dolfin team mailing list archive

["Åsmund Ødegård" <aasmund@xxxxxxxxx>]

On Sun, Mar 16, 2008 at 6:20 PM, Jed Brown <fenics@xxxxxxxx> wrote:
>
> On 16 Mar 2008, aasmund@xxxxxxxxx wrote:
>
>  > On Sun, Mar 16, 2008 at 4:18 PM, osman <osman@xxxxxxxx> wrote:
>  > > Hi,
>  > >  I am trying to build dolfin from mercurial repositories. Just copied it
>  > >  today.
>  > >  I just re-built  boost 1.34.1  locally. installed it
>  > >  in /home/osman/local
>  > >
>  > [snip]
>  > >   Trying to generate pkg-config file for petsc... Can not find alternate
>  > >  for mpicc
>  > >  Can not find alternate for mpicc
>  > >  Can not find alternate for mpicc
>  > [snip]
>  >
>  > This is a bit strange. Could you check what the PCC setting in PETSc
>  > is, it's probably located in
>  >
>  > $PETSC_DIR/bmake/common/variables
>  >
>  > [Johannes, we should add this to the log....]
>  >
>  > Somehow, Petsc is reporting a C compiler (mpicc) that we are not able
>  > to find a c++ version of. Now, this should really be taken care of
>  > more gracefully, by just skipping Petsc as a dependency (and an error
>  > message)
>
>  When you have PETSc download mpich2, my recollection is that it only has
>  the binary mpicc, even if it is really a C++ compiler.  This has caused
>  some confusion for another project I'm involved in.

That doesn't sounds completely correct. At least the mpich I have that
petsc downloaded and installed, have both mpicxx, mpif77, and more.
What do you mean with "mpicc, even if it is really a C++ compiler"? Do
you mean that it use the correct backend (e.g. g++) based on the file
you feed it with? If so, do you have any suggestion on a suitable test
such that I can build that into the logic?

Åsmund