In iodef.xml files, level="5" is used in file definitions. It should be output_level="5"

Level is an attribute of field_definition or anything defined within the "field_definition block". It defines the level of output for the variable. 1 is very high, 10 is very low.
Output_level is an attribute of file_definition or anything defined within the "file_definition block". It defines the maximum level of output that you will find in the file. For example, if output_level = 5, you will output all variables listed in the file if their level is <= 5. Variables with level larger that 5 will not be outputted even if they are listed in the file definition.
This attribute allows you to quickly switch-on/off a set of variable within a file.

in version 1594 of step.F90, diawri is called before diahth.
Whitout key_iomput, diahth variables are therefore outputed before beeing computed...

The correction to the freshwater fluxes estimated from the previous year's budegt is being applied with the wrong sign in sbcfwb.F90. Lines 127 and 128:

IF( MOD( kt-1, kn_fsbc ) == 0 ) THEN

emp (:,:) = emp (:,:) - empold

emps(:,:) = emps(:,:) - empold

ENDIF

should be:

IF( MOD( kt-1, kn_fsbc ) == 0 ) THEN

emp (:,:) = emp (:,:) + empold

emps(:,:) = emps(:,:) + empold

ENDIF

Note this is for case nn_fwb=2 which corrects the annual global mean based on the mean ssh change from the previous year. This is incorrectly commented as case nn_fwb=1 in the namelist and kn_fwb=1 in the header for subroutine sbc_fwb.