Merge poseidon.ndb.mysql.com:/home/tomas/mysql-5.0

into poseidon.ndb.mysql.com:/home/tomas/mysql-5.0-ndb

client/mysqltest.c

@@ -2896,7 +2896,7 @@ static int run_query_stmt(MYSQL *mysql, struct st_query *q, int flags)
 
       /* Allocate array with bind structs, lengths and NULL flags */
       bind= (MYSQL_BIND*)      my_malloc(num_fields * sizeof(MYSQL_BIND),
-                                         MYF(MY_WME | MY_FAE));
+                                         MYF(MY_WME | MY_FAE | MY_ZEROFILL));
       length= (unsigned long*) my_malloc(num_fields * sizeof(unsigned long),
                                          MYF(MY_WME | MY_FAE));
       is_null= (my_bool*)      my_malloc(num_fields * sizeof(my_bool),

include/queues.h

@@ -53,6 +53,7 @@ int resize_queue(QUEUE *queue, uint max_elements);
 void delete_queue(QUEUE *queue);
 void queue_insert(QUEUE *queue,byte *element);
 byte *queue_remove(QUEUE *queue,uint idx);
+#define queue_remove_all(queue) { (queue)->elements= 0; }
 void _downheap(QUEUE *queue,uint idx);
 void queue_fix(QUEUE *queue);
 #define is_queue_inited(queue) ((queue)->root != 0)

innobase/include/lock0lock.h

@@ -612,6 +612,8 @@ extern lock_sys_t*	lock_sys;
 #define LOCK_TABLE	16	/* these type values should be so high that */
 #define	LOCK_REC	32	/* they can be ORed to the lock mode */
 #define LOCK_TABLE_EXP	80	/* explicit table lock (80 = 16 + 64) */
+#define	LOCK_TABLE_TRANSACTIONAL	144
+				/* transactional table lock (144 = 16 + 128)*/
 #define LOCK_TYPE_MASK	0xF0UL	/* mask used to extract lock type from the
 				type_mode field in a lock */
 /* Waiting lock flag */

innobase/include/trx0trx.h

@@ -464,6 +464,10 @@ struct trx_struct{
 	ulint		n_lock_table_exp;/* number of explicit table locks
 					(LOCK TABLES) reserved by the
 					transaction, stored in trx_locks */
+	ulint		n_lock_table_transactional;
+					/* number of transactional table locks
+					(LOCK TABLES..WHERE ENGINE) reserved by
+					the transaction, stored in trx_locks */
 	UT_LIST_NODE_T(trx_t)
 			trx_list;	/* list of transactions */
 	UT_LIST_NODE_T(trx_t)

innobase/lock/lock0lock.c

@@ -2207,7 +2207,8 @@ lock_grant(
                 release it at the end of the SQL statement */
 
                 lock->trx->auto_inc_lock = lock;
-        } else if (lock_get_type(lock) == LOCK_TABLE_EXP) {
+        } else if (lock_get_type(lock) == LOCK_TABLE_EXP ||
+			lock_get_type(lock) == LOCK_TABLE_TRANSACTIONAL) {
 		ut_a(lock_get_mode(lock) == LOCK_S
 			|| lock_get_mode(lock) == LOCK_X);
 	}
@@ -3421,6 +3422,10 @@ lock_table_create(
 		lock->trx->n_lock_table_exp++;
 	}
 
+	if (lock_get_type(lock) == LOCK_TABLE_TRANSACTIONAL) {
+		lock->trx->n_lock_table_transactional++;
+	}
+
 	lock->un_member.tab_lock.table = table;
 
 	UT_LIST_ADD_LAST(un_member.tab_lock.locks, table->locks, lock);
@@ -3458,7 +3463,11 @@ lock_table_remove_low(
 	}
 
 	if (lock_get_type(lock) == LOCK_TABLE_EXP) {
-		lock->trx->n_lock_table_exp--;
+		trx->n_lock_table_exp--;
+	}
+
+        if (lock_get_type(lock) == LOCK_TABLE_TRANSACTIONAL) {
+		trx->n_lock_table_transactional--;
 	}
 
 	UT_LIST_REMOVE(trx_locks, trx->trx_locks, lock);
@@ -3592,7 +3601,8 @@ lock_table(
 				DB_DEADLOCK, or DB_QUE_THR_SUSPENDED */
 	ulint		flags,	/* in: if BTR_NO_LOCKING_FLAG bit is set,
 				does nothing;
-				if LOCK_TABLE_EXP bits are set,
+				if LOCK_TABLE_EXP|LOCK_TABLE_TRANSACTIONAL
+				bits are set,
 				creates an explicit table lock */
 	dict_table_t*	table,	/* in: database table in dictionary cache */
 	ulint		mode,	/* in: lock mode */
@@ -3608,7 +3618,8 @@ lock_table(
 		return(DB_SUCCESS);
 	}
 
-	ut_a(flags == 0 || flags == LOCK_TABLE_EXP);
+	ut_a(flags == 0 || flags == LOCK_TABLE_EXP || 
+					flags == LOCK_TABLE_TRANSACTIONAL);
 
 	trx = thr_get_trx(thr);
 
@@ -3631,7 +3642,7 @@ lock_table(
 		/* Another trx has a request on the table in an incompatible
 		mode: this trx may have to wait */
 
-		err = lock_table_enqueue_waiting(mode, table, thr);
+		err = lock_table_enqueue_waiting(mode | flags, table, thr);
 			
 		lock_mutex_exit_kernel();
 
@@ -3722,7 +3733,8 @@ lock_table_dequeue(
 	ut_ad(mutex_own(&kernel_mutex));
 #endif /* UNIV_SYNC_DEBUG */
 	ut_a(lock_get_type(in_lock) == LOCK_TABLE ||
-		lock_get_type(in_lock) == LOCK_TABLE_EXP);
+		lock_get_type(in_lock) == LOCK_TABLE_EXP ||
+		lock_get_type(in_lock) == LOCK_TABLE_TRANSACTIONAL);
 
 	lock = UT_LIST_GET_NEXT(un_member.tab_lock.locks, in_lock);
 
@@ -3826,7 +3838,9 @@ lock_release_off_kernel(
 			}
 
 			lock_table_dequeue(lock);
-			if (lock_get_type(lock) == LOCK_TABLE_EXP) {
+
+			if (lock_get_type(lock) == LOCK_TABLE_EXP ||
+			    lock_get_type(lock) == LOCK_TABLE_TRANSACTIONAL) {
 				ut_a(lock_get_mode(lock) == LOCK_S
 					|| lock_get_mode(lock) == LOCK_X);
 			}
@@ -3850,6 +3864,7 @@ lock_release_off_kernel(
 
 	ut_a(trx->auto_inc_lock == NULL);
 	ut_a(trx->n_lock_table_exp == 0);
+	ut_a(trx->n_lock_table_transactional == 0);
 }
 
 /*************************************************************************
@@ -3915,6 +3930,7 @@ lock_release_tables_off_kernel(
 	}
 
 	ut_a(trx->n_lock_table_exp == 0);
+	ut_a(trx->n_lock_table_transactional == 0);
 }
 
 /*************************************************************************
@@ -4028,11 +4044,15 @@ lock_table_print(
 	ut_ad(mutex_own(&kernel_mutex));
 #endif /* UNIV_SYNC_DEBUG */
 	ut_a(lock_get_type(lock) == LOCK_TABLE ||
-		lock_get_type(lock) == LOCK_TABLE_EXP);
+		lock_get_type(lock) == LOCK_TABLE_EXP ||
+		lock_get_type(lock) == LOCK_TABLE_TRANSACTIONAL);
 
 	if (lock_get_type(lock) == LOCK_TABLE_EXP) {
 		fputs("EXPLICIT ", file);
+	} else if (lock_get_type(lock) == LOCK_TABLE_TRANSACTIONAL) {
+		fputs("TRANSACTIONAL ", file);
 	}
+
 	fputs("TABLE LOCK table ", file);
 	ut_print_name(file, lock->trx, lock->un_member.tab_lock.table->name);
 	fprintf(file, " trx id %lu %lu",
@@ -4418,6 +4438,7 @@ lock_table_queue_validate(
 
 	while (lock) {
 		ut_a(((lock->trx)->conc_state == TRX_ACTIVE)
+		     || ((lock->trx)->conc_state == TRX_PREPARED)
 		     || ((lock->trx)->conc_state == TRX_COMMITTED_IN_MEMORY));
 	
 		if (!lock_get_wait(lock)) {
@@ -4465,6 +4486,7 @@ lock_rec_queue_validate(
 
 		while (lock) {
 			ut_a(lock->trx->conc_state == TRX_ACTIVE
+			     || lock->trx->conc_state == TRX_PREPARED
 		     	     || lock->trx->conc_state
 						== TRX_COMMITTED_IN_MEMORY);
 	
@@ -4519,6 +4541,7 @@ lock_rec_queue_validate(
 
 	while (lock) {
 		ut_a(lock->trx->conc_state == TRX_ACTIVE
+		     || lock->trx->conc_state == TRX_PREPARED
 		     || lock->trx->conc_state == TRX_COMMITTED_IN_MEMORY);
 		ut_a(trx_in_trx_list(lock->trx));
 	
@@ -4601,6 +4624,7 @@ loop:
 
 	ut_a(trx_in_trx_list(lock->trx));
 	ut_a(lock->trx->conc_state == TRX_ACTIVE
+		     || lock->trx->conc_state == TRX_PREPARED
 		     || lock->trx->conc_state == TRX_COMMITTED_IN_MEMORY);
 	
 	for (i = nth_bit; i < lock_rec_get_n_bits(lock); i++) {

innobase/row/row0mysql.c

@@ -784,7 +784,7 @@ row_lock_table_for_mysql(
 					table handle */
 	dict_table_t*	table,		/* in: table to lock, or NULL
 					if prebuilt->table should be
-					locked as LOCK_TABLE_EXP |
+					locked or a
 					prebuilt->select_lock_type */
 	ulint		mode)		/* in: lock mode of table */
 {
@@ -822,8 +822,14 @@ run_again:
 	if (table) {
 		err = lock_table(0, table, mode, thr);
 	} else {
-		err = lock_table(LOCK_TABLE_EXP, prebuilt->table,
+		if (mode == LOCK_TABLE_TRANSACTIONAL) {
-			prebuilt->select_lock_type, thr);
+			err = lock_table(LOCK_TABLE_TRANSACTIONAL, 
+					prebuilt->table,
+					prebuilt->select_lock_type, thr);
+		} else {
+			err = lock_table(LOCK_TABLE_EXP, prebuilt->table,
+					prebuilt->select_lock_type, thr);
+		}
 	}
 
 	trx->error_state = err;

innobase/trx/trx0trx.c

@@ -153,6 +153,7 @@ trx_create(
 
 	trx->auto_inc_lock = NULL;
 	trx->n_lock_table_exp = 0;
+	trx->n_lock_table_transactional = 0;
 	
 	trx->read_view_heap = mem_heap_create(256);
 	trx->read_view = NULL;
@@ -285,6 +286,7 @@ trx_free(
 	ut_a(!trx->has_search_latch);
 	ut_a(!trx->auto_inc_lock);
 	ut_a(!trx->n_lock_table_exp);
+	ut_a(!trx->n_lock_table_transactional);
 
 	ut_a(trx->dict_operation_lock_mode == 0);
 
@@ -1645,10 +1647,15 @@ trx_print(
 	putc('\n', f);
   	
         if (trx->n_mysql_tables_in_use > 0 || trx->mysql_n_tables_locked > 0) {
+		fprintf(f, "mysql tables in use %lu, locked %lu\n",
+					(ulong) trx->n_mysql_tables_in_use,
+					(ulong) trx->mysql_n_tables_locked);
+	}
 
-                fprintf(f, "mysql tables in use %lu, locked %lu\n",
+	if (trx->n_lock_table_transactional > 0 || trx->n_lock_table_exp > 0) {
-                           (ulong) trx->n_mysql_tables_in_use,
+fprintf(f, "mysql explicit table locks %lu, transactional table locks %lu\n",
-                           (ulong) trx->mysql_n_tables_locked);
+				(ulong) trx->n_lock_table_exp,
+				(ulong) trx->n_lock_table_transactional);
         }
 
 	newline = TRUE;

mysql-test/r/derived.result

@@ -330,3 +330,8 @@ SELECT MIN(price) min, MAX(price) max, AVG(price) avg FROM (SELECT SUBSTRING( MA
 min	max	avg
 10.00	10.00	10
 DROP TABLE t1;
+CREATE TABLE t1 (a char(10), b char(10));
+INSERT INTO t1 VALUES ('root','localhost'), ('root','%');
+SELECT * FROM (SELECT (SELECT a.a FROM t1 AS a WHERE a.a = b.a) FROM t1 AS b) AS c;
+ERROR 21000: Subquery returns more than 1 row
+DROP TABLE t1;

mysql-test/r/index_merge.result

@@ -1,4 +1,4 @@
-drop table if exists t0, t1, t2, t3,t4;
+drop table if exists t0, t1, t2, t3, t4;
 create table t0
 (
 key1 int not null, 
@@ -335,4 +335,55 @@ key1	key2	key3	key4	key5	key6	key7	key8
 select count(*) from t0;
 count(*)
 1021
+drop table t4;
+create table t4 (a int);
+insert into t4 values (1),(4),(3);
+set @save_join_buffer_size=@@join_buffer_size;
+set join_buffer_size= 4000;
+show variables like 'join_buffer_size';
+Variable_name	Value
+join_buffer_size	8228
+explain select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+where (A.key1 < 500000 or A.key2 < 3)
+and   (B.key1 < 500000 or B.key2 < 3);
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	A	index_merge	i1,i2	i1,i2	4,4	NULL	1016	Using sort_union(i1,i2); Using where
+1	SIMPLE	B	index_merge	i1,i2	i1,i2	4,4	NULL	1016	Using sort_union(i1,i2); Using where
+select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+where (A.key1 < 500000 or A.key2 < 3)
+and   (B.key1 < 500000 or B.key2 < 3);
+max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5)
+10240
+update t0 set key1=1;
+explain select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+where (A.key1 = 1 or A.key2 = 1)
+and   (B.key1 = 1 or B.key2 = 1);
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	A	index_merge	i1,i2	i1,i2	4,4	NULL	1020	Using union(i1,i2); Using where
+1	SIMPLE	B	index_merge	i1,i2	i1,i2	4,4	NULL	1020	Using union(i1,i2); Using where
+select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+where (A.key1 = 1 or A.key2 = 1)
+and   (B.key1 = 1 or B.key2 = 1);
+max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5)
+8194
+alter table t0 add filler1 char(200), add filler2 char(200), add filler3 char(200);
+update t0 set key2=1, key3=1, key4=1, key5=1,key6=1,key7=1 where key7 < 500;
+explain select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+from t0 as A, t0 as B 
+where (A.key1 = 1 and A.key2 = 1 and A.key3 = 1 and A.key4=1 and A.key5=1 and A.key6=1 and A.key7 = 1 or A.key8=1)
+and (B.key1 = 1 and B.key2 = 1 and B.key3 = 1 and B.key4=1 and B.key5=1 and B.key6=1 and B.key7 = 1 or B.key8=1);
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	A	index_merge	i1,i2,i3,i4,i5,i6,i7,i8	i2,i3,i4,i5,i6,i8	4,4,4,4,4,4	NULL	16	Using union(intersect(i2,i3,i4,i5,i6),i8); Using where
+1	SIMPLE	B	index_merge	i1,i2,i3,i4,i5,i6,i7,i8	i2,i3,i4,i5,i6,i8	4,4,4,4,4,4	NULL	16	Using union(intersect(i2,i3,i4,i5,i6),i8); Using where
+select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+from t0 as A, t0 as B 
+where (A.key1 = 1 and A.key2 = 1 and A.key3 = 1 and A.key4=1 and A.key5=1 and A.key6=1 and A.key7 = 1 or A.key8=1)
+and (B.key1 = 1 and B.key2 = 1 and B.key3 = 1 and B.key4=1 and B.key5=1 and B.key6=1 and B.key7 = 1 or B.key8=1);
+max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5)
+8186
+set join_buffer_size= @save_join_buffer_size;
 drop table t0, t1, t2, t3, t4;

mysql-test/r/join_outer.result

@@ -853,4 +853,19 @@ id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
 1	SIMPLE	t1	const	PRIMARY	PRIMARY	4	const	1	Using index
 1	SIMPLE	t2	ALL	NULL	NULL	NULL	NULL	2	
 1	SIMPLE	t3	ALL	NULL	NULL	NULL	NULL	2	
+drop table t1,t2;
+create table t1 (a int, b int);
+insert into t1 values (1,1),(2,2),(3,3);
+create table t2 (a int, b int);
+insert into t2 values (1,1), (2,2);
+select * from t2 right join t1 on t2.a=t1.a;
+a	b	a	b
+1	1	1	1
+2	2	2	2
+NULL	NULL	3	3
+select straight_join * from t2 right join t1 on t2.a=t1.a;
+a	b	a	b
+1	1	1	1
+2	2	2	2
+NULL	NULL	3	3
 DROP TABLE t0,t1,t2,t3;

mysql-test/t/derived.test

@@ -214,3 +214,16 @@ CREATE TABLE `t1` ( `itemid` int(11) NOT NULL default '0', `grpid` varchar(15) N
 insert into t1 values (128, 'rozn', 2, now(), 10),(128, 'rozn', 1, now(), 10);
 SELECT MIN(price) min, MAX(price) max, AVG(price) avg FROM (SELECT SUBSTRING( MAX(concat(date_,";",price)), 12) price FROM t1 WHERE itemid=128 AND  grpid='rozn' GROUP BY itemid, grpid, vendor) lastprices;
 DROP TABLE t1;
+
+#
+# Test for bug #7413 "Subquery with non-scalar results participating in
+# select list of derived table crashes server" aka "VIEW with sub query can
+# cause the MySQL server to crash". If we have encountered problem during
+# filling of derived table we should report error and perform cleanup
+# properly.
+#
+CREATE TABLE t1 (a char(10), b char(10));
+INSERT INTO t1 VALUES ('root','localhost'), ('root','%');
+--error 1242
+SELECT * FROM (SELECT (SELECT a.a FROM t1 AS a WHERE a.a = b.a) FROM t1 AS b) AS c;
+DROP TABLE t1;

mysql-test/t/index_merge.test

@@ -1,9 +1,8 @@
 #
 # Index merge tests
 #
-
 --disable_warnings
-drop table if exists t0, t1, t2, t3,t4;
+drop table if exists t0, t1, t2, t3, t4;
 --enable_warnings
 
 # Create and fill a table with simple keys
@@ -278,4 +277,48 @@ delete from t0 where key1 < 3 or key2 < 4;
 select * from t0 where key1 < 3 or key2 < 4;
 select count(*) from t0;
 
-drop table t0, t1, t2, t3, t4; 
+# Test for BUG#4177
+drop table t4;
+create table t4 (a int);
+insert into t4 values (1),(4),(3);
+set @save_join_buffer_size=@@join_buffer_size;
+set join_buffer_size= 4000;
+show variables like 'join_buffer_size';
+explain select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+ from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+  where (A.key1 < 500000 or A.key2 < 3)
+  and   (B.key1 < 500000 or B.key2 < 3);
+
+select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+ from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+  where (A.key1 < 500000 or A.key2 < 3)
+  and   (B.key1 < 500000 or B.key2 < 3);
+
+update t0 set key1=1;
+explain select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+ from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+  where (A.key1 = 1 or A.key2 = 1)
+  and   (B.key1 = 1 or B.key2 = 1);
+
+select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+ from t0 as A force index(i1,i2), t0 as B force index (i1,i2) 
+  where (A.key1 = 1 or A.key2 = 1)
+  and   (B.key1 = 1 or B.key2 = 1);
+
+alter table t0 add filler1 char(200), add filler2 char(200), add filler3 char(200);
+update t0 set key2=1, key3=1, key4=1, key5=1,key6=1,key7=1 where key7 < 500;
+explain select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+ from t0 as A, t0 as B 
+ where (A.key1 = 1 and A.key2 = 1 and A.key3 = 1 and A.key4=1 and A.key5=1 and A.key6=1 and A.key7 = 1 or A.key8=1)
+  and (B.key1 = 1 and B.key2 = 1 and B.key3 = 1 and B.key4=1 and B.key5=1 and B.key6=1 and B.key7 = 1 or B.key8=1);
+
+select max(A.key1 + B.key1 + A.key2 + B.key2 + A.key3 + B.key3 + A.key4 + B.key4 + A.key5 + B.key5) 
+ from t0 as A, t0 as B 
+ where (A.key1 = 1 and A.key2 = 1 and A.key3 = 1 and A.key4=1 and A.key5=1 and A.key6=1 and A.key7 = 1 or A.key8=1)
+  and (B.key1 = 1 and B.key2 = 1 and B.key3 = 1 and B.key4=1 and B.key5=1 and B.key6=1 and B.key7 = 1 or B.key8=1);
+
+set join_buffer_size= @save_join_buffer_size;
+# Test for BUG#4177 ends 
+
+drop table t0, t1, t2, t3, t4;
+

mysql-test/t/join_outer.test

@@ -606,4 +606,14 @@ INSERT INTO t1 VALUES (0);
 SELECT * FROM t0, t1 LEFT JOIN (t2,t3) ON a1=5 WHERE a0=a1 AND a0=1;
 EXPLAIN SELECT * FROM t0, t1 LEFT JOIN (t2,t3) ON a1=5 WHERE a0=a1 AND a0=1;
 
+# Test for BUG#4480
+drop table t1,t2;
+create table t1 (a int, b int);
+insert into t1 values (1,1),(2,2),(3,3);
+create table t2 (a int, b int);
+insert into t2 values (1,1), (2,2);
+
+select * from t2 right join t1 on t2.a=t1.a;
+select straight_join * from t2 right join t1 on t2.a=t1.a;
+
 DROP TABLE t0,t1,t2,t3;

ndb/include/ndbapi/NdbEventOperation.hpp

@@ -64,7 +64,7 @@ class NdbEventOperationImpl;
  *
  * Known issues:
  *
- * When several NdbEventOperation s are tied to the same event in the same
+ * When several NdbEventOperation's are tied to the same event in the same
  * process they will share the circular buffer. The BufferLength will then
  * be the same for all and decided by the first NdbEventOperation 
  * instantiation. Just make sure to instantiate the "largest" one first.
@@ -84,7 +84,7 @@ class NdbEventOperationImpl;
  * replica.  If a node fails events will not be received twice anymore
  * for data in corresponding fragment. Will be optimized in later versions.
  *
- * If a nodefailiure has occured not all events will be recieved
+ * If a node failure has occured not all events will be recieved
  * anymore. Drop NdbEventOperation and Create again after nodes are up
  * again. Will be fixed in later versions.
  *
@@ -97,7 +97,7 @@ class NdbEventOperationImpl;
  *
  * Useful API programs:
  *
- * select_all -d sys 'NDB$EVENTS_0'
+ * ndb_select_all -d sys 'NDB$EVENTS_0'
  * Will show contents in the system table containing created events.
  *
  */
@@ -187,8 +187,19 @@ public:
    */
   NdbDictionary::Event::TableEvent getEventType();
 
+  /**
+   *
+   */
   Uint32 getGCI();
+
+  /**
+   *
+   */
   Uint32 getLatestGCI();
+
+  /*
+   *
+   */
   void print();
 
 private:

server-tools/instance-manager/mysql_connection.cc

@@ -296,7 +296,6 @@ int Mysql_connection_thread::check_user(const char *user, const char *password)
 int Mysql_connection_thread::do_command()
 {
   char *packet;
-  uint old_timeout;
   ulong packet_length;
 
   /* We start to count packets from 0 for each new command */

server-tools/instance-manager/parse.cc

@@ -131,7 +131,6 @@ Command *parse_command(Command_factory *factory, const char *text)
   const char *instance_name;
   uint instance_name_len;
   Command *command;
-  const char *saved_text= text;
 
   Token tok1= shift_token(&text, &word_len);
 

server-tools/instance-manager/thread_registry.cc

@@ -148,6 +148,7 @@ int Thread_registry::cond_wait(Thread_info *info, pthread_cond_t *cond,
 
 void Thread_registry::deliver_shutdown()
 {
+  Thread_info *info;
   struct timespec shutdown_time;
   set_timespec(shutdown_time, 1);
 
@@ -161,7 +162,7 @@ void Thread_registry::deliver_shutdown()
     stopped alarm processing.
   */
   process_alarm(THR_SERVER_ALARM);
-  for (Thread_info *info= head.next; info != &head; info= info->next)
+  for (info= head.next; info != &head; info= info->next)
   {
     pthread_kill(info->thread_id, THREAD_KICK_OFF_SIGNAL);
     /*
@@ -190,7 +191,7 @@ void Thread_registry::deliver_shutdown()
     so this time everybody should be informed (presumably each worker can
     get CPU during shutdown_time.)
   */
-  for (Thread_info *info= head.next; info != &head; info= info->next)
+  for (info= head.next; info != &head; info= info->next)
   {
     pthread_kill(info->thread_id, THREAD_KICK_OFF_SIGNAL);
     if (info->current_cond)

server-tools/instance-manager/user_map.cc

@@ -123,7 +123,6 @@ int User_map::load(const char *password_file_name)
             1 +                               /* for ':' */
             1 +                               /* for newline */
             1];                               /* for trailing zero */
-  uint line_length;
   User *user;
   int rc= 1;
 

sql/ha_innodb.cc

@@ -1356,7 +1356,8 @@ innobase_commit(
 	3. innobase_query_caching_of_table_permitted(),
 	4. innobase_savepoint(),
 	5. ::init_table_handle_for_HANDLER(),
-	6. innobase_start_trx_and_assign_read_view()
+	6. innobase_start_trx_and_assign_read_view(),
+	7. ::transactional_table_lock()
 
 	and it is only set to 0 in a commit or a rollback. If it is 0 we know
 	there cannot be resources to be freed and we could return immediately.
@@ -5134,8 +5135,9 @@ ha_innobase::start_stmt(
 			select_lock_type value. The value of
 			stored_select_lock_type was decided in:
 			1) ::store_lock(),
-			2) ::external_lock(), and
+			2) ::external_lock(), 
-			3) ::init_table_handle_for_HANDLER(). */
+			3) ::init_table_handle_for_HANDLER(), and 
+			4) :.transactional_table_lock(). */
 
 			prebuilt->select_lock_type =
 				prebuilt->stored_select_lock_type;
@@ -5326,6 +5328,94 @@ ha_innobase::external_lock(
 	DBUG_RETURN(0);
 }
 
+/**********************************************************************
+With this function MySQL request a transactional lock to a table when
+user issued query LOCK TABLES..WHERE ENGINE = InnoDB. */
+
+int
+ha_innobase::transactional_table_lock(
+/*==================================*/
+			        /* out: 0 */
+	THD*	thd,		/* in: handle to the user thread */
+	int 	lock_type)	/* in: lock type */
+{
+	row_prebuilt_t* prebuilt = (row_prebuilt_t*) innobase_prebuilt;
+	trx_t*		trx;
+
+  	DBUG_ENTER("ha_innobase::transactional_table_lock");
+	DBUG_PRINT("enter",("lock_type: %d", lock_type));
+
+	/* We do not know if MySQL can call this function before calling
+	external_lock(). To be safe, update the thd of the current table
+	handle. */
+
+	update_thd(thd);
+
+ 	if (prebuilt->table->ibd_file_missing && !current_thd->tablespace_op) {
+	        ut_print_timestamp(stderr);
+	        fprintf(stderr, "  InnoDB error:\n"
+"MySQL is trying to use a table handle but the .ibd file for\n"
+"table %s does not exist.\n"
+"Have you deleted the .ibd file from the database directory under\n"
+"the MySQL datadir?"
+"Look from section 15.1 of http://www.innodb.com/ibman.html\n"
+"how you can resolve the problem.\n",
+				prebuilt->table->name);
+		DBUG_RETURN(HA_ERR_CRASHED);
+	}
+
+	trx = prebuilt->trx;
+
+	prebuilt->sql_stat_start = TRUE;
+	prebuilt->hint_need_to_fetch_extra_cols = 0;
+
+	prebuilt->read_just_key = 0;
+	prebuilt->keep_other_fields_on_keyread = FALSE;
+
+	if (lock_type == F_WRLCK) {
+		prebuilt->select_lock_type = LOCK_X;
+		prebuilt->stored_select_lock_type = LOCK_X;
+	} else if (lock_type == F_RDLCK) {
+		prebuilt->select_lock_type = LOCK_S;
+		prebuilt->stored_select_lock_type = LOCK_S;
+	} else {
+	        ut_print_timestamp(stderr);
+	        fprintf(stderr, "  InnoDB error:\n"
+"MySQL is trying to set transactional table lock with corrupted lock type\n"
+"to table %s, lock type %d does not exist.\n",
+				prebuilt->table->name, lock_type);
+		DBUG_RETURN(HA_ERR_CRASHED);
+	}
+
+	/* MySQL is setting a new transactional table lock */
+
+	/* Set the MySQL flag to mark that there is an active transaction */
+	thd->transaction.all.innodb_active_trans = 1;
+
+	if (thd->in_lock_tables && thd->variables.innodb_table_locks) {
+		ulint	error = DB_SUCCESS;
+
+		error = row_lock_table_for_mysql(prebuilt,NULL, 
+						LOCK_TABLE_TRANSACTIONAL);
+
+		if (error != DB_SUCCESS) {
+			error = convert_error_code_to_mysql(error, user_thd);
+			DBUG_RETURN(error);
+		}
+
+		if (thd->options & (OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) {
+
+			/* Store the current undo_no of the transaction 
+			so that we know where to roll back if we have 
+			to roll back the next SQL statement */
+
+			trx_mark_sql_stat_end(trx);
+		}
+	}
+
+	DBUG_RETURN(0);
+}
+
 /****************************************************************************
 Here we export InnoDB status variables to MySQL.  */
 

sql/ha_innodb.h

@@ -150,6 +150,7 @@ class ha_innobase: public handler
 	int discard_or_import_tablespace(my_bool discard);
   	int extra(enum ha_extra_function operation);
   	int external_lock(THD *thd, int lock_type);
+	int transactional_table_lock(THD *thd, int lock_type);
 	int start_stmt(THD *thd);
 
   	void position(byte *record);

sql/handler.h

@@ -454,6 +454,13 @@ public:
   { return extra(operation); }
   virtual int reset() { return extra(HA_EXTRA_RESET); }
   virtual int external_lock(THD *thd, int lock_type) { return 0; }
+  /*
+    This is called to set transactional table lock to a table.
+    If the handler don't support this, then this function will
+    return HA_ERR_WRONG_COMMAND and MySQL will give
+    ER_ILLEGAL_HA error message.
+  */
+  virtual int transactional_table_lock(THD *thd, int lock_type) {return HA_ERR_WRONG_COMMAND;}
   virtual void unlock_row() {}
   virtual int start_stmt(THD *thd) {return 0;}
   /*

sql/lock.cc

@@ -79,7 +79,7 @@ static MYSQL_LOCK *get_lock_data(THD *thd, TABLE **table,uint count,
 				 bool unlock, TABLE **write_locked);
 static int lock_external(THD *thd, TABLE **table,uint count);
 static int unlock_external(THD *thd, TABLE **table,uint count);
-static void print_lock_error(int error);
+static void print_lock_error(int error, const char *);
 
 
 MYSQL_LOCK *mysql_lock_tables(THD *thd,TABLE **tables,uint count)
@@ -187,7 +187,7 @@ static int lock_external(THD *thd, TABLE **tables, uint count)
 	(*tables)->file->external_lock(thd, F_UNLCK);
 	(*tables)->current_lock=F_UNLCK;
       }
-      print_lock_error(error);
+      print_lock_error(error, (*tables)->file->table_type());
       DBUG_RETURN(error);
     }
     else
@@ -380,7 +380,7 @@ static int unlock_external(THD *thd, TABLE **table,uint count)
     table++;
   } while (--count);
   if (error_code)
-    print_lock_error(error_code);
+    print_lock_error(error_code, (*table)->file->table_type());
   DBUG_RETURN(error_code);
 }
 
@@ -683,7 +683,7 @@ void unlock_table_names(THD *thd, TABLE_LIST *table_list,
 }
 
 
-static void print_lock_error(int error)
+static void print_lock_error(int error, const char *table)
 {
   int textno;
   DBUG_ENTER("print_lock_error");
@@ -695,11 +695,22 @@ static void print_lock_error(int error)
   case HA_ERR_READ_ONLY_TRANSACTION:
     textno=ER_READ_ONLY_TRANSACTION;
     break;
+  case HA_ERR_LOCK_DEADLOCK:
+    textno=ER_LOCK_DEADLOCK;
+    break;
+  case HA_ERR_WRONG_COMMAND:
+    textno=ER_ILLEGAL_HA;
+    break;
   default:
     textno=ER_CANT_LOCK;
     break;
   }
-  my_error(textno,MYF(ME_BELL+ME_OLDWIN+ME_WAITTANG),error);
+
+  if ( textno == ER_ILLEGAL_HA )
+    my_error(textno, MYF(ME_BELL+ME_OLDWIN+ME_WAITTANG), table);
+  else
+    my_error(textno, MYF(ME_BELL+ME_OLDWIN+ME_WAITTANG), error);
+
   DBUG_VOID_RETURN;
 }
 
@@ -927,3 +938,62 @@ bool make_global_read_lock_block_commit(THD *thd)
   thd->exit_cond(old_message);
   DBUG_RETURN(error);
 }
+
+/*
+  Take transactional table lock for all tables in the list
+
+  SYNOPSIS
+    transactional_lock_tables
+    thd			Thread THD
+    tables              list of tables
+    counter             number of tables in the list
+
+  NOTES
+
+  RETURN
+    0  - OK
+    -1 - error
+
+*/
+int transactional_lock_tables(THD *thd, TABLE_LIST *tables, uint counter)
+{ 
+  uint i;
+  int lock_type,error=0;
+  TABLE_LIST *table;
+  TABLE **start,**ptr;
+
+  DBUG_ENTER("transactional_lock_tables");
+
+  if (!(ptr=start=(TABLE**) sql_alloc(sizeof(TABLE*) * counter)))
+      return -1;
+    
+  for (table= tables; table; table= table->next_global)
+  {
+    if (!table->placeholder() && !table->schema_table)
+      *(ptr++)= table->table;
+  }
+
+  for (i=1 ; i <= counter ; i++, start++)
+  {
+    DBUG_ASSERT((*start)->reginfo.lock_type >= TL_READ);
+    lock_type=F_WRLCK;				/* Lock exclusive */
+
+    if ((*start)->db_stat & HA_READ_ONLY ||
+       ((*start)->reginfo.lock_type >= TL_READ &&
+	(*start)->reginfo.lock_type <= TL_READ_NO_INSERT))
+      lock_type=F_RDLCK;
+
+    if ((error=(*start)->file->transactional_table_lock(thd, lock_type)))
+    {
+      print_lock_error(error, (*start)->file->table_type());
+      DBUG_RETURN(-1);
+    }
+    else
+    {
+      (*start)->db_stat &= ~ HA_BLOCK_LOCK;
+      (*start)->current_lock= lock_type;
+    }
+  }
+
+  DBUG_RETURN(0);
+}

sql/mysql_priv.h

@@ -828,6 +828,7 @@ int open_tables(THD *thd, TABLE_LIST *tables, uint *counter);
 int simple_open_n_lock_tables(THD *thd,TABLE_LIST *tables);
 bool open_and_lock_tables(THD *thd,TABLE_LIST *tables);
 int lock_tables(THD *thd, TABLE_LIST *tables, uint counter);
+int transactional_lock_tables(THD *thd, TABLE_LIST *tables, uint counter);
 TABLE *open_temporary_table(THD *thd, const char *path, const char *db,
 			    const char *table_name, bool link_in_list);
 bool rm_temporary_table(enum db_type base, char *path);

sql/opt_range.cc

@@ -777,8 +777,7 @@ QUICK_RANGE_SELECT::~QUICK_RANGE_SELECT()
 
 QUICK_INDEX_MERGE_SELECT::QUICK_INDEX_MERGE_SELECT(THD *thd_param,
                                                    TABLE *table)
-  :cur_quick_it(quick_selects),pk_quick_select(NULL),unique(NULL),
+  :pk_quick_select(NULL), thd(thd_param)
-   thd(thd_param)
 {
   DBUG_ENTER("QUICK_INDEX_MERGE_SELECT::QUICK_INDEX_MERGE_SELECT");
   index= MAX_KEY;
@@ -790,17 +789,14 @@ QUICK_INDEX_MERGE_SELECT::QUICK_INDEX_MERGE_SELECT(THD *thd_param,
 
 int QUICK_INDEX_MERGE_SELECT::init()
 {
-  cur_quick_it.rewind();
+  DBUG_ENTER("QUICK_INDEX_MERGE_SELECT::init");
-  cur_quick_select= cur_quick_it++;
+  DBUG_RETURN(0);
-  return 0;
 }
 
 int QUICK_INDEX_MERGE_SELECT::reset()
 {
-  int result;
   DBUG_ENTER("QUICK_INDEX_MERGE_SELECT::reset");
-  result= cur_quick_select->reset() || prepare_unique();
+  DBUG_RETURN(read_keys_and_merge());
-  DBUG_RETURN(result);
 }
 
 bool
@@ -820,8 +816,12 @@ QUICK_INDEX_MERGE_SELECT::push_quick_back(QUICK_RANGE_SELECT *quick_sel_range)
 
 QUICK_INDEX_MERGE_SELECT::~QUICK_INDEX_MERGE_SELECT()
 {
+  List_iterator_fast<QUICK_RANGE_SELECT> quick_it(quick_selects);
+  QUICK_RANGE_SELECT* quick;
   DBUG_ENTER("QUICK_INDEX_MERGE_SELECT::~QUICK_INDEX_MERGE_SELECT");
-  delete unique;
+  quick_it.rewind();
+  while ((quick= quick_it++))
+    quick->file= NULL;
   quick_selects.delete_elements();
   delete pk_quick_select;
   free_root(&alloc,MYF(0));
@@ -833,7 +833,8 @@ QUICK_ROR_INTERSECT_SELECT::QUICK_ROR_INTERSECT_SELECT(THD *thd_param,
                                                        TABLE *table,
                                                        bool retrieve_full_rows,
                                                        MEM_ROOT *parent_alloc)
-  : cpk_quick(NULL), thd(thd_param), need_to_fetch_row(retrieve_full_rows)
+  : cpk_quick(NULL), thd(thd_param), need_to_fetch_row(retrieve_full_rows),
+    scans_inited(false)
 {
   index= MAX_KEY;
   head= table;
@@ -859,8 +860,9 @@ QUICK_ROR_INTERSECT_SELECT::QUICK_ROR_INTERSECT_SELECT(THD *thd_param,
 
 int QUICK_ROR_INTERSECT_SELECT::init()
 {
-  /* Check if last_rowid was successfully allocated in ctor */
+  DBUG_ENTER("QUICK_ROR_INTERSECT_SELECT::init");
-  return !last_rowid;
+ /* Check if last_rowid was successfully allocated in ctor */
+  DBUG_RETURN(!last_rowid);
 }
 
 
@@ -953,7 +955,7 @@ int QUICK_ROR_INTERSECT_SELECT::init_ror_merged_scan(bool reuse_handler)
   DBUG_ENTER("QUICK_ROR_INTERSECT_SELECT::init_ror_merged_scan");
 
   /* Initialize all merged "children" quick selects */
-  DBUG_ASSERT(!(need_to_fetch_row && !reuse_handler));
+  DBUG_ASSERT(!need_to_fetch_row || reuse_handler);
   if (!need_to_fetch_row && reuse_handler)
   {
     quick= quick_it++;
@@ -995,7 +997,14 @@ int QUICK_ROR_INTERSECT_SELECT::init_ror_merged_scan(bool reuse_handler)
 int QUICK_ROR_INTERSECT_SELECT::reset()
 {
   DBUG_ENTER("QUICK_ROR_INTERSECT_SELECT::reset");
-  DBUG_RETURN(init_ror_merged_scan(TRUE));
+  if (!scans_inited && init_ror_merged_scan(TRUE))
+    DBUG_RETURN(1);
+  scans_inited= true;
+  List_iterator_fast<QUICK_RANGE_SELECT> it(quick_selects);
+  QUICK_RANGE_SELECT *quick;
+  while ((quick= it++))
+    quick->reset();
+  DBUG_RETURN(0);
 }
 
 
@@ -1034,7 +1043,7 @@ QUICK_ROR_INTERSECT_SELECT::~QUICK_ROR_INTERSECT_SELECT()
 
 QUICK_ROR_UNION_SELECT::QUICK_ROR_UNION_SELECT(THD *thd_param,
                                                TABLE *table)
-  :thd(thd_param)
+  : thd(thd_param), scans_inited(false)
 {
   index= MAX_KEY;
   head= table;
@@ -1057,18 +1066,19 @@ QUICK_ROR_UNION_SELECT::QUICK_ROR_UNION_SELECT(THD *thd_param,
 
 int QUICK_ROR_UNION_SELECT::init()
 {
+  DBUG_ENTER("QUICK_ROR_UNION_SELECT::init");
   if (init_queue(&queue, quick_selects.elements, 0,
                  FALSE , QUICK_ROR_UNION_SELECT::queue_cmp,
                  (void*) this))
   {
     bzero(&queue, sizeof(QUEUE));
-    return 1;
+    DBUG_RETURN(1);
   }
 
   if (!(cur_rowid= (byte*)alloc_root(&alloc, 2*head->file->ref_length)))
-    return 1;
+    DBUG_RETURN(1);
   prev_rowid= cur_rowid + head->file->ref_length;
-  return 0;
+  DBUG_RETURN(0);
 }
 
 
@@ -1106,6 +1116,18 @@ int QUICK_ROR_UNION_SELECT::reset()
   int error;
   DBUG_ENTER("QUICK_ROR_UNION_SELECT::reset");
   have_prev_rowid= FALSE;
+  if (!scans_inited)
+  {
+    QUICK_SELECT_I *quick;
+    List_iterator_fast<QUICK_SELECT_I> it(quick_selects);
+    while ((quick= it++))
+    {
+      if (quick->init_ror_merged_scan(FALSE))
+        DBUG_RETURN(1);
+    }
+    scans_inited= true;
+  }
+  queue_remove_all(&queue);
   /*
     Initialize scans for merged quick selects and put all merged quick
     selects into the queue.
@@ -1113,7 +1135,7 @@ int QUICK_ROR_UNION_SELECT::reset()
   List_iterator_fast<QUICK_SELECT_I> it(quick_selects);
   while ((quick= it++))
   {
-    if (quick->init_ror_merged_scan(FALSE))
+    if (quick->reset())
       DBUG_RETURN(1);
     if ((error= quick->get_next()))
     {
@@ -1591,7 +1613,6 @@ int SQL_SELECT::test_quick_select(THD *thd, key_map keys_to_use,
   DBUG_PRINT("enter",("keys_to_use: %lu  prev_tables: %lu  const_tables: %lu",
 		      keys_to_use.to_ulonglong(), (ulong) prev_tables,
 		      (ulong) const_tables));
-
   delete quick;
   quick=0;
   needed_reg.clear_all();
@@ -5553,22 +5574,29 @@ err:
 
 
 /*
-  Fetch all row ids into unique.
+  Perform key scans for all used indexes (except CPK), get rowids and merge 
-
+  them into an ordered non-recurrent sequence of rowids.
+  
+  The merge/duplicate removal is performed using Unique class. We put all
+  rowids into Unique, get the sorted sequence and destroy the Unique.
+  
   If table has a clustered primary key that covers all rows (TRUE for bdb
      and innodb currently) and one of the index_merge scans is a scan on PK,
   then
-    primary key scan rowids are not put into Unique and also
+    rows that will be retrieved by PK scan are not put into Unique and 
-    rows that will be retrieved by PK scan are not put into Unique
+    primary key scan is not performed here, it is performed later separately.
 
   RETURN
     0     OK
     other error
 */
 
-int QUICK_INDEX_MERGE_SELECT::prepare_unique()
+int QUICK_INDEX_MERGE_SELECT::read_keys_and_merge()
 {
+  List_iterator_fast<QUICK_RANGE_SELECT> cur_quick_it(quick_selects);
+  QUICK_RANGE_SELECT* cur_quick;
   int result;
+  Unique *unique;
   DBUG_ENTER("QUICK_INDEX_MERGE_SELECT::prepare_unique");
 
   /* We're going to just read rowids. */
@@ -5583,7 +5611,17 @@ int QUICK_INDEX_MERGE_SELECT::prepare_unique()
   */
   head->file->extra(HA_EXTRA_RETRIEVE_ALL_COLS);
 
-  cur_quick_select->init();
+  cur_quick_it.rewind();
+  cur_quick= cur_quick_it++;
+  DBUG_ASSERT(cur_quick);
+  
+  /*
+    We reuse the same instance of handler so we need to call both init and 
+    reset here.
+  */
+  if (cur_quick->init())
+    DBUG_RETURN(1);
+  cur_quick->reset();
 
   unique= new Unique(refpos_order_cmp, (void *)head->file,
                      head->file->ref_length,
@@ -5592,24 +5630,28 @@ int QUICK_INDEX_MERGE_SELECT::prepare_unique()
     DBUG_RETURN(1);
   for (;;)
   {
-    while ((result= cur_quick_select->get_next()) == HA_ERR_END_OF_FILE)
+    while ((result= cur_quick->get_next()) == HA_ERR_END_OF_FILE)
     {
-      cur_quick_select->range_end();
+      cur_quick->range_end();
-      cur_quick_select= cur_quick_it++;
+      cur_quick= cur_quick_it++;
-      if (!cur_quick_select)
+      if (!cur_quick)
         break;
 
-      if (cur_quick_select->init())
+      if (cur_quick->file->inited != handler::NONE) 
+        cur_quick->file->ha_index_end();
+      if (cur_quick->init())
         DBUG_RETURN(1);
-
       /* QUICK_RANGE_SELECT::reset never fails */
-      cur_quick_select->reset();
+      cur_quick->reset();
     }
 
     if (result)
     {
       if (result != HA_ERR_END_OF_FILE)
+      {
+        cur_quick->range_end();
         DBUG_RETURN(result);
+      }
       break;
     }
 
@@ -5620,8 +5662,8 @@ int QUICK_INDEX_MERGE_SELECT::prepare_unique()
     if (pk_quick_select && pk_quick_select->row_in_ranges())
       continue;
 
-    cur_quick_select->file->position(cur_quick_select->record);
+    cur_quick->file->position(cur_quick->record);
-    result= unique->unique_add((char*)cur_quick_select->file->ref);
+    result= unique->unique_add((char*)cur_quick->file->ref);
     if (result)
       DBUG_RETURN(1);
 
@@ -5629,6 +5671,7 @@ int QUICK_INDEX_MERGE_SELECT::prepare_unique()
 
   /* ok, all row ids are in Unique */
   result= unique->get(head);
+  delete unique;
   doing_pk_scan= FALSE;
   /* start table scan */
   init_read_record(&read_record, thd, head, (SQL_SELECT*) 0, 1, 1);
@@ -5668,6 +5711,7 @@ int QUICK_INDEX_MERGE_SELECT::get_next()
       doing_pk_scan= TRUE;
       if ((result= pk_quick_select->init()))
         DBUG_RETURN(result);
+      pk_quick_select->reset();
       DBUG_RETURN(pk_quick_select->get_next());
     }
   }

sql/opt_range.h

@@ -127,7 +127,8 @@ public:
     reset() should be called when it is certain that row retrieval will be
     necessary. This call may do heavyweight initialization like buffering first
     N records etc. If reset() call fails get_next() must not be called.
-
+    Note that reset() may be called several times if this quick select 
+    executes in a subselect.
     RETURN
       0      OK
       other  Error code
@@ -274,6 +275,10 @@ public:
     next=0;
     range= NULL;
     cur_range= NULL;
+    /*
+      Note: in opt_range.cc there are places where it is assumed that this
+      function always succeeds 
+    */
     return 0;
   }
   int init();
@@ -388,21 +393,15 @@ public:
   /* range quick selects this index_merge read consists of */
   List<QUICK_RANGE_SELECT> quick_selects;
 
-  /* quick select which is currently used for rows retrieval */
-  List_iterator_fast<QUICK_RANGE_SELECT> cur_quick_it;
-  QUICK_RANGE_SELECT* cur_quick_select;
-
   /* quick select that uses clustered primary key (NULL if none) */
   QUICK_RANGE_SELECT* pk_quick_select;
 
   /* true if this select is currently doing a clustered PK scan */
   bool  doing_pk_scan;
 
-  Unique  *unique;
   MEM_ROOT alloc;
-
   THD *thd;
-  int prepare_unique();
+  int read_keys_and_merge();
 
   /* used to get rows collected in Unique */
   READ_RECORD read_record;
@@ -465,6 +464,8 @@ public:
   MEM_ROOT alloc; /* Memory pool for this and merged quick selects data. */
   THD *thd;       /* current thread */
   bool need_to_fetch_row; /* if true, do retrieve full table records. */
+  /* in top-level quick select, true if merged scans where initialized */
+  bool scans_inited; 
 };
 
 
@@ -514,6 +515,7 @@ public:
   uint rowid_length;    /* table rowid length */
 private:
   static int queue_cmp(void *arg, byte *val1, byte *val2);
+  bool scans_inited; 
 };
 
 

sql/sql_base.cc

@@ -1864,8 +1864,7 @@ bool open_and_lock_tables(THD *thd, TABLE_LIST *tables)
 
 static void relink_tables_for_multidelete(THD *thd)
 {
-  if (thd->lex->all_selects_list->next_select_in_list() ||
+  if (thd->lex->all_selects_list->next_select_in_list())
-      thd->lex->time_zone_tables_used)
   {
     for (SELECT_LEX *sl= thd->lex->all_selects_list;
 	 sl;

sql/sql_lex.cc

@@ -30,15 +30,6 @@
 */
 sys_var_long_ptr trg_new_row_fake_var(0, 0);
 
-/*
-  Fake table list object, pointer to which is used as special value for
-  st_lex::time_zone_tables_used indicating that we implicitly use time
-  zone tables in this statement but real table list was not yet created.
-  Pointer to it is also returned by my_tz_get_tables_list() as indication
-  of transient error;
-*/
-TABLE_LIST fake_time_zone_tables_list;
-
 /* Macros to look like lex */
 
 #define yyGet()		*(lex->ptr++)
@@ -1911,6 +1902,31 @@ void st_lex::first_lists_tables_same()
 }
 
 
+/*
+  Add implicitly used time zone description tables to global table list
+  (if needed).
+
+  SYNOPSYS
+    st_lex::add_time_zone_tables_to_query_tables()
+      thd - pointer to current thread context
+
+  RETURN VALUE
+   TRUE  - error
+   FALSE - success
+*/
+
+bool st_lex::add_time_zone_tables_to_query_tables(THD *thd)
+{
+  /* We should not add these tables twice */
+  if (!time_zone_tables_used)
+  {
+    time_zone_tables_used= my_tz_get_table_list(thd, &query_tables_last);
+    if (time_zone_tables_used == &fake_time_zone_tables_list)
+      return TRUE;
+  }
+  return FALSE;
+}
+
 /*
   Link table back that was unlinked with unlink_first_table()
 

sql/sql_lex.h

@@ -87,6 +87,7 @@ enum enum_sql_command {
   SQLCOM_PREPARE, SQLCOM_EXECUTE, SQLCOM_DEALLOCATE_PREPARE,
   SQLCOM_CREATE_VIEW, SQLCOM_DROP_VIEW,
   SQLCOM_CREATE_TRIGGER, SQLCOM_DROP_TRIGGER,
+  SQLCOM_LOCK_TABLES_TRANSACTIONAL,
   /* This should be the last !!! */
   SQLCOM_END
 };
@@ -735,9 +736,8 @@ typedef struct st_lex
   /* Names of user variables holding parameters (in EXECUTE) */
   List<LEX_STRING> prepared_stmt_params; 
   /*
-    If points to fake_time_zone_tables_list indicates that time zone
+    Points to part of global table list which contains time zone tables
-    tables are implicitly used by statement, also is used for holding
+    implicitly used by the statement.
-    list of those tables after they are opened.
   */
   TABLE_LIST *time_zone_tables_used;
   sp_head *sphead;
@@ -802,6 +802,7 @@ typedef struct st_lex
     *(table->prev_global= query_tables_last)= table;
     query_tables_last= &table->next_global;
   }
+  bool add_time_zone_tables_to_query_tables(THD *thd);
 
   bool can_be_merged();
   bool can_use_merged();
@@ -810,7 +811,6 @@ typedef struct st_lex
   bool need_correct_ident();
 } LEX;
 
-extern TABLE_LIST fake_time_zone_tables_list;
 struct st_lex_local: public st_lex
 {
   static void *operator new(size_t size)

sql/sql_parse.cc

@@ -2123,19 +2123,6 @@ mysql_execute_command(THD *thd)
   }
 #endif /* !HAVE_REPLICATION */
 
-  if (lex->time_zone_tables_used)
-  {
-    TABLE_LIST *tmp;
-    if ((tmp= my_tz_get_table_list(thd, &lex->query_tables_last)) ==
-        &fake_time_zone_tables_list)
-    {
-      DBUG_RETURN(-1);
-    }
-    lex->time_zone_tables_used= tmp;
-    if (!all_tables)
-      all_tables= tmp;
-  }
-
   /*
     When option readonly is set deny operations which change tables.
     Except for the replication thread and the 'super' users.
@@ -3266,6 +3253,27 @@ create_error:
       thd->options&= ~(ulong) (OPTION_TABLE_LOCK);
     thd->in_lock_tables=0;
     break;
+  case SQLCOM_LOCK_TABLES_TRANSACTIONAL:
+  {
+    uint counter = 0;
+
+    if (check_db_used(thd, all_tables))
+      goto error;
+    if (check_table_access(thd, LOCK_TABLES_ACL | SELECT_ACL, all_tables, 0))
+      goto error;
+
+    thd->in_lock_tables=1;
+    thd->options|= OPTION_TABLE_LOCK;
+
+    if (open_tables(thd, all_tables, &counter) == 0 &&
+        transactional_lock_tables(thd, all_tables, counter) == 0)
+        send_ok(thd);
+    else
+      thd->options&= ~(ulong) (OPTION_TABLE_LOCK);
+
+    thd->in_lock_tables=0;
+    break;
+  }
   case SQLCOM_CREATE_DB:
   {
     char *alias;

sql/sql_select.cc

@@ -64,6 +64,7 @@ static void best_extension_by_limited_search(JOIN *join,
                                              uint prune_level);
 static uint determine_search_depth(JOIN* join);
 static int join_tab_cmp(const void* ptr1, const void* ptr2);
+static int join_tab_cmp_straight(const void* ptr1, const void* ptr2);
 /*
   TODO: 'find_best' is here only temporarily until 'greedy_search' is
   tested and approved.
@@ -3678,22 +3679,26 @@ choose_plan(JOIN *join, table_map join_tables)
 {
   uint search_depth= join->thd->variables.optimizer_search_depth;
   uint prune_level=  join->thd->variables.optimizer_prune_level;
-
+  bool straight_join= join->select_options & SELECT_STRAIGHT_JOIN;
   DBUG_ENTER("choose_plan");
 
-  if (join->select_options & SELECT_STRAIGHT_JOIN)
+  /*
+    if (SELECT_STRAIGHT_JOIN option is set)
+      reorder tables so dependent tables come after tables they depend 
+      on, otherwise keep tables in the order they were specified in the query 
+    else
+      Apply heuristic: pre-sort all access plans with respect to the number of
+      records accessed.
+  */
+  qsort(join->best_ref + join->const_tables, join->tables - join->const_tables,
+        sizeof(JOIN_TAB*), straight_join?join_tab_cmp_straight:join_tab_cmp);
+  
+  if (straight_join)
   {
     optimize_straight_join(join, join_tables);
   }
   else
   {
-    /*
-      Heuristic: pre-sort all access plans with respect to the number of
-      records accessed.
-    */
-    qsort(join->best_ref + join->const_tables, join->tables - join->const_tables,
-          sizeof(JOIN_TAB*), join_tab_cmp);
-
     if (search_depth == MAX_TABLES+2)
     { /*
         TODO: 'MAX_TABLES+2' denotes the old implementation of find_best before
@@ -3750,6 +3755,23 @@ join_tab_cmp(const void* ptr1, const void* ptr2)
 }
 
 
+/* 
+  Same as join_tab_cmp, but for use with SELECT_STRAIGHT_JOIN.
+*/
+
+static int
+join_tab_cmp_straight(const void* ptr1, const void* ptr2)
+{
+  JOIN_TAB *jt1= *(JOIN_TAB**) ptr1;
+  JOIN_TAB *jt2= *(JOIN_TAB**) ptr2;
+
+  if (jt1->dependent & jt2->table->map)
+    return 1;
+  if (jt2->dependent & jt1->table->map)
+    return -1;
+  return jt1 > jt2 ? 1 : (jt1 < jt2 ? -1 : 0);
+}
+
 /*
   Heuristic procedure to automatically guess a reasonable degree of
   exhaustiveness for the greedy search procedure.
@@ -3832,7 +3854,7 @@ optimize_straight_join(JOIN *join, table_map join_tables)
   uint idx= join->const_tables;
   double    record_count= 1.0;
   double    read_time=    0.0;
-
+ 
   for (JOIN_TAB **pos= join->best_ref + idx ; (s= *pos) ; pos++)
   {
     /* Find the best access method from 's' to the current partial plan */

sql/sql_yacc.yy

@@ -4269,7 +4269,8 @@ simple_expr:
 	  { $$= create_func_contains($3, $5); }
 	| CONVERT_TZ_SYM '(' expr ',' expr ',' expr ')'
 	  {
-	    Lex->time_zone_tables_used= &fake_time_zone_tables_list;
+            if (Lex->add_time_zone_tables_to_query_tables(YYTHD))
+              YYABORT;
 	    $$= new Item_func_convert_tz($3, $5, $7);
 	  }
 	| CURDATE optional_braces
@@ -7307,8 +7308,9 @@ internal_variable_name:
               If this is time_zone variable we should open time zone
               describing tables 
             */
-            if (tmp == &sys_time_zone)
+            if (tmp == &sys_time_zone &&
-	      Lex->time_zone_tables_used= &fake_time_zone_tables_list;
+                lex->add_time_zone_tables_to_query_tables(YYTHD))
+              YYABORT;
 	  }
 	  else
 	  {
@@ -7413,8 +7415,8 @@ lock:
 	{
 	  Lex->sql_command=SQLCOM_LOCK_TABLES;
 	}
-	table_lock_list
+	table_lock_list lock_engine_opt
-	{}
+	{} 
 	;
 
 table_or_tables:
@@ -7440,6 +7442,15 @@ lock_option:
 	| READ_SYM LOCAL_SYM { $$= TL_READ; }
         ;
 
+lock_engine_opt:
+	/* empty */	
+	| WHERE 
+	{
+	  Lex->sql_command=SQLCOM_LOCK_TABLES_TRANSACTIONAL;
+	}
+	ENGINE_SYM opt_equal storage_engines
+	;
+
 unlock:
 	UNLOCK_SYM table_or_tables { Lex->sql_command=SQLCOM_UNLOCK_TABLES; }
         ;

sql/tztime.cc

@@ -1439,6 +1439,12 @@ tz_init_table_list(TABLE_LIST *tz_tabs, TABLE_LIST ***global_next_ptr)
 }
 
 
+/*
+  Fake table list object, pointer to which is returned by
+  my_tz_get_tables_list() as indication of error.
+*/
+TABLE_LIST fake_time_zone_tables_list;
+
 /*
   Create table list with time zone related tables and add it to the end
   of global table list.

sql/tztime.h

@@ -64,6 +64,7 @@ extern Time_zone * my_tz_find(const String *name, TABLE_LIST *tz_tables);
 extern my_bool     my_tz_init(THD *org_thd, const char *default_tzname, my_bool bootstrap);
 extern void        my_tz_free();
 
+extern TABLE_LIST fake_time_zone_tables_list;
 
 /*
   Check if we have pointer to the beggining of list of implictly used