Timo Heister

I am glad that this is the issue. Can you open a PR to fix GridGenerator::half_hyper_shell?

I thought about this some more and I have to admit that I am confused. The flat faces are all on a plane through the origin and it should not...

Does the issue disappear with an additional global refinement?

Yes, like i thought. Can you reduce the test case to a single call to normal_vector given the cell?

Can you create a minimal example that just contains a single call to normal_vector and post it here?

Yeah. It can still be simplified a bit, but this is probably good enough. @luca-heltai Do you have a suggestion for fixing this problem? See the last comment for a...

Great, it would be great to test a few more cases, yes. I would suggest we try all possible cases inside ASPECT (including solving a Stokes system) but add some...

Two ways to verify and we should do both: - if you remove all finer cells from the triangulation then the active version should be identical - solve a Stokes...

Your ASPECT version seems to be quite old. Can you retry with the current development version, please? Is this the output without any changes to the .prm file?

@zjiaqi2018 Can you try cookbooks/initial-condition-S20RTS/S20RTS.prm with the current ASPECT master and run it for a while to check if it crashes for you as well?