Avoiding "Lock wait timeout exceeded"

[Richard]

Hello all,
I am having a locking issue. Here is my environment:
OS: Solaris 10 SPARC
MySQL: 5.1.22-rc running as a slave using ROW based replication.

Table information:
`scratch`.`scratch_table` = MyISAM
`prod`.`table1` = InnoDB

Here are the queries I run. Changed for simplicity.

set SESSION transaction isolation level read uncommitted;

UPDATE `scratch`.`scratch_table` a INNER JOIN `prod`.`table1` b ON a.`value1` = b.`value1` AND a.`value2` = b.`value2`
SET a.value3 = b.value3 WHERE a.value4 = 'X' and b.value4 = 'Y';

At the time of the locks it is highly likely that `prod`.`table1` was being updated through replication and the rows I was joining on were being updated or deleted. Nothing would be running queries against `scratch`.`scratch_table`. I do not care about dirty reads. Any data is good, but a timeout is bad. I thought by setting the transaction isolation I would be fine but either I am wrong or I am doing something incorrectly. I also tried setting innodb_locks_unsafe_for_binlog=1 but that caused the DB to crash. Any thoughts on what I can do to avoid the lock and the timeout it causes?

Thanks,
Richard