Bug in bdytides.F90

[nemo_user]

The presence of dta_bdy with 2D variables(ssh/u2d/v2d) is inconsistent and/or dta_bdy is just not allocated with such variables to be used in bdytides.F90

! If time splitting, save data at first barotropic iteration

IF ( PRESENT(kit) ) THEN
               IF ( lk_first_btstp ) THEN ! Save slow varying open boundary data:
                  dta_bdy_s(ib_bdy)%ssh(1:ilen0(1)) = dta_bdy(ib_bdy)%ssh(1:ilen0(1))
                  dta_bdy_s(ib_bdy)%u2d(1:ilen0(2)) = dta_bdy(ib_bdy)%u2d(1:ilen0(2))
                  dta_bdy_s(ib_bdy)%v2d(1:ilen0(3)) = dta_bdy(ib_bdy)%v2d(1:ilen0(3))

               ELSE ! Initialize arrays from slow varying open boundary data:            
                  dta_bdy(ib_bdy)%ssh(1:ilen0(1)) = dta_bdy_s(ib_bdy)%ssh(1:ilen0(1))
                  dta_bdy(ib_bdy)%u2d(1:ilen0(2)) = dta_bdy_s(ib_bdy)%u2d(1:ilen0(2))
                  dta_bdy(ib_bdy)%v2d(1:ilen0(3)) = dta_bdy_s(ib_bdy)%v2d(1:ilen0(3))
               ENDIF
            ENDIF

[jchanut]

The problem is not very clear from your description.
I found that, indeed, if you have tides but you do not use flather scheme, there is a problem. Because dta_bdy%ssh is not allocated in that case.

[jchanut]

Corrected at revision 4758.
This change allows to have tides and not only a Flather scheme at open boundaries. This removes the above issue due to unallocated ssh arrays.
Still, reading ssh data at open boundaries is required even if a Flather scheme is not used (should be changed in the future although such a situation in realistic applications is unlikely).
​http://forge.ipsl.jussieu.fr/nemo/changeset/4758