Fix for bug#38269: pushbuild gives valgrind error in

ha_statistic_increment for rpl_temporary

Problem: in some cases master send a special event to reconnecting
slave to keep slave's temporary tables (see #17284) and they still 
have references to the "old" SQL slave thread and use them to access
thread's data.

Fix: set temporary tables thread references to the actual SQL slave
thread in such cases.


mysql-test/suite/rpl/t/disabled.def:
  Fix for bug#38269: pushbuild gives valgrind error in 
  ha_statistic_increment for rpl_temporary
    - rpl_temporary test enabled.
sql/log_event.cc:
  Fix for bug#38269: pushbuild gives valgrind error in 
  ha_statistic_increment for rpl_temporary
    - if we keep slave's temporary tables after reconnect,
  set their thread pointers (->in_use) to the current SQL thread.

mysql-test/suite/rpl/t/disabled.def

@@ -12,5 +12,4 @@
 
 rpl_redirect               : Failure is sporadic and and the test is superfluous (mats)
 rpl_innodb_bug28430        : Failure on Solaris Bug #36793
-rpl_temporary              : BUG#38269 2008-07-21 Sven valgrind error in pushbuild
 rpl_flushlog_loop          : BUG#37733 2008-07-23 Sven disabled in 5.1-bugteam. the bug has been fixed in 5.1-rpl: please re-enable when that gets pushed to main

sql/log_event.cc

@@ -3350,6 +3350,17 @@ int Start_log_event_v3::do_apply_event(Relay_log_info const *rli)
       close_temporary_tables(thd);
       cleanup_load_tmpdir();
     }
+    else
+    {
+      /*
+        Set all temporary tables thread references to the current thread
+        as they may point to the "old" SQL slave thread in case of its
+        restart.
+      */
+      TABLE *table;
+      for (table= thd->temporary_tables; table; table= table->next)
+        table->in_use= thd;
+    }
     break;
 
     /*