The read_group behaviour has changed and read_group no longer sum on the field you are grouping.
Instead agregate on the field we want to sum and use no groupby.
For future version, grouping on a user such as what is done in batch mode is a better way to process.
if definition.computation_mode == 'sum':
field_name = definition.field_id.name
- res = obj.read_group(cr, uid, domain, [field_name], [field_name], context=context)
+ # TODO for master: group on user field in batch mode
+ res = obj.read_group(cr, uid, domain, [field_name], [], context=context)
new_value = res and res[0][field_name] or 0.0
else: # computation mode = count