> It seems that there is a logical bug in "KellyErrorEstimator" class, it
> takes a vector field (each field can be multi-component itself), so it
> should takes a vector boundary conditions (each one for each field), but
> it just take one neumann bc object and so use same bc for all fields.
Function objects represent, in general, vector-valued functions. If we want
a scalar function then that is just a special case of that.
> to see more about what happen refer to "error_estimator.cc": line 1331
> or line 1344, you can see that the same boundary condition is taken into
> account for all "n_solution_vectors"
It queries the boundary conditions as vector of length n_components, just
as needed.
> PS1: in this estimator for dirichlet faces, no action is considered, my
> question: does this mean that this estimator may not work as well for
> Dirichlet problems (in constrast to neumann ones)?
No, it should work just as fine. It's just that there is no edge residual
error associated with Dirichlet boundary conditions.
W.
-------------------------------------------------------------------------
Wolfgang Bangerth email: [EMAIL PROTECTED]
www: http://www.math.tamu.edu/~bangerth/
_______________________________________________
