#2030: Multi-block boundaries leave uninitialized boundary points ------------------------------------+--------------------------------------- Reporter: eschnett | Owner: Type: defect | Status: new Priority: major | Milestone: Component: EinsteinToolkit thorn | Version: development version Resolution: | Keywords: ------------------------------------+---------------------------------------
Comment (by hinder):
The first bug mentioned is one that we have discussed before, many years ago. One solution was to apply the BCs twice, which would probably need to be handled in Kranc. (I just looked, and cannot find a Kranc issue about this, unfortunately.) This was the reason that for a long time, my McLachlan Llama BBH parameter files used poison = 0, because setting the uninitialised points to zero was better than having them as NaN, especially when the boundary was causally disconnected. However, recently, I was not able to reproduce the problem. I thought that maybe something had been fixed/changed, though this was perhaps optimistic. Do you have a parameter file and a method for identifying that this problem has occurred?
Presumably the long-term solution is for Cactus to keep track of variables read and written in different parts of the grid. Is there a short-term solution that doesn't involve "clever" scheduling, and applying the BCs twice?
As I understand it, this problem occurs only on initial data, not during evolution. Is that your understanding as well?