1) Replaced calls to "float" by calls to "real" and "dble" in
"phys_cosp". "call construct_cosp_gridbox(float(itap),..." was a bug
since the corresponding dummy argument has the type "double
precision". (And "float" is not in the Fortran standard.)
2) Modifications for the program "create_etat0_limit" only, "gcm" is
not touched:
2.1) Removed generic interface "startget" in module "startvar". Replaced
calls to "startget" by calls to "startget_phys2d", "startget_dyn" and
"startget_phys1d".
2.2) Simplified the interface of "startget_dyn" and made it more secure by
removing arguments for sizes of arrays and using assumed-shape array
arguments.
2.3) Corrected bug in call to "startget_dyn" for northward velocity. See
ticket #25 in Trac.
2.4) Collected procedures "inter_barxy", "inter_barx", "inter_bary",
"ord_coord" and "ord_coordm" in a module. "inter_barxy" is the only
public procedure in the module. Rewrote those five procedures: removed
arguments for sizes of arrays; used assumed-shape array arguments;
used array expressions; translated some spaghetti code in "inter_bary"
into structured code; corrected bug in "inter_bary", described by
ticket #26 in Trac.
2.5) Corrected a bug in "inter_bary": "y0" was initialized at 0
instead of -90. This bug made values at the south pole wrong. (This bug
has been acknowledeged by P. Le Van.)
2.6) Made the variable "champ" in procedure "limit_netcdf" an array of
rank 2. Thus it can be an argument in the call to the newly-secured
"inter_barxy".
2.7) The files "start.nc", "startphy.nc" and "limit.nc" are impacted
by this revision. There is a significant change in the variable "vcov"
of "start.nc". See
http://web.lmd.jussieu.fr/~lglmd/Rev_1293/index.html. Note also the
changed value of "vcov" at the south pole. There is very little change
in other variables, in all three files.
