a possible bug in row-based replication

[Wei Li]

Hi,

I am looking at the following situation and wonder whether this is a
possible bug or I missed some information.

insert into foo values (...);
insert into foo values (...);

#060804 11:08:29 server id 1  end_log_pos 20229         Table_map:
`test`.`t1` mapped to number 16
BINLOG '
HY3TRBMBAAAAJwAAAAVPAAAAABAAAAAAAAAABHRlc3QAAnQxAAED
';
# at 20415
#060804 11:08:29 server id 1  end_log_pos 20263         Write_rows: table id 16
BINLOG '
HY3TRBQBAAAAIgAAACdPAAAQABAAAAAAAAEAAf/9DgIAAA==
';
# at 20449
#060804 11:08:29 server id 1  end_log_pos 20302         Table_map:
`test`.`t1` mapped to number 16

BINLOG '
HY3TRBMBAAAAJwAAAE5PAAAAABAAAAAAAAAABHRlc3QAAnQxAAED
';
# at 20488
#060804 11:21:17 server id 2  end_log_pos 20537         Rotate to
slave-relay-bin.000004  pos: 4

The problem is that there is a rotation before the last table_map
event and the last Row_log_event.  During the rotation,
flush_relay_log_info() will be called and group_relay_log_name/pos
will be updated unexpectedly.

Thanks,
Wei