From d4450e66964e14ac85bc40c567a510dd86353f60 Mon Sep 17 00:00:00 2001
From: "dlenev@mysql.com" <>
Date: Sun, 2 Jul 2006 01:51:10 +0400
Subject: [PATCH 1/2] Fix for bug#18437 "Wrong values inserted with a before
 update trigger on NDB table".

SQL-layer was not marking fields which were used in triggers as such. As
result these fields were not always properly retrieved/stored by handler
layer. So one might got wrong values or lost changes in triggers for NDB,
Federated and possibly InnoDB tables.
This fix solves the problem by marking fields used in triggers
appropriately.

Also this patch contains the following cleanup of ha_ndbcluster code:

We no longer rely on reading LEX::sql_command value in handler in order
to determine if we can enable optimization which allows us to handle REPLACE
statement in more efficient way by doing replaces directly in write_row()
method without reporting error to SQL-layer.
Instead we rely on SQL-layer informing us whether this optimization
applicable by calling handler::extra() method with
HA_EXTRA_WRITE_CAN_REPLACE flag.
As result we no longer apply this optimzation in cases when it should not
be used (e.g. if we have on delete triggers on table) and use in some
additional cases when it is applicable (e.g. for LOAD DATA REPLACE).

Finally this patch includes fix for bug#20728 "REPLACE does not work
correctly for NDB table with PK and unique index".

This was yet another problem which was caused by improper field mark-up.
During row replacement fields which weren't explicity used in REPLACE
statement were not marked as fields to be saved (updated) so they have
retained values from old row version. The fix is to mark all table
fields as set for REPLACE statement. Note that in 5.1 we already solve
this problem by notifying handler that it should save values from all
fields only in case when real replacement happens.
---
 include/my_base.h               |  11 ++-
 mysql-test/r/federated.result   |  28 ++++++++
 mysql-test/r/ndb_replace.result |  47 ++++++++++++-
 mysql-test/r/ndb_trigger.result | 119 ++++++++++++++++++++++++++++++++
 mysql-test/t/federated.test     |  42 +++++++++++
 mysql-test/t/ndb_replace.test   |  37 ++++++++++
 mysql-test/t/ndb_trigger.test   |  92 ++++++++++++++++++++++++
 sql/ha_ndbcluster.cc            |  27 ++++----
 sql/item.cc                     |   9 ++-
 sql/mysql_priv.h                |   6 +-
 sql/sql_delete.cc               |   6 ++
 sql/sql_insert.cc               |  77 ++++++++++++++++++++-
 sql/sql_load.cc                 |  10 +++
 sql/sql_parse.cc                |   7 +-
 sql/sql_table.cc                |  19 ++---
 sql/sql_trigger.cc              |  44 +++++++++++-
 sql/sql_trigger.h               |   8 +++
 sql/sql_update.cc               |   6 ++
 18 files changed, 556 insertions(+), 39 deletions(-)
 create mode 100644 mysql-test/r/ndb_trigger.result
 create mode 100644 mysql-test/t/ndb_trigger.test

diff --git a/include/my_base.h b/include/my_base.h
index 076eed9998f..05ba38e77eb 100644
--- a/include/my_base.h
+++ b/include/my_base.h
@@ -152,7 +152,16 @@ enum ha_extra_function {
     other fields intact. When this is off (by default) InnoDB will use memcpy
     to overwrite entire row.
   */
-  HA_EXTRA_KEYREAD_PRESERVE_FIELDS
+  HA_EXTRA_KEYREAD_PRESERVE_FIELDS,
+  /*
+    Informs handler that write_row() which tries to insert new row into the
+    table and encounters some already existing row with same primary/unique
+    key can replace old row with new row instead of reporting error (basically
+    it informs handler that we do REPLACE instead of simple INSERT).
+    Off by default.
+  */
+  HA_EXTRA_WRITE_CAN_REPLACE,
+  HA_EXTRA_WRITE_CANNOT_REPLACE
 };
 
 	/* The following is parameter to ha_panic() */
diff --git a/mysql-test/r/federated.result b/mysql-test/r/federated.result
index f11da4ee62f..07396d17ed5 100644
--- a/mysql-test/r/federated.result
+++ b/mysql-test/r/federated.result
@@ -1601,6 +1601,34 @@ fld_cid	fld_name	fld_parentid	fld_delt
 5	Torkel	0	0
 DROP TABLE federated.t1;
 DROP TABLE federated.bug_17377_table;
+drop table if exists federated.t1;
+create table federated.t1 (a int, b int, c int);
+drop table if exists federated.t1;
+drop table if exists federated.t2;
+create table federated.t1 (a int,  b int, c int) engine=federated connection='mysql://root@127.0.0.1:SLAVE_PORT/federated/t1';
+create trigger federated.t1_bi before insert on federated.t1 for each row set new.c= new.a * new.b;
+create table federated.t2 (a int, b int);
+insert into federated.t2 values (13, 17), (19, 23);
+insert into federated.t1 (a, b) values (1, 2), (3, 5), (7, 11);
+select * from federated.t1;
+a	b	c
+1	2	2
+3	5	15
+7	11	77
+delete from federated.t1;
+insert into federated.t1 (a, b) select * from federated.t2;
+select * from federated.t1;
+a	b	c
+13	17	221
+19	23	437
+delete from federated.t1;
+load data infile '../std_data_ln/loaddata5.dat' into table federated.t1 fields terminated by '' enclosed by '' ignore 1 lines (a, b);
+select * from federated.t1;
+a	b	c
+3	4	12
+5	6	30
+drop tables federated.t1, federated.t2;
+drop table federated.t1;
 DROP TABLE IF EXISTS federated.t1;
 DROP DATABASE IF EXISTS federated;
 DROP TABLE IF EXISTS federated.t1;
diff --git a/mysql-test/r/ndb_replace.result b/mysql-test/r/ndb_replace.result
index cdfcd6a7a43..4d63c397d60 100644
--- a/mysql-test/r/ndb_replace.result
+++ b/mysql-test/r/ndb_replace.result
@@ -30,7 +30,8 @@ REPLACE INTO t1 (i,j) VALUES (17,2);
 SELECT * from t1 ORDER BY i;
 i	j	k
 3	1	42
-17	2	24
+17	2	NULL
+DROP TABLE t1;
 CREATE TABLE t2 (a INT(11) NOT NULL,
 b INT(11) NOT NULL,
 c INT(11) NOT NULL,
@@ -52,3 +53,47 @@ SELECT * FROM t2 ORDER BY id;
 a	b	c	x	y	z	id	i
 1	1	1	b	b	b	5	2
 DROP TABLE t2;
+drop table if exists t1;
+create table t1 (pk int primary key, apk int unique, data int) engine=ndbcluster;
+insert into t1 values (1, 1, 1), (2, 2, 2), (3, 3, 3);
+replace into t1 (pk, apk) values (4, 1), (5, 2);
+select * from t1 order by pk;
+pk	apk	data
+3	3	3
+4	1	NULL
+5	2	NULL
+delete from t1;
+insert into t1 values (1, 1, 1), (2, 2, 2), (3, 3, 3);
+replace into t1 (pk, apk) values (1, 4), (2, 5);
+select * from t1 order by pk;
+pk	apk	data
+1	4	NULL
+2	5	NULL
+3	3	3
+delete from t1;
+insert into t1 values (1, 1, 1), (4, 4, 4), (6, 6, 6);
+load data infile '../std_data_ln/loaddata5.dat' replace into table t1 fields terminated by '' enclosed by '' ignore 1 lines (pk, apk);
+select * from t1 order by pk;
+pk	apk	data
+1	1	1
+3	4	NULL
+5	6	NULL
+delete from t1;
+insert into t1 values (1, 1, 1), (3, 3, 3), (5, 5, 5);
+load data infile '../std_data_ln/loaddata5.dat' replace into table t1 fields terminated by '' enclosed by '' ignore 1 lines (pk, apk);
+select * from t1 order by pk;
+pk	apk	data
+1	1	1
+3	4	NULL
+5	6	NULL
+delete from t1;
+insert into t1 values (1, 1, 1), (2, 2, 2), (3, 3, 3);
+replace into t1 (pk, apk) select 4, 1;
+replace into t1 (pk, apk) select 2, 4;
+select * from t1 order by pk;
+pk	apk	data
+2	4	NULL
+3	3	3
+4	1	NULL
+drop table t1;
+End of 5.0 tests.
diff --git a/mysql-test/r/ndb_trigger.result b/mysql-test/r/ndb_trigger.result
new file mode 100644
index 00000000000..27f83df70c9
--- /dev/null
+++ b/mysql-test/r/ndb_trigger.result
@@ -0,0 +1,119 @@
+drop table if exists t1, t2, t3;
+create table t1 (id int primary key, a int not null, b decimal (63,30) default 0) engine=ndb;
+create table t2 (op char(1), a int not null, b decimal (63,30));
+create table t3 select 1 as i;
+create trigger t1_bu before update on t1 for each row
+begin
+insert into t2 values ("u", old.a, old.b);
+set new.b = old.b + 10;
+end;//
+create trigger t1_bd before delete on t1 for each row
+begin
+insert into t2 values ("d", old.a, old.b);
+end;//
+insert into t1 values (1, 1, 1.05), (2, 2, 2.05), (3, 3, 3.05), (4, 4, 4.05);
+update t1 set a=5 where a != 3;
+select * from t1 order by id;
+id	a	b
+1	5	11.050000000000000000000000000000
+2	5	12.050000000000000000000000000000
+3	3	3.050000000000000000000000000000
+4	5	14.050000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+u	1	1.050000000000000000000000000000
+u	2	2.050000000000000000000000000000
+u	4	4.050000000000000000000000000000
+delete from t2;
+update t1, t3 set a=6 where a = 5;
+select * from t1 order by id;
+id	a	b
+1	6	21.050000000000000000000000000000
+2	6	22.050000000000000000000000000000
+3	3	3.050000000000000000000000000000
+4	6	24.050000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+u	5	11.050000000000000000000000000000
+u	5	12.050000000000000000000000000000
+u	5	14.050000000000000000000000000000
+delete from t2;
+delete from t1 where a != 3;
+select * from t1 order by id;
+id	a	b
+3	3	3.050000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+d	6	21.050000000000000000000000000000
+d	6	22.050000000000000000000000000000
+d	6	24.050000000000000000000000000000
+delete from t2;
+insert into t1 values (1, 1, 1.05), (2, 2, 2.05), (4, 4, 4.05);
+delete t1 from t1, t3 where a != 3;
+select * from t1 order by id;
+id	a	b
+3	3	3.050000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+d	1	1.050000000000000000000000000000
+d	2	2.050000000000000000000000000000
+d	4	4.050000000000000000000000000000
+delete from t2;
+insert into t1 values (4, 4, 4.05);
+insert into t1 (id, a) values (4, 1), (3, 1) on duplicate key update a= a + 1;
+select * from t1 order by id;
+id	a	b
+3	4	13.050000000000000000000000000000
+4	5	14.050000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+u	3	3.050000000000000000000000000000
+u	4	4.050000000000000000000000000000
+delete from t2;
+delete from t3;
+insert into t3 values (4), (3);
+insert into t1 (id, a) (select i, 1 from t3) on duplicate key update a= a + 1;
+select * from t1 order by id;
+id	a	b
+3	5	23.050000000000000000000000000000
+4	6	24.050000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+u	4	13.050000000000000000000000000000
+u	5	14.050000000000000000000000000000
+delete from t2;
+replace into t1 (id, a) values (4, 1), (3, 1);
+select * from t1 order by id;
+id	a	b
+3	1	0.000000000000000000000000000000
+4	1	0.000000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+d	5	23.050000000000000000000000000000
+d	6	24.050000000000000000000000000000
+delete from t1;
+delete from t2;
+insert into t1 values (3, 1, 1.05), (4, 1, 2.05);
+replace into t1 (id, a) (select i, 2 from t3);
+select * from t1 order by id;
+id	a	b
+3	2	0.000000000000000000000000000000
+4	2	0.000000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+d	1	1.050000000000000000000000000000
+d	1	2.050000000000000000000000000000
+delete from t1;
+delete from t2;
+insert into t1 values (3, 1, 1.05), (5, 2, 2.05);
+load data infile '../std_data_ln/loaddata5.dat' replace into table t1 fields terminated by '' enclosed by '' ignore 1 lines (id, a);
+select * from t1 order by id;
+id	a	b
+3	4	0.000000000000000000000000000000
+5	6	0.000000000000000000000000000000
+select * from t2 order by op, a, b;
+op	a	b
+d	1	1.050000000000000000000000000000
+d	2	2.050000000000000000000000000000
+drop tables t1, t2, t3;
+End of 5.0 tests
diff --git a/mysql-test/t/federated.test b/mysql-test/t/federated.test
index 80b31c610a2..6e53b1e3adb 100644
--- a/mysql-test/t/federated.test
+++ b/mysql-test/t/federated.test
@@ -1310,4 +1310,46 @@ connection slave;
 DROP TABLE federated.bug_17377_table;
 
 
+# 
+# Additional test for bug#18437 "Wrong values inserted with a before
+# update trigger on NDB table". SQL-layer didn't properly inform
+# handler about fields which were read and set in triggers. In some
+# cases this resulted in incorrect (garbage) values of OLD variables
+# and lost changes to NEW variables.
+# Since for federated engine only operation which is affected by wrong
+# fields mark-up is handler::write_row() this file constains coverage
+# for ON INSERT triggers only. Tests for other types of triggers reside
+# in ndb_trigger.test.
+#
+--disable_warnings
+drop table if exists federated.t1;
+--enable_warnings
+create table federated.t1 (a int, b int, c int);
+connection master;
+--disable_warnings
+drop table if exists federated.t1;
+drop table if exists federated.t2;
+--enable_warnings
+--replace_result $SLAVE_MYPORT SLAVE_PORT
+eval create table federated.t1 (a int,  b int, c int) engine=federated connection='mysql://root@127.0.0.1:$SLAVE_MYPORT/federated/t1';
+create trigger federated.t1_bi before insert on federated.t1 for each row set new.c= new.a * new.b;
+create table federated.t2 (a int, b int);
+insert into federated.t2 values (13, 17), (19, 23);
+# Each of three statements should correctly set values for all three fields
+# insert
+insert into federated.t1 (a, b) values (1, 2), (3, 5), (7, 11);
+select * from federated.t1;
+delete from federated.t1;
+# insert ... select
+insert into federated.t1 (a, b) select * from federated.t2;
+select * from federated.t1;
+delete from federated.t1;
+# load
+load data infile '../std_data_ln/loaddata5.dat' into table federated.t1 fields terminated by '' enclosed by '' ignore 1 lines (a, b);
+select * from federated.t1;
+drop tables federated.t1, federated.t2;
+
+connection slave;
+drop table federated.t1;
+
 source include/federated_cleanup.inc;
diff --git a/mysql-test/t/ndb_replace.test b/mysql-test/t/ndb_replace.test
index 94a11f7dfb2..476a607ed44 100644
--- a/mysql-test/t/ndb_replace.test
+++ b/mysql-test/t/ndb_replace.test
@@ -39,6 +39,7 @@ INSERT  INTO t1 VALUES (1,1,23),(2,2,24);
 REPLACE INTO t1 (j,k) VALUES (1,42);
 REPLACE INTO t1 (i,j) VALUES (17,2);
 SELECT * from t1 ORDER BY i;
+DROP TABLE t1;
 
 # bug#19906
 CREATE TABLE t2 (a INT(11) NOT NULL,
@@ -64,4 +65,40 @@ SELECT * FROM t2 ORDER BY id;
 
 DROP TABLE t2;
 
+#
+# Bug #20728 "REPLACE does not work correctly for NDB table with PK and
+#             unique index"
+#
+--disable_warnings
+drop table if exists t1;
+--enable_warnings
+create table t1 (pk int primary key, apk int unique, data int) engine=ndbcluster;
+# Test for plain replace which updates pk
+insert into t1 values (1, 1, 1), (2, 2, 2), (3, 3, 3);
+replace into t1 (pk, apk) values (4, 1), (5, 2);
+select * from t1 order by pk;
+delete from t1;
+# Another test for plain replace which doesn't touch pk
+insert into t1 values (1, 1, 1), (2, 2, 2), (3, 3, 3);
+replace into t1 (pk, apk) values (1, 4), (2, 5);
+select * from t1 order by pk;
+delete from t1;
+# Test for load data replace which updates pk
+insert into t1 values (1, 1, 1), (4, 4, 4), (6, 6, 6);
+load data infile '../std_data_ln/loaddata5.dat' replace into table t1 fields terminated by '' enclosed by '' ignore 1 lines (pk, apk);
+select * from t1 order by pk;
+delete from t1;
+# Now test for load data replace which doesn't touch pk
+insert into t1 values (1, 1, 1), (3, 3, 3), (5, 5, 5);
+load data infile '../std_data_ln/loaddata5.dat' replace into table t1 fields terminated by '' enclosed by '' ignore 1 lines (pk, apk);
+select * from t1 order by pk;
+delete from t1;
+# Finally test for both types of replace ... select
+insert into t1 values (1, 1, 1), (2, 2, 2), (3, 3, 3);
+replace into t1 (pk, apk) select 4, 1;
+replace into t1 (pk, apk) select 2, 4;
+select * from t1 order by pk;
+# Clean-up
+drop table t1;
 
+--echo End of 5.0 tests.
diff --git a/mysql-test/t/ndb_trigger.test b/mysql-test/t/ndb_trigger.test
new file mode 100644
index 00000000000..2521ef17842
--- /dev/null
+++ b/mysql-test/t/ndb_trigger.test
@@ -0,0 +1,92 @@
+# Tests which involve triggers and NDB storage engine
+--source include/have_ndb.inc
+--source include/not_embedded.inc
+
+# 
+# Test for bug#18437 "Wrong values inserted with a before update
+# trigger on NDB table". SQL-layer didn't properly inform handler
+# about fields which were read and set in triggers. In some cases 
+# this resulted in incorrect (garbage) values of OLD variables and
+# lost changes to NEW variables.
+# You can find similar tests for ON INSERT triggers in federated.test
+# since this engine so far is the only engine in MySQL which cares
+# about field mark-up during handler::write_row() operation.
+#
+
+--disable_warnings
+drop table if exists t1, t2, t3;
+--enable_warnings
+
+create table t1 (id int primary key, a int not null, b decimal (63,30) default 0) engine=ndb;
+create table t2 (op char(1), a int not null, b decimal (63,30));
+create table t3 select 1 as i;
+	
+delimiter //;
+create trigger t1_bu before update on t1 for each row
+begin
+  insert into t2 values ("u", old.a, old.b);
+  set new.b = old.b + 10;
+end;//
+create trigger t1_bd before delete on t1 for each row
+begin
+  insert into t2 values ("d", old.a, old.b);
+end;//
+delimiter ;//
+insert into t1 values (1, 1, 1.05), (2, 2, 2.05), (3, 3, 3.05), (4, 4, 4.05);
+
+# Check that usual update works as it should
+update t1 set a=5 where a != 3;
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t2;
+# Check that everything works for multi-update
+update t1, t3 set a=6 where a = 5;
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t2;
+# Check for delete
+delete from t1 where a != 3;
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t2;
+# Check for multi-delete
+insert into t1 values (1, 1, 1.05), (2, 2, 2.05), (4, 4, 4.05);
+delete t1 from t1, t3 where a != 3;
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t2;
+# Check for insert ... on duplicate key update
+insert into t1 values (4, 4, 4.05);
+insert into t1 (id, a) values (4, 1), (3, 1) on duplicate key update a= a + 1;
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t2;
+# Check for insert ... select ... on duplicate key update
+delete from t3;
+insert into t3 values (4), (3);
+insert into t1 (id, a) (select i, 1 from t3) on duplicate key update a= a + 1;
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t2;
+# Check for replace
+replace into t1 (id, a) values (4, 1), (3, 1);
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t1;
+delete from t2;
+# Check for replace ... select ...
+insert into t1 values (3, 1, 1.05), (4, 1, 2.05);
+replace into t1 (id, a) (select i, 2 from t3);
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+delete from t1;
+delete from t2;
+# Check for load data replace
+insert into t1 values (3, 1, 1.05), (5, 2, 2.05);
+load data infile '../std_data_ln/loaddata5.dat' replace into table t1 fields terminated by '' enclosed by '' ignore 1 lines (id, a);
+select * from t1 order by id;
+select * from t2 order by op, a, b;
+
+drop tables t1, t2, t3;
+
+--echo End of 5.0 tests
diff --git a/sql/ha_ndbcluster.cc b/sql/ha_ndbcluster.cc
index 46ab5b88624..074e8fb9371 100644
--- a/sql/ha_ndbcluster.cc
+++ b/sql/ha_ndbcluster.cc
@@ -3212,20 +3212,11 @@ int ha_ndbcluster::extra(enum ha_extra_function operation)
     break;
   case HA_EXTRA_IGNORE_DUP_KEY:       /* Dup keys don't rollback everything*/
     DBUG_PRINT("info", ("HA_EXTRA_IGNORE_DUP_KEY"));
-    if (current_thd->lex->sql_command == SQLCOM_REPLACE && !m_has_unique_index)
-    {
-      DBUG_PRINT("info", ("Turning ON use of write instead of insert"));
-      m_use_write= TRUE;
-    } else 
-    {
-      DBUG_PRINT("info", ("Ignoring duplicate key"));
-      m_ignore_dup_key= TRUE;
-    }
+    DBUG_PRINT("info", ("Ignoring duplicate key"));
+    m_ignore_dup_key= TRUE;
     break;
   case HA_EXTRA_NO_IGNORE_DUP_KEY:
     DBUG_PRINT("info", ("HA_EXTRA_NO_IGNORE_DUP_KEY"));
-    DBUG_PRINT("info", ("Turning OFF use of write instead of insert"));
-    m_use_write= FALSE;
     m_ignore_dup_key= FALSE;
     break;
   case HA_EXTRA_RETRIEVE_ALL_COLS:    /* Retrieve all columns, not just those
@@ -3255,7 +3246,19 @@ int ha_ndbcluster::extra(enum ha_extra_function operation)
   case HA_EXTRA_KEYREAD_PRESERVE_FIELDS:
     DBUG_PRINT("info", ("HA_EXTRA_KEYREAD_PRESERVE_FIELDS"));
     break;
-
+  case HA_EXTRA_WRITE_CAN_REPLACE:
+    DBUG_PRINT("info", ("HA_EXTRA_WRITE_CAN_REPLACE"));
+    if (!m_has_unique_index)
+    {
+      DBUG_PRINT("info", ("Turning ON use of write instead of insert"));
+      m_use_write= TRUE;
+    }
+    break;
+  case HA_EXTRA_WRITE_CANNOT_REPLACE:
+    DBUG_PRINT("info", ("HA_EXTRA_WRITE_CANNOT_REPLACE"));
+    DBUG_PRINT("info", ("Turning OFF use of write instead of insert"));
+    m_use_write= FALSE;
+    break;
   }
   
   DBUG_RETURN(0);
diff --git a/sql/item.cc b/sql/item.cc
index 24efc1f106f..6e26c204c0b 100644
--- a/sql/item.cc
+++ b/sql/item.cc
@@ -5350,9 +5350,14 @@ void Item_insert_value::print(String *str)
 void Item_trigger_field::setup_field(THD *thd, TABLE *table,
                                      GRANT_INFO *table_grant_info)
 {
+  /*
+    There is no sense in marking fields used by trigger with current value
+    of THD::query_id since it is completely unrelated to the THD::query_id
+    value for statements which will invoke trigger. So instead we use
+    Table_triggers_list::mark_fields_used() method which is called during
+    execution of these statements.
+  */
   bool save_set_query_id= thd->set_query_id;
-
-  /* TODO: Think more about consequences of this step. */
   thd->set_query_id= 0;
   /*
     Try to find field by its name and if it will be found
diff --git a/sql/mysql_priv.h b/sql/mysql_priv.h
index 3bb371b6004..798ca3b8967 100644
--- a/sql/mysql_priv.h
+++ b/sql/mysql_priv.h
@@ -726,9 +726,7 @@ bool mysql_alter_table(THD *thd, char *new_db, char *new_name,
                        TABLE_LIST *table_list,
                        List<create_field> &fields,
                        List<Key> &keys,
-                       uint order_num, ORDER *order,
-                       enum enum_duplicates handle_duplicates,
-                       bool ignore,
+                       uint order_num, ORDER *order, bool ignore,
                        ALTER_INFO *alter_info, bool do_send_ok);
 bool mysql_recreate_table(THD *thd, TABLE_LIST *table_list, bool do_send_ok);
 bool mysql_create_like_table(THD *thd, TABLE_LIST *table,
@@ -764,6 +762,8 @@ bool mysql_insert(THD *thd,TABLE_LIST *table,List<Item> &fields,
                   bool ignore);
 int check_that_all_fields_are_given_values(THD *thd, TABLE *entry,
                                            TABLE_LIST *table_list);
+void mark_fields_used_by_triggers_for_insert_stmt(THD *thd, TABLE *table,
+                                                  enum_duplicates duplic);
 bool mysql_prepare_delete(THD *thd, TABLE_LIST *table_list, Item **conds);
 bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds,
                   SQL_LIST *order, ha_rows rows, ulonglong options,
diff --git a/sql/sql_delete.cc b/sql/sql_delete.cc
index af20b770c56..98bd3ee5a36 100644
--- a/sql/sql_delete.cc
+++ b/sql/sql_delete.cc
@@ -194,6 +194,10 @@ bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds,
   deleted=0L;
   init_ftfuncs(thd, select_lex, 1);
   thd->proc_info="updating";
+
+  if (table->triggers)
+    table->triggers->mark_fields_used(thd, TRG_EVENT_DELETE);
+
   while (!(error=info.read_record(&info)) && !thd->killed &&
 	 !thd->net.report_error)
   {
@@ -507,6 +511,8 @@ multi_delete::initialize_tables(JOIN *join)
 	transactional_tables= 1;
       else
 	normal_tables= 1;
+      if (tbl->triggers)
+        tbl->triggers->mark_fields_used(thd, TRG_EVENT_DELETE);
     }
     else if ((tab->type != JT_SYSTEM && tab->type != JT_CONST) &&
              walk == delete_tables)
diff --git a/sql/sql_insert.cc b/sql/sql_insert.cc
index 8ffc6f53a43..c5ff8061c94 100644
--- a/sql/sql_insert.cc
+++ b/sql/sql_insert.cc
@@ -241,6 +241,33 @@ static int check_update_fields(THD *thd, TABLE_LIST *insert_table_list,
 }
 
 
+/*
+  Mark fields used by triggers for INSERT-like statement.
+
+  SYNOPSIS
+    mark_fields_used_by_triggers_for_insert_stmt()
+      thd     The current thread
+      table   Table to which insert will happen
+      duplic  Type of duplicate handling for insert which will happen
+
+  NOTE
+    For REPLACE there is no sense in marking particular fields
+    used by ON DELETE trigger as to execute it properly we have
+    to retrieve and store values for all table columns anyway.
+*/
+
+void mark_fields_used_by_triggers_for_insert_stmt(THD *thd, TABLE *table,
+                                                  enum_duplicates duplic)
+{
+  if (table->triggers)
+  {
+    table->triggers->mark_fields_used(thd, TRG_EVENT_INSERT);
+    if (duplic == DUP_UPDATE)
+      table->triggers->mark_fields_used(thd, TRG_EVENT_UPDATE);
+  }
+}
+
+
 bool mysql_insert(THD *thd,TABLE_LIST *table_list,
                   List<Item> &fields,
                   List<List_item> &values_list,
@@ -401,6 +428,17 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list,
   thd->proc_info="update";
   if (duplic != DUP_ERROR || ignore)
     table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
+  if (duplic == DUP_REPLACE)
+  {
+    if (!table->triggers || !table->triggers->has_delete_triggers())
+      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
+    /*
+      REPLACE should change values of all columns so we should mark
+      all columns as columns to be set. As nice side effect we will
+      retrieve columns which values are needed for ON DELETE triggers.
+    */
+    table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
+  }
   /*
     let's *try* to start bulk inserts. It won't necessary
     start them as values_list.elements should be greater than
@@ -429,6 +467,8 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list,
     error= 1;
   }
 
+  mark_fields_used_by_triggers_for_insert_stmt(thd, table, duplic);
+
   if (table_list->prepare_where(thd, 0, TRUE) ||
       table_list->prepare_check_option(thd))
     error= 1;
@@ -599,6 +639,9 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list,
   thd->next_insert_id=0;			// Reset this if wrongly used
   if (duplic != DUP_ERROR || ignore)
     table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY);
+  if (duplic == DUP_REPLACE &&
+      (!table->triggers || !table->triggers->has_delete_triggers()))
+    table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE);
 
   /* Reset value of LAST_INSERT_ID if no rows where inserted */
   if (!info.copied && thd->insert_id_used)
@@ -1902,7 +1945,8 @@ bool delayed_insert::handle_inserts(void)
 {
   int error;
   ulong max_rows;
-  bool using_ignore=0, using_bin_log=mysql_bin_log.is_open();
+  bool using_ignore= 0, using_opt_replace= 0;
+  bool using_bin_log= mysql_bin_log.is_open();
   delayed_row *row;
   DBUG_ENTER("handle_inserts");
 
@@ -1964,6 +2008,13 @@ bool delayed_insert::handle_inserts(void)
       table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
       using_ignore=1;
     }
+    if (info.handle_duplicates == DUP_REPLACE &&
+        (!table->triggers ||
+         !table->triggers->has_delete_triggers()))
+    {
+      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
+      using_opt_replace= 1;
+    }
     thd.clear_error(); // reset error for binlog
     if (write_record(&thd, table, &info))
     {
@@ -1976,6 +2027,11 @@ bool delayed_insert::handle_inserts(void)
       using_ignore=0;
       table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY);
     }
+    if (using_opt_replace)
+    {
+      using_opt_replace= 0;
+      table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE);
+    }
     if (row->query && row->log_query && using_bin_log)
     {
       Query_log_event qinfo(&thd, row->query, row->query_length, 0, FALSE);
@@ -2221,6 +2277,12 @@ select_insert::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
   thd->cuted_fields=0;
   if (info.ignore || info.handle_duplicates != DUP_ERROR)
     table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
+  if (info.handle_duplicates == DUP_REPLACE)
+  {
+    if (!table->triggers || !table->triggers->has_delete_triggers())
+      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
+    table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
+  }
   thd->no_trans_update= 0;
   thd->abort_on_warning= (!info.ignore &&
                           (thd->variables.sql_mode &
@@ -2230,6 +2292,10 @@ select_insert::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
          check_that_all_fields_are_given_values(thd, table, table_list)) ||
         table_list->prepare_where(thd, 0, TRUE) ||
         table_list->prepare_check_option(thd));
+
+  if (!res)
+    mark_fields_used_by_triggers_for_insert_stmt(thd, table,
+                                                 info.handle_duplicates);
   DBUG_RETURN(res);
 }
 
@@ -2395,6 +2461,7 @@ bool select_insert::send_eof()
 
   error= (!thd->prelocked_mode) ? table->file->end_bulk_insert():0;
   table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY);
+  table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE);
 
   /*
     We must invalidate the table in the query cache before binlog writing
@@ -2624,6 +2691,12 @@ select_create::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
   thd->cuted_fields=0;
   if (info.ignore || info.handle_duplicates != DUP_ERROR)
     table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
+  if (info.handle_duplicates == DUP_REPLACE)
+  {
+    if (!table->triggers || !table->triggers->has_delete_triggers())
+      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
+    table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
+  }
   if (!thd->prelocked_mode)
     table->file->start_bulk_insert((ha_rows) 0);
   thd->no_trans_update= 0;
@@ -2663,6 +2736,7 @@ bool select_create::send_eof()
   else
   {
     table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY);
+    table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE);
     VOID(pthread_mutex_lock(&LOCK_open));
     mysql_unlock_tables(thd, lock);
     /*
@@ -2696,6 +2770,7 @@ void select_create::abort()
   if (table)
   {
     table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY);
+    table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE);
     enum db_type table_type=table->s->db_type;
     if (!table->s->tmp_table)
     {
diff --git a/sql/sql_load.cc b/sql/sql_load.cc
index eaee5edf9f1..40e1e6b07aa 100644
--- a/sql/sql_load.cc
+++ b/sql/sql_load.cc
@@ -225,6 +225,8 @@ bool mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list,
       DBUG_RETURN(TRUE);
   }
 
+  mark_fields_used_by_triggers_for_insert_stmt(thd, table, handle_duplicates);
+
   uint tot_length=0;
   bool use_blobs= 0, use_vars= 0;
   List_iterator_fast<Item> it(fields_vars);
@@ -357,6 +359,13 @@ bool mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list,
     if (ignore ||
 	handle_duplicates == DUP_REPLACE)
       table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
+    if (handle_duplicates == DUP_REPLACE)
+    {
+      if (!table->triggers ||
+          !table->triggers->has_delete_triggers())
+        table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
+      table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
+    }
     if (!thd->prelocked_mode)
       table->file->start_bulk_insert((ha_rows) 0);
     table->copy_blobs=1;
@@ -381,6 +390,7 @@ bool mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list,
       error= 1;
     }
     table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY);
+    table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE);
     table->next_number_field=0;
   }
   ha_enable_transaction(thd, TRUE);
diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc
index ebbe6e6f558..359cd8f9500 100644
--- a/sql/sql_parse.cc
+++ b/sql/sql_parse.cc
@@ -3046,8 +3046,7 @@ end_with_restore_list:
 			       lex->key_list,
 			       select_lex->order_list.elements,
                                (ORDER *) select_lex->order_list.first,
-			       lex->duplicates, lex->ignore, &lex->alter_info,
-                               1);
+			       lex->ignore, &lex->alter_info, 1);
       }
       break;
     }
@@ -6979,7 +6978,7 @@ bool mysql_create_index(THD *thd, TABLE_LIST *table_list, List<Key> &keys)
   DBUG_RETURN(mysql_alter_table(thd,table_list->db,table_list->table_name,
 				&create_info, table_list,
 				fields, keys, 0, (ORDER*)0,
-				DUP_ERROR, 0, &alter_info, 1));
+                                0, &alter_info, 1));
 }
 
 
@@ -6997,7 +6996,7 @@ bool mysql_drop_index(THD *thd, TABLE_LIST *table_list, ALTER_INFO *alter_info)
   DBUG_RETURN(mysql_alter_table(thd,table_list->db,table_list->table_name,
 				&create_info, table_list,
 				fields, keys, 0, (ORDER*)0,
-				DUP_ERROR, 0, alter_info, 1));
+                                0, alter_info, 1));
 }
 
 
diff --git a/sql/sql_table.cc b/sql/sql_table.cc
index 275cfbaa088..9f23345f69c 100644
--- a/sql/sql_table.cc
+++ b/sql/sql_table.cc
@@ -35,9 +35,7 @@ const char *primary_key_name="PRIMARY";
 static bool check_if_keyname_exists(const char *name,KEY *start, KEY *end);
 static char *make_unique_key_name(const char *field_name,KEY *start,KEY *end);
 static int copy_data_between_tables(TABLE *from,TABLE *to,
-				    List<create_field> &create,
-				    enum enum_duplicates handle_duplicates,
-                                    bool ignore,
+                                    List<create_field> &create, bool ignore,
 				    uint order_num, ORDER *order,
 				    ha_rows *copied,ha_rows *deleted);
 static bool prepare_blob_field(THD *thd, create_field *sql_field);
@@ -3128,8 +3126,7 @@ bool mysql_alter_table(THD *thd,char *new_db, char *new_name,
                        HA_CREATE_INFO *create_info,
                        TABLE_LIST *table_list,
                        List<create_field> &fields, List<Key> &keys,
-                       uint order_num, ORDER *order,
-                       enum enum_duplicates handle_duplicates, bool ignore,
+                       uint order_num, ORDER *order, bool ignore,
                        ALTER_INFO *alter_info, bool do_send_ok)
 {
   TABLE *table,*new_table=0;
@@ -3724,8 +3721,7 @@ bool mysql_alter_table(THD *thd,char *new_db, char *new_name,
   {
     new_table->timestamp_field_type= TIMESTAMP_NO_AUTO_SET;
     new_table->next_number_field=new_table->found_next_number_field;
-    error=copy_data_between_tables(table,new_table,create_list,
-				   handle_duplicates, ignore,
+    error=copy_data_between_tables(table, new_table, create_list, ignore,
 				   order_num, order, &copied, &deleted);
   }
   thd->last_insert_id=next_insert_id;		// Needed for correct log
@@ -3948,7 +3944,6 @@ end_temporary:
 static int
 copy_data_between_tables(TABLE *from,TABLE *to,
 			 List<create_field> &create,
-			 enum enum_duplicates handle_duplicates,
                          bool ignore,
 			 uint order_num, ORDER *order,
 			 ha_rows *copied,
@@ -4051,8 +4046,7 @@ copy_data_between_tables(TABLE *from,TABLE *to,
   */
   from->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
   init_read_record(&info, thd, from, (SQL_SELECT *) 0, 1,1);
-  if (ignore ||
-      handle_duplicates == DUP_REPLACE)
+  if (ignore)
     to->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
   thd->row_count= 0;
   restore_record(to, s->default_values);        // Create empty record
@@ -4079,8 +4073,7 @@ copy_data_between_tables(TABLE *from,TABLE *to,
     }
     if ((error=to->file->write_row((byte*) to->record[0])))
     {
-      if ((!ignore &&
-	   handle_duplicates != DUP_REPLACE) ||
+      if (!ignore ||
 	  (error != HA_ERR_FOUND_DUPP_KEY &&
 	   error != HA_ERR_FOUND_DUPP_UNIQUE))
       {
@@ -4158,7 +4151,7 @@ bool mysql_recreate_table(THD *thd, TABLE_LIST *table_list,
   DBUG_RETURN(mysql_alter_table(thd, NullS, NullS, &create_info,
                                 table_list, lex->create_list,
                                 lex->key_list, 0, (ORDER *) 0,
-                                DUP_ERROR, 0, &lex->alter_info, do_send_ok));
+                                0, &lex->alter_info, do_send_ok));
 }
 
 
diff --git a/sql/sql_trigger.cc b/sql/sql_trigger.cc
index f943b014118..667fdf733e9 100644
--- a/sql/sql_trigger.cc
+++ b/sql/sql_trigger.cc
@@ -1014,8 +1014,15 @@ bool Table_triggers_list::check_n_load(THD *thd, const char *db,
         }
 
         /*
-          Let us bind Item_trigger_field objects representing access to fields
-          in old/new versions of row in trigger to Field objects in table being
+          Gather all Item_trigger_field objects representing access to fields
+          in old/new versions of row in trigger into lists containing all such
+          objects for the triggers with same action and timing.
+        */
+        triggers->trigger_fields[lex.trg_chistics.event]
+                                [lex.trg_chistics.action_time]=
+          (Item_trigger_field *)(lex.trg_table_fields.first);
+        /*
+          Also let us bind these objects to Field objects in table being
           opened.
 
           We ignore errors here, because if even something is wrong we still
@@ -1527,6 +1534,39 @@ bool Table_triggers_list::process_triggers(THD *thd, trg_event_type event,
 }
 
 
+/*
+  Mark fields of subject table which we read/set in its triggers as such.
+
+  SYNOPSIS
+    mark_fields_used()
+      thd    Current thread context
+      event  Type of event triggers for which we are going to inspect
+
+  DESCRIPTION
+    This method marks fields of subject table which are read/set in its
+    triggers as such (by setting Field::query_id equal to THD::query_id)
+    and thus informs handler that values for these fields should be
+    retrieved/stored during execution of statement.
+*/
+
+void Table_triggers_list::mark_fields_used(THD *thd, trg_event_type event)
+{
+  int action_time;
+  Item_trigger_field *trg_field;
+
+  for (action_time= 0; action_time < (int)TRG_ACTION_MAX; action_time++)
+  {
+    for (trg_field= trigger_fields[event][action_time]; trg_field;
+         trg_field= trg_field->next_trg_field)
+    {
+      /* We cannot mark fields which does not present in table. */
+      if (trg_field->field_idx != (uint)-1)
+        table->field[trg_field->field_idx]->query_id = thd->query_id;
+    }
+  }
+}
+
+
 /*
   Trigger BUG#14090 compatibility hook
 
diff --git a/sql/sql_trigger.h b/sql/sql_trigger.h
index b67c22e0588..e736c3e0e1a 100644
--- a/sql/sql_trigger.h
+++ b/sql/sql_trigger.h
@@ -25,6 +25,11 @@ class Table_triggers_list: public Sql_alloc
 {
   /* Triggers as SPs grouped by event, action_time */
   sp_head           *bodies[TRG_EVENT_MAX][TRG_ACTION_MAX];
+  /*
+    Heads of the lists linking items for all fields used in triggers
+    grouped by event and action_time.
+  */
+  Item_trigger_field *trigger_fields[TRG_EVENT_MAX][TRG_ACTION_MAX];
   /*
     Copy of TABLE::Field array with field pointers set to TABLE::record[1]
     buffer instead of TABLE::record[0] (used for OLD values in on UPDATE
@@ -82,6 +87,7 @@ public:
     record1_field(0), table(table_arg)
   {
     bzero((char *)bodies, sizeof(bodies));
+    bzero((char *)trigger_fields, sizeof(trigger_fields));
     bzero((char *)&subject_table_grants, sizeof(subject_table_grants));
   }
   ~Table_triggers_list();
@@ -119,6 +125,8 @@ public:
 
   void set_table(TABLE *new_table);
 
+  void mark_fields_used(THD *thd, trg_event_type event);
+
   friend class Item_trigger_field;
   friend int sp_cache_routines_and_add_tables_for_triggers(THD *thd, LEX *lex,
                                                             TABLE_LIST *table);
diff --git a/sql/sql_update.cc b/sql/sql_update.cc
index c2b7624c9e7..f282cf19de1 100644
--- a/sql/sql_update.cc
+++ b/sql/sql_update.cc
@@ -433,6 +433,9 @@ int mysql_update(THD *thd,
                                (MODE_STRICT_TRANS_TABLES |
                                 MODE_STRICT_ALL_TABLES)));
 
+  if (table->triggers)
+    table->triggers->mark_fields_used(thd, TRG_EVENT_UPDATE);
+
   while (!(error=info.read_record(&info)) && !thd->killed)
   {
     if (!(select && select->skip_record()))
@@ -755,6 +758,9 @@ reopen_tables:
         DBUG_RETURN(TRUE);
       }
 
+      if (table->triggers)
+        table->triggers->mark_fields_used(thd, TRG_EVENT_UPDATE);
+
       DBUG_PRINT("info",("setting table `%s` for update", tl->alias));
       /*
         If table will be updated we should not downgrade lock for it and

From d6f47c31b660e7ed955edb26ebc6c7816e82f34f Mon Sep 17 00:00:00 2001
From: "dlenev@mysql.com" <>
Date: Thu, 6 Jul 2006 13:33:23 +0400
Subject: [PATCH 2/2] After merge fixes for patch solving bug#18437 "Wrong
 values inserted with a before update trigger on NDB table".

Two main changes:
- We use TABLE::read_set/write_set bitmaps for marking fields used by
  statement instead of Field::query_id in 5.1.
- Now when we mark columns used by statement we take into account columns
  used by table's triggers instead of marking all columns as used if table
  has triggers.
---
 mysql-test/r/federated.result |  6 +--
 mysql-test/t/federated.test   |  6 +--
 sql/ha_partition.cc           | 21 ++++++++++
 sql/item.cc                   | 10 ++---
 sql/log_event.cc              |  1 +
 sql/mysql_priv.h              |  2 -
 sql/sql_insert.cc             | 74 ++++++++---------------------------
 sql/sql_load.cc               | 14 ++-----
 sql/sql_trigger.cc            | 11 ++++--
 sql/sql_trigger.h             |  6 +--
 sql/table.cc                  | 36 +++++------------
 11 files changed, 72 insertions(+), 115 deletions(-)

diff --git a/mysql-test/r/federated.result b/mysql-test/r/federated.result
index 638695cd5b9..0669ef98874 100644
--- a/mysql-test/r/federated.result
+++ b/mysql-test/r/federated.result
@@ -1611,20 +1611,20 @@ create trigger federated.t1_bi before insert on federated.t1 for each row set ne
 create table federated.t2 (a int, b int);
 insert into federated.t2 values (13, 17), (19, 23);
 insert into federated.t1 (a, b) values (1, 2), (3, 5), (7, 11);
-select * from federated.t1;
+select * from federated.t1 order by a;
 a	b	c
 1	2	2
 3	5	15
 7	11	77
 delete from federated.t1;
 insert into federated.t1 (a, b) select * from federated.t2;
-select * from federated.t1;
+select * from federated.t1 order by a;
 a	b	c
 13	17	221
 19	23	437
 delete from federated.t1;
 load data infile '../std_data_ln/loaddata5.dat' into table federated.t1 fields terminated by '' enclosed by '' ignore 1 lines (a, b);
-select * from federated.t1;
+select * from federated.t1 order by a;
 a	b	c
 3	4	12
 5	6	30
diff --git a/mysql-test/t/federated.test b/mysql-test/t/federated.test
index 9144f2965df..5c464e1210f 100644
--- a/mysql-test/t/federated.test
+++ b/mysql-test/t/federated.test
@@ -1391,15 +1391,15 @@ insert into federated.t2 values (13, 17), (19, 23);
 # Each of three statements should correctly set values for all three fields
 # insert
 insert into federated.t1 (a, b) values (1, 2), (3, 5), (7, 11);
-select * from federated.t1;
+select * from federated.t1 order by a;
 delete from federated.t1;
 # insert ... select
 insert into federated.t1 (a, b) select * from federated.t2;
-select * from federated.t1;
+select * from federated.t1 order by a;
 delete from federated.t1;
 # load
 load data infile '../std_data_ln/loaddata5.dat' into table federated.t1 fields terminated by '' enclosed by '' ignore 1 lines (a, b);
-select * from federated.t1;
+select * from federated.t1 order by a;
 drop tables federated.t1, federated.t2;
 
 connection slave;
diff --git a/sql/ha_partition.cc b/sql/ha_partition.cc
index 7929257d608..2a3589bf65c 100644
--- a/sql/ha_partition.cc
+++ b/sql/ha_partition.cc
@@ -4665,6 +4665,27 @@ int ha_partition::extra(enum ha_extra_function operation)
      */
     break;
   }
+  case HA_EXTRA_WRITE_CAN_REPLACE:
+  case HA_EXTRA_WRITE_CANNOT_REPLACE:
+  {
+    /*
+      Informs handler that write_row() can replace rows which conflict
+      with row being inserted by PK/unique key without reporting error
+      to the SQL-layer.
+
+      This optimization is not safe for partitioned table in general case
+      since we may have to put new version of row into partition which is
+      different from partition in which old version resides (for example
+      when we partition by non-PK column or by some column which is not
+      part of unique key which were violated).
+      And since NDB which is the only engine at the moment that supports
+      this optimization handles partitioning on its own we simple disable
+      it here. (BTW for NDB this optimization is safe since it supports
+      only KEY partitioning and won't use this optimization for tables
+      which have additional unique constraints).
+    */
+    break;
+  }
   default:
   {
     /* Temporary crash to discover what is wrong */
diff --git a/sql/item.cc b/sql/item.cc
index bf570da7caf..c4a3ca3f596 100644
--- a/sql/item.cc
+++ b/sql/item.cc
@@ -5437,11 +5437,11 @@ void Item_trigger_field::setup_field(THD *thd, TABLE *table,
                                      GRANT_INFO *table_grant_info)
 {
   /*
-    There is no sense in marking fields used by trigger with current value
-    of THD::query_id since it is completely unrelated to the THD::query_id
-    value for statements which will invoke trigger. So instead we use
-    Table_triggers_list::mark_fields_used() method which is called during
-    execution of these statements.
+    It is too early to mark fields used here, because before execution
+    of statement that will invoke trigger other statements may use same
+    TABLE object, so all such mark-up will be wiped out.
+    So instead we do it in Table_triggers_list::mark_fields_used()
+    method which is called during execution of these statements.
   */
   enum_mark_columns save_mark_used_columns= thd->mark_used_columns;
   thd->mark_used_columns= MARK_COLUMNS_NONE;
diff --git a/sql/log_event.cc b/sql/log_event.cc
index 36805e0043d..c922c592112 100644
--- a/sql/log_event.cc
+++ b/sql/log_event.cc
@@ -6112,6 +6112,7 @@ int Write_rows_log_event::do_before_row_operations(TABLE *table)
   thd->lex->sql_command= SQLCOM_REPLACE;
 
   table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);  // Needed for ndbcluster
+  table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);  // Needed for ndbcluster
   table->file->extra(HA_EXTRA_IGNORE_NO_KEY);   // Needed for ndbcluster
   /*
     TODO: the cluster team (Tomas?) says that it's better if the engine knows
diff --git a/sql/mysql_priv.h b/sql/mysql_priv.h
index e06a5ad59ed..a7c46577ab3 100644
--- a/sql/mysql_priv.h
+++ b/sql/mysql_priv.h
@@ -908,8 +908,6 @@ bool mysql_insert(THD *thd,TABLE_LIST *table,List<Item> &fields,
                   bool ignore);
 int check_that_all_fields_are_given_values(THD *thd, TABLE *entry,
                                            TABLE_LIST *table_list);
-void mark_fields_used_by_triggers_for_insert_stmt(THD *thd, TABLE *table,
-                                                  enum_duplicates duplic);
 bool mysql_prepare_delete(THD *thd, TABLE_LIST *table_list, Item **conds);
 bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds,
                   SQL_LIST *order, ha_rows rows, ulonglong options,
diff --git a/sql/sql_insert.cc b/sql/sql_insert.cc
index cfc7d4ff3b0..b27b9e8f13e 100644
--- a/sql/sql_insert.cc
+++ b/sql/sql_insert.cc
@@ -181,9 +181,6 @@ static int check_insert_fields(THD *thd, TABLE_LIST *table_list,
       }
     }
   }
-  if (table->found_next_number_field)
-    table->mark_auto_increment_column();
-  table->mark_columns_needed_for_insert();
   // For the values we need select_priv
 #ifndef NO_EMBEDDED_ACCESS_CHECKS
   table->grant.want_privilege= (SELECT_ACL & ~table->grant.privilege);
@@ -255,33 +252,6 @@ static int check_update_fields(THD *thd, TABLE_LIST *insert_table_list,
 }
 
 
-/*
-  Mark fields used by triggers for INSERT-like statement.
-
-  SYNOPSIS
-    mark_fields_used_by_triggers_for_insert_stmt()
-      thd     The current thread
-      table   Table to which insert will happen
-      duplic  Type of duplicate handling for insert which will happen
-
-  NOTE
-    For REPLACE there is no sense in marking particular fields
-    used by ON DELETE trigger as to execute it properly we have
-    to retrieve and store values for all table columns anyway.
-*/
-
-void mark_fields_used_by_triggers_for_insert_stmt(THD *thd, TABLE *table,
-                                                  enum_duplicates duplic)
-{
-  if (table->triggers)
-  {
-    table->triggers->mark_fields_used(thd, TRG_EVENT_INSERT);
-    if (duplic == DUP_UPDATE)
-      table->triggers->mark_fields_used(thd, TRG_EVENT_UPDATE);
-  }
-}
-
-
 bool mysql_insert(THD *thd,TABLE_LIST *table_list,
                   List<Item> &fields,
                   List<List_item> &values_list,
@@ -442,17 +412,9 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list,
   thd->proc_info="update";
   if (duplic != DUP_ERROR || ignore)
     table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
-  if (duplic == DUP_REPLACE)
-  {
-    if (!table->triggers || !table->triggers->has_delete_triggers())
-      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
-    /*
-      REPLACE should change values of all columns so we should mark
-      all columns as columns to be set. As nice side effect we will
-      retrieve columns which values are needed for ON DELETE triggers.
-    */
-    table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
-  }
+  if (duplic == DUP_REPLACE &&
+      (!table->triggers || !table->triggers->has_delete_triggers()))
+    table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
   /*
     let's *try* to start bulk inserts. It won't necessary
     start them as values_list.elements should be greater than
@@ -481,7 +443,7 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list,
     error= 1;
   }
 
-  mark_fields_used_by_triggers_for_insert_stmt(thd, table, duplic);
+  table->mark_columns_needed_for_insert();
 
   if (table_list->prepare_where(thd, 0, TRUE) ||
       table_list->prepare_check_option(thd))
@@ -2346,12 +2308,9 @@ select_insert::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
   thd->cuted_fields=0;
   if (info.ignore || info.handle_duplicates != DUP_ERROR)
     table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
-  if (info.handle_duplicates == DUP_REPLACE)
-  {
-    if (!table->triggers || !table->triggers->has_delete_triggers())
-      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
-    table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
-  }
+  if (info.handle_duplicates == DUP_REPLACE &&
+      (!table->triggers || !table->triggers->has_delete_triggers()))
+    table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
   thd->no_trans_update= 0;
   thd->abort_on_warning= (!info.ignore &&
                           (thd->variables.sql_mode &
@@ -2363,8 +2322,8 @@ select_insert::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
         table_list->prepare_check_option(thd));
 
   if (!res)
-    mark_fields_used_by_triggers_for_insert_stmt(thd, table,
-                                                 info.handle_duplicates);
+    table->mark_columns_needed_for_insert();
+
   DBUG_RETURN(res);
 }
 
@@ -2840,12 +2799,9 @@ select_create::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
   thd->cuted_fields=0;
   if (info.ignore || info.handle_duplicates != DUP_ERROR)
     table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
-  if (info.handle_duplicates == DUP_REPLACE)
-  {
-    if (!table->triggers || !table->triggers->has_delete_triggers())
-      table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
-    table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
-  }
+  if (info.handle_duplicates == DUP_REPLACE &&
+      (!table->triggers || !table->triggers->has_delete_triggers()))
+    table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
   if (!thd->prelocked_mode)
     table->file->ha_start_bulk_insert((ha_rows) 0);
   thd->no_trans_update= 0;
@@ -2853,8 +2809,10 @@ select_create::prepare(List<Item> &values, SELECT_LEX_UNIT *u)
                           (thd->variables.sql_mode &
                            (MODE_STRICT_TRANS_TABLES |
                             MODE_STRICT_ALL_TABLES)));
-  DBUG_RETURN(check_that_all_fields_are_given_values(thd, table,
-                                                     table_list));
+  if (check_that_all_fields_are_given_values(thd, table, table_list))
+    DBUG_RETURN(1);
+  table->mark_columns_needed_for_insert();
+  DBUG_RETURN(0);
 }
 
 
diff --git a/sql/sql_load.cc b/sql/sql_load.cc
index 07a33d0551d..db6e3422478 100644
--- a/sql/sql_load.cc
+++ b/sql/sql_load.cc
@@ -187,9 +187,6 @@ bool mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list,
   table= table_list->table;
   transactional_table= table->file->has_transactions();
 
-  if (table->found_next_number_field)
-    table->mark_auto_increment_column();
-
   if (!fields_vars.elements)
   {
     Field **field;
@@ -232,7 +229,7 @@ bool mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list,
       DBUG_RETURN(TRUE);
   }
 
-  mark_fields_used_by_triggers_for_insert_stmt(thd, table, handle_duplicates);
+  table->mark_columns_needed_for_insert();
 
   uint tot_length=0;
   bool use_blobs= 0, use_vars= 0;
@@ -364,13 +361,10 @@ bool mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list,
     if (ignore ||
 	handle_duplicates == DUP_REPLACE)
       table->file->extra(HA_EXTRA_IGNORE_DUP_KEY);
-    if (handle_duplicates == DUP_REPLACE)
-    {
-      if (!table->triggers ||
-          !table->triggers->has_delete_triggers())
+    if (handle_duplicates == DUP_REPLACE &&
+        (!table->triggers ||
+         !table->triggers->has_delete_triggers()))
         table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE);
-      table->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
-    }
     if (!thd->prelocked_mode)
       table->file->ha_start_bulk_insert((ha_rows) 0);
     table->copy_blobs=1;
diff --git a/sql/sql_trigger.cc b/sql/sql_trigger.cc
index 79b8290c077..e125c49dcbf 100644
--- a/sql/sql_trigger.cc
+++ b/sql/sql_trigger.cc
@@ -1543,12 +1543,12 @@ bool Table_triggers_list::process_triggers(THD *thd, trg_event_type event,
 
   DESCRIPTION
     This method marks fields of subject table which are read/set in its
-    triggers as such (by setting Field::query_id equal to THD::query_id)
+    triggers as such (by properly updating TABLE::read_set/write_set)
     and thus informs handler that values for these fields should be
     retrieved/stored during execution of statement.
 */
 
-void Table_triggers_list::mark_fields_used(THD *thd, trg_event_type event)
+void Table_triggers_list::mark_fields_used(trg_event_type event)
 {
   int action_time;
   Item_trigger_field *trg_field;
@@ -1560,9 +1560,14 @@ void Table_triggers_list::mark_fields_used(THD *thd, trg_event_type event)
     {
       /* We cannot mark fields which does not present in table. */
       if (trg_field->field_idx != (uint)-1)
-        table->field[trg_field->field_idx]->query_id = thd->query_id;
+      {
+        bitmap_set_bit(table->read_set, trg_field->field_idx);
+        if (trg_field->get_settable_routine_parameter())
+          bitmap_set_bit(table->write_set, trg_field->field_idx);
+      }
     }
   }
+  table->file->column_bitmaps_signal();
 }
 
 
diff --git a/sql/sql_trigger.h b/sql/sql_trigger.h
index 55744dc4d2c..13a919c09ca 100644
--- a/sql/sql_trigger.h
+++ b/sql/sql_trigger.h
@@ -125,7 +125,7 @@ public:
 
   void set_table(TABLE *new_table);
 
-  void mark_fields_used(THD *thd, trg_event_type event);
+  void mark_fields_used(trg_event_type event);
 
   friend class Item_trigger_field;
   friend int sp_cache_routines_and_add_tables_for_triggers(THD *thd, LEX *lex,
@@ -140,10 +140,6 @@ private:
                                      const char *db_name,
                                      LEX_STRING *old_table_name,
                                      LEX_STRING *new_table_name);
-  friend void st_table::mark_columns_needed_for_insert(void);
-  friend void st_table::mark_columns_needed_for_update(void);
-  friend void st_table::mark_columns_needed_for_delete(void);
-
 };
 
 extern const LEX_STRING trg_action_time_type_names[];
diff --git a/sql/table.cc b/sql/table.cc
index a96ca0da881..0320890ed4a 100644
--- a/sql/table.cc
+++ b/sql/table.cc
@@ -3925,16 +3925,7 @@ void st_table::mark_auto_increment_column()
 void st_table::mark_columns_needed_for_delete()
 {
   if (triggers)
-  {
-    if (triggers->bodies[TRG_EVENT_DELETE][TRG_ACTION_BEFORE] ||
-        triggers->bodies[TRG_EVENT_DELETE][TRG_ACTION_AFTER])
-    {
-      /* TODO: optimize to only add columns used by trigger */
-      use_all_columns();
-      return;
-    }
-  }
-
+    triggers->mark_fields_used(TRG_EVENT_DELETE);
   if (file->ha_table_flags() & HA_REQUIRES_KEY_COLUMNS_FOR_DELETE)
   {
     Field **reg_field;
@@ -3985,15 +3976,7 @@ void st_table::mark_columns_needed_for_update()
 {
   DBUG_ENTER("mark_columns_needed_for_update");
   if (triggers)
-  {
-    if (triggers->bodies[TRG_EVENT_UPDATE][TRG_ACTION_BEFORE] ||
-        triggers->bodies[TRG_EVENT_UPDATE][TRG_ACTION_AFTER])
-    {
-      /* TODO: optimize to only add columns used by trigger */
-      use_all_columns();
-      DBUG_VOID_RETURN;
-    }
-  }
+    triggers->mark_fields_used(TRG_EVENT_UPDATE);
   if (file->ha_table_flags() & HA_REQUIRES_KEY_COLUMNS_FOR_DELETE)
   {
     /* Mark all used key columns for read */
@@ -4036,13 +4019,14 @@ void st_table::mark_columns_needed_for_insert()
 {
   if (triggers)
   {
-    if (triggers->bodies[TRG_EVENT_INSERT][TRG_ACTION_BEFORE] ||
-        triggers->bodies[TRG_EVENT_INSERT][TRG_ACTION_AFTER])
-    {
-      /* TODO: optimize to only add columns used by trigger */
-      use_all_columns();
-      return;
-    }
+    /*
+      We don't need to mark columns which are used by ON DELETE and
+      ON UPDATE triggers, which may be invoked in case of REPLACE or
+      INSERT ... ON DUPLICATE KEY UPDATE, since before doing actual
+      row replacement or update write_record() will mark all table
+      fields as used.
+    */
+    triggers->mark_fields_used(TRG_EVENT_INSERT);
   }
   if (found_next_number_field)
     mark_auto_increment_column();