diff --git a/mysql-test/r/subselect.result b/mysql-test/r/subselect.result
index 26fe129feed..49cf73677b1 100644
--- a/mysql-test/r/subselect.result
+++ b/mysql-test/r/subselect.result
@@ -4555,8 +4555,6 @@ SELECT * FROM t1
 WHERE EXISTS (SELECT DISTINCT a FROM t2 WHERE t1.a < t2.a ORDER BY b);
 pk	a
 1	10
-3	30
-2	20
 DROP TABLE t1,t2;
 CREATE TABLE t1 (a INT, b INT, PRIMARY KEY (a), KEY b (b));
 INSERT INTO t1 VALUES (1,NULL), (9,NULL);
diff --git a/sql/sql_select.cc b/sql/sql_select.cc
index f2007f609e0..c097c4d16ef 100644
--- a/sql/sql_select.cc
+++ b/sql/sql_select.cc
@@ -1482,12 +1482,19 @@ JOIN::optimize()
         DBUG_RETURN(1);
       }
     }
-    
+    /*
+      Calculate a possible 'limit' of table rows for 'GROUP BY': 'need_tmp'
+      implies that there will be more postprocessing so the specified
+      'limit' should not be enforced yet in the call to
+      'test_if_skip_sort_order'.
+    */
+    const ha_rows limit = need_tmp ? HA_POS_ERROR : unit->select_limit_cnt;
+
     if (!(select_options & SELECT_BIG_RESULT) &&
         ((group_list &&
           (!simple_group ||
            !test_if_skip_sort_order(&join_tab[const_tables], group_list,
-                                    unit->select_limit_cnt, 0, 
+                                    limit, 0,
                                     &join_tab[const_tables].table->
                                     keys_in_use_for_group_by))) ||
          select_distinct) &&