Save and clear run context before executing a stored function or trigger and restore it afterwards.

This allows us to use statement replication with functions and triggers
The following things are fixed with this patch:
- NOW() and automatic timestamps takes the value from the main event for functions and triggers (which allows these to replicate with statement level logging)
- No side effects for triggers or functions with auto-increment values(), last_insert_id(), rand() or found_rows()
- Triggers can't return result sets

Fixes bugs:
#12480: NOW() is not constant in a trigger
#12481: Using NOW() in a stored function breaks statement based replication
#12482: Triggers has side effects with auto_increment values
#11587: trigger causes lost connection error


mysql-test/r/trigger.result:
  Added test fpr big
mysql-test/t/sp-error.test:
  Changed error message numbers
mysql-test/t/trigger.test:
  Added test for trigger returning result (#11587)
sql/item_func.cc:
  Store the first used seed value for RAND() value.
  (This makes rand() replicatable in functions and triggers)
  Save and clear run context before executing a stored function and restore it afterwards.
  This removes side effects of stored functions for RAND(), auto-increment values and NOW() and makes most stored function replicatable
sql/share/errmsg.txt:
  Reuse error message also for triggers
sql/sp_head.cc:
  If in function or trigger, don't change value of NOW()
  (This allows us to use statement replication with functions that directly or indirectly uses timestamps)
sql/sql_class.cc:
  Added framework for storing and retrieving run context while exceuting triggers or stored functions.
sql/sql_class.h:
  Added framework for storing and retrieving run context while exceuting triggers or stored functions.
sql/sql_parse.cc:
  If in function or trigger, don't change value of NOW()
  (This allows us to use statement replication with functions that directly or indirectly uses timestamps)
sql/sql_trigger.cc:
  Moved process_triggers function from sql_trigger.h
  Use reset/restore sub_statement_state while executing triggers to avoid side effects and make them replicatable
sql/sql_trigger.h:
  Moved process_triggers function from sql_trigger.h
  Use reset/restore sub_statement_state while executing triggers to avoid side effects and make them replicatable
sql/sql_yacc.yy:
  Give error message if trigger can return a result set (Bug #11587)
tests/fork_big2.pl:
  Removed return from end of lines
mysql-test/r/rpl_trigger.result:
  New BitKeeper file ``mysql-test/r/rpl_trigger.result''
mysql-test/t/rpl_trigger.test:
  New BitKeeper file ``mysql-test/t/rpl_trigger.test''

1 files changed, 2 insertions, 49 deletions

diff --git a/sql/sql_trigger.h b/sql/sql_trigger.h
index e2ed4bcc0f4..d9b39cc3034 100644
--- a/sql/sql_trigger.h
+++ b/sql/sql_trigger.h
@@ -20,6 +20,7 @@
 
   QQ: Will it be merged into TABLE in future ?
 */
+
 class Table_triggers_list: public Sql_alloc
 {
   /* Triggers as SPs grouped by event, action_time */
@@ -76,55 +77,7 @@ public:
   bool drop_trigger(THD *thd, TABLE_LIST *table);
   bool process_triggers(THD *thd, trg_event_type event,
                         trg_action_time_type time_type,
-                        bool old_row_is_record1)
-  {
-    int res= 0;
-
-    if (bodies[event][time_type])
-    {
-      bool save_in_sub_stmt= thd->in_sub_stmt;
-#ifndef EMBEDDED_LIBRARY
-      /* Surpress OK packets in case if we will execute statements */
-      my_bool nsok= thd->net.no_send_ok;
-      thd->net.no_send_ok= TRUE;
-#endif
-
-      if (old_row_is_record1)
-      {
-        old_field= record1_field;
-        new_field= table->field;
-      }
-      else
-      {
-        new_field= record1_field;
-        old_field= table->field;
-      }
-
-      /*
-        FIXME: We should juggle with security context here (because trigger
-        should be invoked with creator rights).
-      */
-      /*
-	We disable binlogging, as in SP/functions, even though currently
-        triggers can't do updates. When triggers can do updates, someone
-        should add such a trigger to rpl_sp.test to verify that the update
-        does NOT go into binlog.
-      */
-      tmp_disable_binlog(thd);
-      thd->in_sub_stmt= TRUE;
-
-      res= bodies[event][time_type]->execute_function(thd, 0, 0, 0);
-
-      thd->in_sub_stmt= save_in_sub_stmt;
-      reenable_binlog(thd);
-
-#ifndef EMBEDDED_LIBRARY
-      thd->net.no_send_ok= nsok;
-#endif
-    }
-
-    return res;
-  }
+                        bool old_row_is_record1);
   bool get_trigger_info(THD *thd, trg_event_type event,
                         trg_action_time_type time_type,
                         LEX_STRING *trigger_name, LEX_STRING *trigger_stmt,