Abort model cleanly in ifsinterface and standalone on errors #661
Conversation
…f trying in clean end in case of namelist errors in oce_mesh
JanStreffing
left a comment
JanStreffing
left a comment
There was a problem hiding this comment.
Excellent changes. This should make missing namelists etc. easier to catch. I have just a few small changes requested.
| ! TODO logic below is also convoluted really not really for standalone | ||
| #else ! | ||
| ! TODO logic below is convoluted, COMM that is passed should be used for MPI_ABORT | ||
| ! changes are easy but need to be tested with coupled configurations |
There was a problem hiding this comment.
Does that mean changes are were easy to make or would be easy?
There was a problem hiding this comment.
It handles core parts and there are other parts like ice modules, blowup that may need more inspection on aborts, clean kind to stop normally vs really abort kind. this can be easy to do once we go through those parts, but i would leave it here for now, going into all those parts to me is another issue. @JanStreffing thanks for confirming communicator in oifs-fesom case, then nothing about that configuration to worry about with these changes.
There was a problem hiding this comment.
i want to do one small test with OpenIFS-FESOM2 before I merge.
removed an abort commented out line.
Some of old features lost in my branches that didn't make it into main, part 1.
TODO in another PR: