Context

additional loops were added for vectorial optimization in ldfslp.F90/ldf_slp.

Their "do ji=... / do jj=.." have not been replaced by DO loops rewriting :

https://forge.ipsl.jussieu.fr/nemo/browser/NEMO/trunk/src/OCE/LDF/ldfslp.F90 lines 247 and 323

Analysis

They are useless now. They can be removed. It won't change the results.

Fix

...

In dtatsd.F90, the 4 error status (ierr0, ierr1, ierr2, ierr3) on allocation are used to detect allocation problems: if the sum of these 4 flags >0 then ans error is detected and the program stops.

This is done even in case of ln_tint false (ierr1, ierr3 not initialized. In some case, the run stops for bad reason, just because ierr1 or ierr3 take dummy random values.

1. limwri, limwri_dimg, does not consider the standard EVP case. Output values are interpolated as it used to be with VP rheology.
2. in order to fix this bug, using lk_lim2_vp instead of directly the CPP key, array tmv must be declared in both cases even if it is used only for evp case.
3. So in dom_ice_2.F90, it is probably a good idea to eliminate CPP keys in the declaration part of the module, and to let the allocation routine decide whether the arrays are to be allocated or not.

• impacted routine : limwri_2.F90, limwi_dimg_2.h90, dom_ice_2.F90

• I attach my corrected version of these files