag.

> Format in d/copyright is in http, where https is prefered.
> 
> I'm not sure whether it is possible to reproduce the following files
> (and
> whether is is required at all):
> 
> - petsc-3.8.0/share/petsc/datafiles/matrices/ contain suspicious
> files in binary
>   format. Quite possible related to some computational stuff, but i'm
> not sure.

These are example matrices used by
petsc/matlab/generatePetscTestFiles.m to test the matlab interface to
PETSc.


> - petsc-3.8.0/share/petsc/datafiles/meshes/*.exo are in CDF format,
> which is
>   rather a binary format.
> 
> - petsc-3.8.0/share/petsc/datafiles/meshes/*.med are in HDF format,
> which is
>   rather a binary format.
> 
> - petsc-3.8.0/share/petsc/datafiles/meshes/*.cgns are in ADF format,
> which is
>   rather a binary format for database storing.
> 
> - petsc-3.8.0/share/petsc/datafiles/meshes/horse.ply.bz2 is an
> archived file,
>   containing some binary data.
> 
> - petsc-
3.8.0/petsc/src/dm/impls/moab/examples/tests/input/ex3_in.h5m 
> is a
>  
 binary file in HDF format.

These are data files and meshes defining the geometries of various 2D
and 3D objects provided for use in examples and tests.    The tests and
examples that use them are located below 
/usr/lib/petscdir/petsc3.8/x86_64-linux-gnu-real/share/petsc/examples/src/
(there are different tests for the various PETSc components)

You can view the .exo fils in paraview, 
and also the horse.ply after decompressing.  

gmsh shows you the mesh in the .msh and .med files.

cgns is a format used by some computational fluid dynamics software.
The cgns library is in debian. You can view and plot using cgnsview.



> 
> - petsc-3.8.0/petsc/src/dm/impls/moab/examples/tests/input/ex3_in.h5m 
> is a
> binary file in HDF format.

dm/impls/moab's ex3 test generates a h5m file (when its -io option is
used).  I think input/ex3_in.h5m is a record of what it is supposed to
generate, which could be used to confirm that PETSc is working
correctly. You can inspect the contents using HDFCompass.  HDF5 is a
general file format for storing complex datasets.  It's very general,
so you organise your data according to your needs using your own
schema.  h5m is Moab's HDF5 file format, using their own schema to
store data in an hdf5 file,
https://www.mcs.anl.gov/~fathom/moab-docs/h5mmain.html
It's used to store finite element mesh information (vertices and
connectivity, geometry and topology).  We don't have moab in debian but
it's LGPL (https://bitbucket.org/fathomteam/moab/src), so we couldpackage it, if anyone really wants to check what's in 
petsc/src/dm/impls/moab/examples/tests/input/ex3_in.h5m



> Also d/copyritgh is not full. I've found the following file:
> 
> petsc-3.8.0/src/ksp/ksp/examples/tutorials/ex63.cxx, which is
> licensed under
> BSD-2-clause and copyrighted 2011 Sandia, but it is absent from
> d/copyright.

Thanks, I'll update copyright.

Drew