AMR in schwarzschild with polar boundary condition

[LuisFLongo]

Hello everyone
I am interested in using the Schwarzschild coordinate class with AMR.
As I am including the whole z>0 plane, I am including the pole in the simulation.
For this reason I am using the 'polar' BC at $$\theta=0$$ and 'reflective' at $$\theta=\pi/2$$
When I added an user defined AMR condition, I got the following message:

"
The use of AMR with any 'polar' or 'polar_wedge' boundary
flags is currently unsupported
"
which I traced back to src/bvals/utils/check_polar.cpp

I took the black-box approach to run the simulation by basically commenting this warning/shut-down off.
So far the code hasn't crash or raised any apparent problems.

May some one give me an overview of why are 'polar' or 'polar_wedge' boundary flags "not allowed" in AMR ?

[tomidakn]

The polar(_wedge) boundary takes average of all the EMFs on the pole because they should be the same. For this procedure, all the MeshBlocks contacting the pole at the same radius must be on the same refinement level, but currently no logic to maintain this condition is implemented. So, when some (not all) the MeshBlocks contacting the pole are refined, it will probably crash. As long as you do not refine the region near the pole, it may work, but it is not a guaranteed behavior.

[LuisFLongo]

may I just not refine the mesh block around the poles then? and get away with this?

[tomidakn]

Maybe, or maybe not. Even you do not flag a MeshBlock to be refined, AMR may automatically refine it when its neighbor MeshBlock is refined to a higher level, because our implementation allows only one level difference between neighboring MeshBlocks. So if you refine MeshBlocks far away from the pole, it may work, but it is not guaranteed. Alternatively, if you flag all the MeshBlocks around the pole even when only one MeshBlock is flagged, it should work. I'd appreciate if you could implement such a logic.