bzr push into mysql-trunk branch (tor.didriksen:3588 to 3589) Bug#52123

[Tor Didriksen]

 3589 Tor Didriksen	2011-02-02 [merge]
      Empty merge of Backport of fix for Bug#52123

 3588 Ole John Aske	2011-02-02
      Fix for bug#56690 'Wrong results with subquery with 
                         GROUP BY inside < ANY clause'
            
      A subquery of the form :
      
       1) SELECT ... WHERE <column> < ANY(select <column>...)
      
      Is transformed into the form by the optimizer:
      
        2) SELECT ... WHERE <column> < (select MAX(<column>)...)
      
      The Min/Max aggregation of subquery columns is implemented in 
      'class select_max_min_finder_subselect'. 
      
      The handling of NULL values in this implementation was incorrect as
      they was interpreted as a NULL-value being '>' than any other value being
      compared. This is incorrect wrt. SQL semantics which specifies that
      NULL values are 'undefined' and should be removed as soon as a non-NULL
      value is encountered.
                  
      This fix changes implementation of all 
      select_max_min_finder_subselect::cmp_<type>() methods to follow the
      correct SQL semantics as described above. (Which also simplifies the
      logic IMHO)
                  
      It also changes the methods to be 'private' within 
      class select_max_min_finder_subselect.

    modified:
      mysql-test/include/subquery.inc
      mysql-test/r/subquery_nomat_nosj.result
      mysql-test/r/subquery_none.result
      sql/sql_class.cc
      sql/sql_class.h
No bundle (reason: useless for push emails).