Remove gcc2.8 workaround with DoFSubstructAccessor
-Replace verbatim implementation of setw(int) in parameter-handler.cc
- when a reasonable (linkable) library implementation is available
- again.
-
-
Use the C++ standard string classes instead of the libg++ String
class.
Let the FE objects take DoF-iterators rather than Tria-iterators.
+
Check usage of vectors: functions should assume that the objects have
the right size already.
-Set dirichlet bc to the solution vector after solving (setting it
- beforehand is not enough, the search directions for CG contain
- components in the directions of these nodes, even if they are not
- coupled to the other dofs. Strange.)
-Update DataIO for ucd i/o of boundary flags and material values.
-Update DataIO for ucd i/o of boundary flags and material values.
+Update DataIO for ucd i/o of boundary flags and material values:
+ is it correct if for the number of elements I give the number
+ of cells + number of faces with boundary id .neq. 0 ? Or must
+ I give the number of cells?
+
+
+Update DoFHandler::distribute_cell_to_dof_vector for higher order
+ elements (it can presently only handle elements with the only dofs
+ on the vertices).
+Use auto_ptr when available for the DataOut::boundary_faces function.