Missing warning about incorrect axisymmetric BC setups

[hapfang]

This is a reminder to fix the warning regarding the incorrect axisymmetric BC setups. I am running an axisymmetric case recently, and got some issues when the first edge/face of related element was not aligned along the axis. As a result, Nek5000 produces a bunch of NaNs without any specific warning about the "A " BC setting. I was told this should not be the case.
It looks like I am not able to assign the issue. If someone could, please assign it to me, Dillon, and/or Ananias. Thanks.

[stgeke]

There is a check (CHKAXCB) in setup_topo. Can you please investigate why it didn't work.

[dshaver-ANL]

This just came up again. This happens with third-party meshes because setup_topo is called early in nek_init, but the BCs aren't assigned until at least usrdat or usrdat2. It doesn't look like setup_topo actually needs the A BC applied, so I think the best solution would be to move or add a call to CHKAXCB somewhere after usrdat2. We would also need to update exo2nek, gms2nek, etc. to rotate any elements with axis BCs.

[stgeke]

I don't get it - what needs to be updated in exo2nek etc?

[dshaver-ANL]

The meshing software makes no guarantees about how any particular element is oriented, but the A BC always needs to be on face 1. We can guarantee this is satisfied by checking for it when converting a mesh. The element orientation is defined by the order the nodes are specified, so we'd need to specify the corresponding elements in the correct order such that A is always on face 1 (see figure). I don't know the specifics of how each converter works so I don't know how easy this is to do, but it is fairly simple to do with an rea mesh.

The alternative is to dump the rea file and load it into prenek, which I'm pretty sure does the rotation for you.