dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On Tue, 2010-08-17 at 21:05 +0200, Anders Logg wrote:
> On Tue, Aug 17, 2010 at 07:59:25PM +0100, Garth N. Wells wrote:
> > On Tue, 2010-08-17 at 20:49 +0200, Anders Logg wrote:
> > > On Tue, Aug 17, 2010 at 07:48:13PM +0100, Garth N. Wells wrote:
> > > > On Tue, 2010-08-17 at 20:39 +0200, Anders Logg wrote:
> > > > > On Tue, Aug 17, 2010 at 10:53:05AM -0700, Johan Hake wrote:
> > > > > > On a related note:
> > > > > >
> > > > > > Is it possible to not print the PETSC_ARCH when it is empty.
> > > > > >
> > > > > >   message(STATUS "PETSC_ARCH=${PETSC_ARCH}")
> > > > > >
> > > > > > While it is informative, it is kind off anoying to get
> > > > > >
> > > > > >   PETSC_ARCH=
> > > > > >
> > > > > > when configuring PETSc.
> > > > >
> > > > > Try again now.
> > > > >
> > > > > Does the PETSc test work for you?
> > > > >
> > > >
> > > > Now it's broken.
> > > >
> > > > Garth
> > >
> > > Was it something I did?
> > >
> > > I've only made changes to clean up, add comments etc.
> > >
> >
> > Maybe it wasn't working 100% before. I get:
> >
> > Performing Test MULTIPASS_TEST_1_petsc_works_minimal
> > -- Performing Test MULTIPASS_TEST_1_petsc_works_minimal - Failed
> > -- Performing Test MULTIPASS_TEST_2_petsc_works_allincludes
> > -- Performing Test MULTIPASS_TEST_2_petsc_works_allincludes - Success
> > -- PETSc requires extra include paths, but links correctly with only
> > interface libraries.  This is an unexpected configuration (but it seems
> > to work fine).
> >
> > Seems to work though.
> >
> > It would be nice if PETSc would provide a petsc-config.cmake file - the
> > Trilinos one is excellent. The PETSc system is horrible for extracting
> > build data.
> 
> I get this:
> 
> - petsc_lib_dir /home/logg/scratch/lib/fenics-dev/lib
> -- Performing Test MULTIPASS_TEST_1_petsc_works_minimal
> -- Performing Test MULTIPASS_TEST_1_petsc_works_minimal - Failed
> -- Performing Test MULTIPASS_TEST_2_petsc_works_allincludes
> -- Performing Test MULTIPASS_TEST_2_petsc_works_allincludes - Failed
> -- Performing Test MULTIPASS_TEST_3_petsc_works_alllibraries
> -- Performing Test MULTIPASS_TEST_3_petsc_works_alllibraries - Failed
> -- Performing Test MULTIPASS_TEST_4_petsc_works_all
> -- Performing Test MULTIPASS_TEST_4_petsc_works_all - Failed
> -- PETSc could not be used, maybe the install is broken.
> -- Found PETSc:
> -/home/logg/scratch/lib/fenics-dev/include;/usr/lib/openmpi/include;/usr/lib/openmpi/lib;/usr/include
> 
> It seems to indicate both that PETSc could not be used and that PETSc
> is found... :-)
> 

Problem is that PETSc can create just one lib (petsc), or a collection
of libs (petscksp, etc) which is a nightmare for a configure script to
figure out and test.

> Should we try to write our own script instead? We could reuse the
> first part. It's not that hard to extract the flags from the PETSc
> bmake stuff. We've done it before.
> 

Yes, we can. I already simplified a part that wasn't working by
incorporating some of our old configure scripts. 

If the PETSc test 'works' for now, I'd prefer that we all look first at
the structure of the CMake system and how it's set up. Once we're happy
with that, we can go through and improve the tests one-by-one.

Garth


> --
> Anders