WITH GROUP BY ON DUPLICATED FIELDS

If a GROUP BY clause contains duplicates of the same field
and group range access is used to resolve the query, the optimizer
incorrectly believed that the range access would not provide 
the ordered output required to do grouping. Thus, the already
ordered rows were ordered once more using filesort (which yields
correct result but is not optimal). end_send() then thought 
that group range access method had not been used (see 
is_using_loose_index_scan()) and therefore did not copy 
non-aggregated fields to the resultset.

The fix is to remove duplicated (and redundant) fields from 
the GROUP BY and ORDER BY clauses, which in turns makes 
the optimizer realize that the output of the range access method
is, in fact, ordered.