bzjan

I realize the question is quite old by now, but I believe we may have worked with the same somewhat deprecated subdivision code. To get that working here is a...

Thanks for clearing the index issue up. I assumed that local index and node index were the same thing. To viusalize that, I plotted them for serial and parallel code:...

I was using the `FESubdivision` elements because of their C1 smoothness. What would be better: Use other C1 smooth elements (are there any?) or track down the bug in the...

Thank you very much for the summary! Unfortunately I am working on curved 2d surfaces in 3d. Looks like I will need to track down the bug in `build_parallel_solution_vector()`

From what I found it seems that the halo/ghost elements have empty dof_indices vectors. So in the following lines: `elem_soln.resize(dof_indices.size()); for (std::size_t i=0; i

Attached is my test code: [debug_build_parallel_solution_vector.zip](https://github.com/libMesh/libmesh/files/4258484/debug_build_parallel_solution_vector.zip). It outputs the nodal solutions of all elements first and later prints all nodes that have differing values for the system.solution vector and the...

Realizing, that I did not answer your questions: 1) `dof_indices.size()` is zero. 2) So `elem_soln.size()` is zero as well 3) Since the loop `for (std::size_t i=0; i

I found two different solutions for my problem. 1) If you don't want to save the values of the ghost nodes and only look for a correctly interpolated solution vector,...

@vikramvgarg I am familiar with the example you mention. Are you suggesting, that instead of working with ghost cells for boundary conditions it would be better to use the penalty...

@vikramvgarg How would I set the the boundary value to something else but zero? The 0 in your quoted lines is the component index, right?