Gap Locks or not - mysql

Question

CREATE TABLE `z` (
`a` int(11) NOT NULL,
`b` int(11) DEFAULT NULL,
PRIMARY KEY (`a`),
KEY `b` (`b`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1;
INSERT INTO `test`.`z` (`a`, `b`) VALUES (1, 1);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (3, 1);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (5, 3);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (7, 6);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (10, 8);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (20, 30);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (50, 60);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (45, 90);
INSERT INTO `test`.`z` (`a`, `b`) VALUES (2, 99);
transaction_isolation = REPEATABLE-READ
tx_isolation = REPEATABLE-READ
SESSION 1
begin;
select * from z where b = 60 for update;
SESSION 2
begin;
insert into z select 4, 90; /* why this can't be insert? */
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
insert into z select 46, 90; /* But this can be insert? */
Query OK, 1 row affected (0.00 sec)
Records: 1 Duplicates: 0 Warnings: 0
As the above step:
one can be insert, but the other can't be. they're all the same value - 90.

Answer 1

Yes, this is a gap lock.
The two inserts are not actually trying to insert the same value, because secondary indexes also store a copy of the primary key at the end (that is how an index lookup finds the original row if more columns are referenced than are in the index).
In the index b, the rows in question would be sorted like this.
b a
60, 50
90, 4
90, 45
90, 46
From this illustration, you can see that (4,90) belongs in the gap immediately after (60,50), but (46,90) is not in that gap.