bug#28570: handler::index_read() is called with different find_flag when

ORDER BY is used

The range analysis module did not correctly signal to the 
handler that a range represents a ref (EQ_RANGE flag). This causes 
non-range queries like 
SELECT ... FROM ... WHERE keypart_1=const, ..., keypart_n=const 
ORDER BY ... FOR UPDATE
to wait for a lock unneccesarily if another running transaction uses
SELECT ... FOR UPDATE on the same table.

Fixed by setting EQ_RANGE for all range accesses that represent 
an equality predicate. 


mysql-test/r/innodb_mysql.result:
  bug#28570: Test Result
mysql-test/t/innodb_mysql.test:
  bug#28570: Test Case
sql/handler.cc:
  bug#28570: Updated comment
sql/opt_range.cc:
  bug#28570: Removed the criterion that key has to be unique (HA_NOSAME) in 
  order for the EQ_RANGE flag to be set. It is sufficient that the range 
  represent a ref access.

1 files changed, 29 insertions, 0 deletions

diff --git a/mysql-test/t/innodb_mysql.test b/mysql-test/t/innodb_mysql.test
index 63431e10bbf..e27db9944fe 100644
--- a/mysql-test/t/innodb_mysql.test
+++ b/mysql-test/t/innodb_mysql.test
@@ -840,5 +840,34 @@ DISCONNECT con2;
 DROP PROCEDURE p1;
 DROP TABLE t1;
 
+#
+# Bug #28570: handler::index_read() is called with different find_flag when 
+# ORDER BY is used
+#
+
+CREATE TABLE t1 (
+  a INT,
+  b INT,
+  KEY (b)
+) ENGINE=InnoDB;
+
+INSERT INTO t1 VALUES (1,10), (2,10), (2,20), (3,30);
+
+START TRANSACTION;
+SELECT * FROM t1 WHERE b=20 FOR UPDATE;
+
+--connect (conn2, localhost, root,,test)
+
+# This statement gives a "failed: 1205: Lock wait timeout exceeded; try 
+# restarting transaction" message when the bug is present.
+START TRANSACTION;
+SELECT * FROM t1 WHERE b=10 ORDER BY A FOR UPDATE;
+ROLLBACK;
+
+--disconnect conn2
+--connection default
+
+ROLLBACK;
+DROP TABLE t1;
 
 --echo End of 5.0 tests