Fix SolutionTransfer when coarsening from FE_Nothing #8860
Conversation
source/numerics/solution_transfer.cc
Outdated
| { | ||
| tmp.reinit(dofs_per_cell, true); | ||
| const unsigned int old_index = | ||
| pointerstruct->second.active_fe_index; |
There was a problem hiding this comment.
Isn't old_index already available above? Edit: and you don't need it at all?!
There was a problem hiding this comment.
Yes, forgot to remove after restructuring. Fixed now.
|
/rebuild |
|
Looks good! I am glad you found the bug. :-) |
| const unsigned int dofs_per_cell = this->get_fe().dofs_per_cell; | ||
| if (dofs_per_cell == 0) | ||
| { | ||
| interpolated_values = 0; |
There was a problem hiding this comment.
@masterleinad : I'm way late to this game, but isn't it the case that if you end in this branch, interpolated_values needs to have size zero anyway? If so, one could just skip the if and jump straight to the else. If you agree, I can write that patch.
There was a problem hiding this comment.
No, in the test for this pull request the size of interpolated_values is 4 since the new element is a FE_Q<2>(1).
There was a problem hiding this comment.
Oh, I see, it's the other way around. Thanks for confirming!
I was seeing some problems when trying to use adaptive refinement where some cells were using
FE_Nothing. This pull request makes sure that coarsening works and the contributions from these cells are ignored.