Fixed bug mdev-10053.

The implementation of the walk method for the class Item_in_subselect
was missing. As a result the method never traversed the left operand
of any IN subquery predicate.

Item_exists_subselect::exists2in_processor() that performs the
Exist-To-In transformation calls the walk method to collect info
on outer references. As the walk method did not traverse the
left operands of the IN subqueries the outer references there
were not taken into account and some subqueries that were actually
correlated were marked as uncorrelated. It could lead to an
attempt of the materialization of such a subquery.

Also added a cleanup for some test cases merged from 5.5.

1 files changed, 28 insertions, 0 deletions

diff --git a/mysql-test/t/subselect_exists2in.test b/mysql-test/t/subselect_exists2in.test
index 9450ef71494..a4fdbe5c50b 100644
--- a/mysql-test/t/subselect_exists2in.test
+++ b/mysql-test/t/subselect_exists2in.test
@@ -758,6 +758,34 @@ deallocate prepare stmt;
 drop view v1;
 drop table t1,t2;
 
+--echo #
+--echo #MDEV-10053: EXIST to IN transformation turned down
+--echo #
+
+CREATE TABLE t1 (
+ pk INT, f1 INT NOT NULL, f2 VARCHAR(3), f3 INT NULL, PRIMARY KEY(pk))
+ENGINE=MyISAM;
+INSERT INTO t1 VALUES (1,1,'foo',8), (2,5,'bar',7);
+
+let $q=
+SELECT STRAIGHT_JOIN sq1.f2 
+  FROM ( SELECT * FROM t1 ) AS sq1
+    WHERE EXISTS ( SELECT * FROM t1 AS sq2 
+                     WHERE sq1.`pk` IN ( SELECT f1 FROM t1 ) AND sq2.f1 = sq1.f1 );
+
+set @optimizer_switch_save=@@optimizer_switch;
+
+set optimizer_switch='exists_to_in=off';
+eval explain extended $q;
+eval $q;
+set optimizer_switch='exists_to_in=on';
+eval explain extended $q;
+eval $q;
+
+set optimizer_switch= @optimizer_switch_save;
+ 
+DROP TABLE t1;
+
 --echo # End of 10.0 tests
 
 #restore defaults