[dune-pdelab] [dune-pdelab-commit] dune-pdelab r2233 - trunk/dune/pdelab/localoperator

[Christian Engwer]

> > I guess, I'm just too dumb, but I don't understand the problem?!
> 
> As pointed out, this is a hack. I am under the impression that the
> problem is in the Interface. Probably it is not yet finished and still
> in the transition from Navier-Stokes with fixed rho to Navier-Stokes
> with variable rho (patch r1954). In the NavierStokesParameterTraits it reads: 
> 
> typedef Dune::FieldVector<DomainField,dimDomain> Domain;
> 
> Unfortunately, I have only a PDELab:LocalVector in jacobian_volume of
> NavierStokesMass and  I do not sure which position to take in the
> entity for varying rho. I would leave to the original other of patch
> r1954 to decide this. (In the meanwhile for pdelab-howto at least the
> old functionality with constant rho works.)

OK, I had a look at the NavierStokesMass operator... it is built upon
the PowerL2 operator and just adjusts the weight... obviously this is
behaviour is broken. The operator should, like for the spatial
operator, evaluate rho inside the integration loop.

Before your patch, the code didn't compile, becaues it was broken, now
it compiles, but is still broken, I'm not convinced that this is what
we want.

Besides, we have the ElementGeometry, which should know about the
coordinate type and the dimension.

Christian