← Back to team overview

dolfin team mailing list archive

Re: [HG DOLFIN] Fix bug in init: only init when needed. Unit tests work again now.

 

Quoting Anders Logg <logg@xxxxxxxxx>:

> On Fri, Nov 14, 2008 at 09:24:37PM +0100, DOLFIN wrote:
> > One or more new changesets pushed to the primary dolfin repository.
> > A short summary of the last three changesets is included below.
> > 
> > changeset:   5158:8f5618bc84be42a36517969786370689478b2fe2
> > tag:         tip
> > user:        Anders Logg <logg@xxxxxxxxx>
> > date:        Fri Nov 14 21:24:28 2008 +0100
> > files:       dolfin/mesh/Mesh.cpp sandbox/misc/main.cpp
> test/unit/function/cpp/test.cpp test/unit/mesh/cpp/test.cpp
> > description:
> > Fix bug in init: only init when needed. Unit tests work again now.
> 
> I think we're back on track again. The mesh-ordering/init strategy is
> worked out (at least I think it looks good), and there is no call to
> mesh.order() in any of the demos.

The function

void PointCell::order(Cell& cell) const
{
  error("PointCell::order() not defined.");
}

in dolfin/mesh/PointCell.cpp kills the assembly over facet integrals.
Should it throw an error or should it simply do nothing?

Kristian

> Two demos are still failing:
> 
>   pde/periodic: segmentation fault
>   pde/sym-dirichlet-bc: assemble_system missing
> 
> Is anyone planning to fix assemble_system? I guess I need to look at
> PeriodicBC.
> 
> -- 
> Anders
> 




Follow ups

References