Bug #26418: Slave out of sync after

 CREATE/DROP TEMPORARY TABLE + ROLLBACK on master

The transaction ability of the storage engines of
the tables on the replication master and the replication
slave must generally be the same.
When the storage engine type of the slave is 
non-transactional then transactions on the master that 
mix update of transactional and non-transactional tables
should be avoided because they will cause inconsistency of
the data between the master's transactional table and the
slave's non-transactional table.

The effect described by this bug is actually expected.
A detailed test case is added (to be merged later to
the updated rpl_ddl.test), as there was no coverage 
by the existing tests. 
Some code cleanup is also added by this change.

1 files changed, 1 insertions, 8 deletions

diff --git a/sql/sql_trigger.cc b/sql/sql_trigger.cc
index e15003ab243..6e8a40e68a2 100644
--- a/sql/sql_trigger.cc
+++ b/sql/sql_trigger.cc
@@ -306,14 +306,7 @@ end:
 
   if (!result)
   {
-    if (mysql_bin_log.is_open())
-    {
-      thd->clear_error();
-
-      /* Such a statement can always go directly to binlog, no trans cache. */
-      thd->binlog_query(THD::STMT_QUERY_TYPE,
-                        stmt_query.ptr(), stmt_query.length(), FALSE, FALSE);
-    }
+    write_bin_log(thd, TRUE, stmt_query.ptr(), stmt_query.length());
   }
 
   VOID(pthread_mutex_unlock(&LOCK_open));