[Mitgcm-support] ecco e23: swapping of obcs control.

[mitgcm-support at dev.mitgcm.org]

Interesting finding:

because OBCS are applied for consistency in obcs_init_variables, the
obcs part of the control vector must also be applied. Later in s/r
packages_init_variables,  the arrays for all the xx_ control terms are
initialized to zero. I've included an initialization for xx_obcs arrays
there. However, this initialization comes after the obcs control terms
have already been needed once. This leads to a problem when the xx_obcs
arrays are swapped, a.k.a. when they are only read once every control
period. My solution, which seems to work, is to put obcs_init_variables
after ecco_init in packages_init_variables. As far as I can tell, this
problem should only relate to the obcs package because none of the other
packages are applied to the physical fields at this early stage of a
model run.  

So the obcs control vector "swaps" accurately now, and I can send you
the code when you want to check it in.

Jake