Improvement in slowproc_update

[maignan]

There are several calls to flioinqv when reading PFT maps:

..
CALL flioinqv (fid,v_n="time_counter",l_ex=l_ex,nb_dims=nb_dim,len_dims=l_d_w)
... 
CALL flioinqv (fid,v_n="lon",l_ex=l_ex,nb_dims=nb_dim,len_dims=l_d_w) 
...
CALL flioinqv (fid,v_n="lat",l_ex=l_ex,nb_dims=nb_dim,len_dims=l_d_w) 
...
CALL flioinqv (fid,v_n="veget",l_ex=l_ex,nb_dims=nb_dim,len_dims=l_d_w) 
...

If the requested field doesn't exist, flioinqv doesn't stop but returns a l_ex value which is not TRUE.

Unfortunately this value is not tested in slowproc_update which can then go on with weird l_d_w dimension values, causing a later crash.

This value has to be tested and the program stopped using a call isplerr_p(3).

It is to be noted that readdim2.f90 uses a similar flinquery_var subroutine. We should look for some harmonization.

[ajornet]

flioinqv is now checked at revision [2925/perso/albert.jornet/ORCHIDEE-MICT]

It remains to check file source:branches/ORCHIDEE-MICT/ORCHIDEE/src_driver/readdim2.f90 the variable flinquery_var

[jgipsl]

Done in the trunk rev [2929]. This is as commit 2925 above but also remove IF (exv) THEN further down in the code. This is always true in the modified version.

Nothing will be done for readdim2.