Minor bug in ORCA2_LIM with Agrif

[MAMM]

Hello,

In case it is of interest. I was having problems running ORCA2_LIM jointly with an AGRIF zoom in the Indian Ocean. The model was blowing after just 4 time steps or so with wild SSHs and velocities adjacent to the north fold.

All I had to do to solve this problem was to

1) replace the following three lines in dom_oce.F90:

INTEGER, PUBLIC,               DIMENSION(jpidta) ::   mi0, mi1  
INTEGER, PUBLIC,               DIMENSION(jpjdta) ::   mj0, mj1  
REAL(wp), PUBLIC, DIMENSION(jpiglo) ::   tpol, fpol

by the alternative lines:

INTEGER, PUBLIC, ALLOCATABLE, SAVE, DIMENSION(:) ::   mi0, mi1 
INTEGER, PUBLIC, ALLOCATABLE, SAVE, DIMENSION(:) ::   mj0, mj1
REAL(wp), PUBLIC, ALLOCATABLE, SAVE, DIMENSION(:) ::   tpol, fpol

and

2) add the following two lines in function dom_oce_alloc

ALLOCATE( mi0(jpidta), mi1(jpidta), mj0(jpjdta), mj1(jpjdta), STAT=ierr(11) )
ALLOCATE( tpol(jpiglo), fpol(jpiglo), STAT=ierr(12) )

Maybe someone who knows better than I how NEMO+AGRIF work together would like to check that these changes make sense and that the original formulation in nemo_v3_4 is, as I believe, not entirely correct. My architecture is pgf90_linux.

Thank you,

Miguel Angel