dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, Sep 04, 2006 at 02:56:07AM +0200, Johan Jansson wrote:
> On Sun, Sep 03, 2006 at 06:15:23PM +0200, DOLFIN wrote:
> > This is an automated message from the FEniCS compile bot.
> > Compilation failed with configure options --enable-debug --enable-petsc --with-petsc-dir=/home/dolfin/local/lib/petsc.
> > Revision number and compilation log follow below.
> > 
> 
> ...
> 
> n/local/lib/petsc -I/home/dolfin/local/lib/petsc/bmake/linux-gnu-cxx-real-debug -I/home/dolfin/local/lib/petsc/include -I/home/dolfin/local/lib/petsc/externalpackages/mpich2-1.0.3/linux-gnu-cxx-real-debug/include -I/usr/include/libxml2 -I../../src/modules/ -I../../src/pydolfin -DHAVE_PETSC_H=1 -DDEBUG=1 -g -O2 -Wall -Werror -ansi -pedantic -Wno-long-long -std=c++98 -MT _dolfin_la-dolfin_wrap.lo -MD -MP -MF .deps/_dolfin_la-dolfin_wrap.Tpo -c dolfin_wrap.cpp  -fPIC -DPIC -o .libs/_dolfin_la-dolfin_wrap.o
> > dolfin_wrap.cpp: In function 'PyObject* _wrap_FEM_lump(PyObject*, PyObject*)':
> > dolfin_wrap.cpp:50408: error: no matching function for call to 'dolfin::FEM::lump(const dolfin::uBlasMatrix<dolfin::ublas_sparse_matrix>&, dolfin::uBlasVector&)'
> > ../../src/kernel/fem/dolfin/FEM.h:91: note: candidates are: static void dolfin::FEM::lump(const A&, X&) [with A = dolfin::PETScMatrix, X = dolfin::PETScVector]
> > make[3]: *** [_dolfin_la-dolfin_wrap.lo] Error 1
> > make[3]: Leaving directory `/home/dolfin/tmp/cron/compilebot/dolfin/src/pydolfin'
> 
> ...
> 
> I've now removed the generated interface from the distribution. This
> means that the interface has to be regenerated when compiling
> PyDOLFIN, which implies that SWIG is required. This needed to be done
> to support choosing between PETSc and uBlas on configure.
> 
> Perhaps it would be possible to supply a generated interface for one
> of the options, but it requires a bit of work, and doesn't feel like a
> clean solution. SWIG is a standard tool and very easy to get hold of,
> so it's not a steep requirement.
> 
>   Johan
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev

ok for me.

Is there a simple solution for letting users know that they may
disable PyDOLFIN if SWIG is not present? We're using AC_PROG_SWIG to
check for SWIG so if SWIG is not present, it will just print an error
message without letting us print an informative message about
disabling PyDOLFIN.

/Anders