some comments on output from configure.py when downloading external packages

[Lisandro Dalcin]

On Wed, Aug 19, 2009 at 6:42 PM, Barry Smith<bsmith at mcs.anl.gov> wrote:
>
> On Aug 15, 2009, at 6:40 PM, Lisandro Dalcin wrote:
>
>> When asking configure.py to download Hypre and ML, we get output like
>> this:
>>
>> 1) for Hypre ...
>> You must register to use hypre at
>> http://www.llnl.gov/CASC/hypre/download/hyprebeta_cur_agree.html
>> Once you have registered, config/configure.py will continue and
>> download and install hypre for you
>> 2) for ML ...
>> You must register to use ml at http://trilinos.sandia.gov/
>> Once you have registered, config/configure.py will continue and
>> download and install ml for you
>>
>> These messages are wrong, as the configuration does not stop.
>> Secondly, a thing that I strongly object, is that after running
>> configure, in home directory I now I have two files '.hypre_license'
>> and '.ml_license' ... IMHO, configure.py should not write stuff in my
>> home directory... why not write these files in $PETSC_DIR... Even
>> more, do we still need these files if configure.py downloads and
>> installs the stuff anyway?
>
>   I have simplified it, removed the use of the home directory and changed
> the messages to simply ask the person to register (no longer imply that they
> have to register if it does not work)
>

Many, many thanks, Barry ...

>>
>> After looking at the websites of Hypre and ML, it seems that both are
>> LGPL... So... Do we still need at all that license stuff in PETSc's
>> build system for these two external packages?
>>
>> Lastly, this annoying thing about ML:
>>
>>        UNABLE to CONFIGURE with GIVEN OPTIONS    (see configure.log
>> for details):
>>
>> ---------------------------------------------------------------------------------------
>> Installing ML requires explicit root directory of MPI
>> Run config/configure.py again with the additional argument
>> --with-mpi-dir=rootdir
>>
>> Are you really sure this issue is hard to fix? Is it worth to give it a
>> look?
>
>   If you want to fix it and can fix it, feel free. We think it is a design
> decision of the ML folks, but maybe you can modify ml.py so that it works.
> Or you could send a fix of ml's configure to the ml people so they can get
> it into Trilinos.
>

I'll take a look ASAP... In fact, I asked about this because ML's
./configure --help shows this
....
 --enable-mpi            MPI support
...
  --with-mpi-compilers=PATH
                          use MPI compilers mpicc, mpif77, and mpicxx, mpic++
                          or mpiCC in the specified path or in the default
                          path if no path is specified. Enables MPI
  --with-mpi=MPIROOT      use MPI root directory (enables MPI)
  --with-mpi-libs="LIBS"  MPI libraries ["-lmpi"]
  --with-mpi-incdir=DIR   MPI include directory [MPIROOT/include] Do not use -I
  --with-mpi-libdir=DIR   MPI library directory [MPIROOT/lib] Do not use -L

So it seems that should be possible... I'll just need to be careful
about using either --with-mpi-compilers or
--with-mpi-[libs|incdir|libdir] depending on the info I can get from
BuildSystem... And also take into account MPIUNI, of course...



>>
>>
>> --
>> Lisandro Dalcín
>> ---------------
>> Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
>> Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
>> Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
>> PTLC - Güemes 3450, (3000) Santa Fe, Argentina
>> Tel/Fax: +54-(0)342-451.1594
>
>



-- 
Lisandro Dalcín
---------------
Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
PTLC - Güemes 3450, (3000) Santa Fe, Argentina
Tel/Fax: +54-(0)342-451.1594