Comment 24 for bug 1240098

a)

=== modified file 'Percona-Server/storage/innobase/handler/ha_innodb.cc'
--- Percona-Server/storage/innobase/handler/ha_innodb.cc 2013-10-27 16:53:20 +0000
+++ Percona-Server/storage/innobase/handler/ha_innodb.cc 2013-11-11 15:29:45 +0000
@@ -8303,8 +8303,8 @@
    bool key_appended = false;
  trx_t *trx = thd_to_trx(thd);

- if (table_share && table_share->tmp_table != NO_TMP_TABLE) {
- WSREP_DEBUG("skipping tmp table DML: THD: %lu tmp: %d SQL: %s",
+ if (table_share && table_share->tmp_table != NO_TMP_TABLE && thd->lex->sql_command != SQLCOM_CREATE_TABLE) {
+ WSREP_DEBUG("skipping tmp table DML: THD: %lu tmp: %d SQL: %s Command %d",
        wsrep_thd_thread_id(thd),
        table_share->tmp_table,
        (wsrep_thd_query(thd)) ?

fixes it.

But, the skipping DML clause was specifically added to prevent
memory leak reported in https://bugs.launchpad.net/codership-mysql/+bug/1112514

So, need to check if the regresses here.