(*)Shelfwave_set_OBC_data works with grid rotation

[Hallberg-NOAA]

Modified shelfwave_set_OBC_data() to work properly with grid rotation of 90, 180 or 270 degrees. With these changes the ESMG shelfwave test case is now giving bitwise identical answers under grid rotation. This commit also corrects a recently added problem with potentially unallocated arrays being used to test whether a pointer is set to the unrotated tracer reservoirs in cases where there are no tracers. This commit does change (and fix) answers when grid rotation is used, but answers are bitwise identical when there is no grid rotation, and there are no changes to any public interfaces.

[kshedstrom]

This code looks fine and I've seen that it works for @Hallberg-NOAA. However, it also depends on changes to MOM_open_boundary.F90 which I don't have, so I can't directly test this yet.

[Hallberg-NOAA]

This commit is ready for review, but some of the bug fixes in PR #905 and PR #907 are beneficial to this case and should be handled first.

[kshedstrom]

This fails like Tidal Bay does, but runs fine with this little change to MOM.F90:

diff --git a/src/core/MOM.F90 b/src/core/MOM.F90
index 989d12e37..1d97df9b8 100644
--- a/src/core/MOM.F90
+++ b/src/core/MOM.F90
@@ -3115,11 +3115,13 @@ subroutine initialize_MOM(Time, Time_init, param_file, dirs, CS, &
           OBC_in%cff_normal_u => CS%OBC%cff_normal_u
           OBC_in%cff_normal_v => CS%OBC%cff_normal_v
         endif
-        if (any(CS%OBC%tracer_x_reservoirs_used)) then
-          OBC_in%tres_x => CS%OBC%tres_x
-        endif
-        if (any(CS%OBC%tracer_y_reservoirs_used)) then
-          OBC_in%tres_y => CS%OBC%tres_y
+        if (CS%OBC%ntr > 0) then
+          if (any(CS%OBC%tracer_x_reservoirs_used)) then
+            OBC_in%tres_x => CS%OBC%tres_x
+          endif
+          if (any(CS%OBC%tracer_y_reservoirs_used)) then
+            OBC_in%tres_y => CS%OBC%tres_y
+          endif
         endif
       ! else
         ! Do we need to swap around the u- and v-components?

I approve once the case of no tracers is handled.

[Hallberg-NOAA]

I have updated this PR to address this problem with an unassociated being referenced with the following even simpler fix:

@@ -3114,12 +3114,12 @@ subroutine initialize_MOM(Time, Time_init, param_file, dirs, CS, &
           OBC_in%ry_oblique_v => CS%OBC%ry_oblique_v
           OBC_in%cff_normal_u => CS%OBC%cff_normal_u
           OBC_in%cff_normal_v => CS%OBC%cff_normal_v
         endif
-        if (any(CS%OBC%tracer_x_reservoirs_used)) then
+        if (associated(CS%OBC%tres_x)) then
           OBC_in%tres_x => CS%OBC%tres_x
         endif
-        if (any(CS%OBC%tracer_y_reservoirs_used)) then
+        if (associated(CS%OBC%tres_y)) then
           OBC_in%tres_y => CS%OBC%tres_y
         endif
       ! else
         ! Do we need to swap around the u- and v-components?

[marshallward]

@kshedstrom I'm going to tentatively approve this based on your previous feedback, but please let us know if the associated() change does not fix the issue.

[kshedstrom]

Yes, I approve the new version.

[Hallberg-NOAA]

This revised PR has passed pipeline testing at https://gitlab.gfdl.noaa.gov/ogrp/mom6ci/MOM6/-/pipelines/27714.