Replicate_Do_DB option causes slave server to commit empty InnoDB transactions.

Hi,

I'm using the Replicate_Do_DB option to replicate a DB with a very small amount of activity (less then one transaction per second). Another DB running on the master server runs many more transactions per second. In the output of 'show full processlist' on the slave, I can see the slave process constantly switching between COMMIT and 'closing tables' for the DB that should be excluded by the Replicate_Do_DB option. I did not have this issue when using MariaDB 5.5.27.

I enabled the general query log and it shows BEGIN and COMMIT repeatedly but no queries that actually alter any tables:

140527 12:13:35 1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
140527 12:13:36 1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT /* implicit, from Xid_log_event */
1721 Query BEGIN
1721 Query COMMIT /* implicit, from Xid_log_event */
1721 Query BEGIN
1721 Query COMMIT
1721 Query BEGIN
1721 Query COMMIT

This generates a lot of I/O contention that I've had to mitigate on this development server by setting innodb_flush_log_at_trx_commit to 0. I wouldn't be comfortable with that setting on a production DB server though.

Regards,
Bill