Bug#56423: Different count with SELECT and CREATE SELECT queries

This is a regression from the fix for bug no 38999. A storage engine capable
of reading only a subset of a table's columns updates corresponding bits in
the read buffer to signal that it has read NULL values for the corresponding
columns. It cannot, and should not, update any other bits. Bug no 38999
occurred because the implementation of UPDATE statements compare the NULL bits
using memcmp, inadvertently comparing bits that were never requested from the
storage engine. The regression was caused by the storage engine trying to
alleviate the situation by writing to all NULL bits, even those that it had no
knowledge of. This has devastating effects for the index merge algorithm,
which relies on all NULL bits, except those explicitly requested, being left
unchanged.

The fix reverts the fix for bug no 38999 in both InnoDB and InnoDB plugin and
changes the server's method of comparing records. For engines that always read
entire rows, we proceed as usual. For engines capable of reading only select
columns, the record buffers are now compared on a column by column basis. An
assertion was also added so that non comparable buffers are never read. Some
relevant copy-pasted code was also consolidated in a new function.

mysql-test/include/index_merge2.inc

@@ -343,3 +343,55 @@ explain select * from t1 where (key3 > 30 and key3<35) or (key2 >32 and key2 < 4
 select * from t1 where (key3 > 30 and key3<35) or (key2 >32 and key2 < 40);
 drop table t1;
 
+--echo #
+--echo # Bug#56423: Different count with SELECT and CREATE SELECT queries
+--echo #
+
+CREATE TABLE t1 (
+  a INT,
+  b INT,
+  c INT,
+  d INT,
+  PRIMARY KEY (a),
+  KEY (c),
+  KEY bd (b,d)
+);
+
+INSERT INTO t1 VALUES
+(1, 0, 1, 0),
+(2, 1, 1, 1),
+(3, 1, 1, 1),
+(4, 0, 1, 1);
+
+EXPLAIN
+SELECT a
+FROM t1
+WHERE c = 1 AND b = 1 AND d = 1;
+
+CREATE TABLE t2 ( a INT )
+SELECT a
+FROM t1
+WHERE c = 1 AND b = 1 AND d = 1;
+
+SELECT * FROM t2;
+
+DROP TABLE t1, t2;
+
+CREATE TABLE t1( a INT, b INT, KEY(a), KEY(b) );
+INSERT INTO t1 VALUES (1, 2), (1, 2), (1, 2), (1, 2);
+SELECT * FROM t1 FORCE INDEX(a, b) WHERE a = 1 AND b = 2;
+
+DROP TABLE t1;
+
+--echo # Code coverage of fix.
+CREATE TABLE t1 ( a INT NOT NULL AUTO_INCREMENT PRIMARY KEY, b INT);
+INSERT INTO t1 (b) VALUES (1);
+UPDATE t1 SET b = 2 WHERE a = 1;
+SELECT * FROM t1;
+
+CREATE TABLE t2 ( a INT NOT NULL AUTO_INCREMENT PRIMARY KEY, b VARCHAR(1) );
+INSERT INTO t2 (b) VALUES ('a');
+UPDATE t2 SET b = 'b' WHERE a = 1;
+SELECT * FROM t2;
+
+DROP TABLE t1, t2;

mysql-test/r/index_merge_innodb.result

@@ -324,6 +324,61 @@ key1	key2	key3
 38	38	38
 39	39	39
 drop table t1;
+#
+# Bug#56423: Different count with SELECT and CREATE SELECT queries
+#
+CREATE TABLE t1 (
+a INT,
+b INT,
+c INT,
+d INT,
+PRIMARY KEY (a),
+KEY (c),
+KEY bd (b,d)
+);
+INSERT INTO t1 VALUES
+(1, 0, 1, 0),
+(2, 1, 1, 1),
+(3, 1, 1, 1),
+(4, 0, 1, 1);
+EXPLAIN
+SELECT a
+FROM t1
+WHERE c = 1 AND b = 1 AND d = 1;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1	index_merge	c,bd	c,bd	5,10	NULL	1	Using intersect(c,bd); Using where; Using index
+CREATE TABLE t2 ( a INT )
+SELECT a
+FROM t1
+WHERE c = 1 AND b = 1 AND d = 1;
+SELECT * FROM t2;
+a
+2
+3
+DROP TABLE t1, t2;
+CREATE TABLE t1( a INT, b INT, KEY(a), KEY(b) );
+INSERT INTO t1 VALUES (1, 2), (1, 2), (1, 2), (1, 2);
+SELECT * FROM t1 FORCE INDEX(a, b) WHERE a = 1 AND b = 2;
+a	b
+1	2
+1	2
+1	2
+1	2
+DROP TABLE t1;
+# Code coverage of fix.
+CREATE TABLE t1 ( a INT NOT NULL AUTO_INCREMENT PRIMARY KEY, b INT);
+INSERT INTO t1 (b) VALUES (1);
+UPDATE t1 SET b = 2 WHERE a = 1;
+SELECT * FROM t1;
+a	b
+1	2
+CREATE TABLE t2 ( a INT NOT NULL AUTO_INCREMENT PRIMARY KEY, b VARCHAR(1) );
+INSERT INTO t2 (b) VALUES ('a');
+UPDATE t2 SET b = 'b' WHERE a = 1;
+SELECT * FROM t2;
+a	b
+1	b
+DROP TABLE t1, t2;
 #---------------- 2-sweeps read Index merge test 2 -------------------------------
 SET SESSION STORAGE_ENGINE = InnoDB;
 drop table if exists t1;

mysql-test/r/index_merge_myisam.result

@@ -1158,6 +1158,61 @@ key1	key2	key3
 38	38	38
 39	39	39
 drop table t1;
+#
+# Bug#56423: Different count with SELECT and CREATE SELECT queries
+#
+CREATE TABLE t1 (
+a INT,
+b INT,
+c INT,
+d INT,
+PRIMARY KEY (a),
+KEY (c),
+KEY bd (b,d)
+);
+INSERT INTO t1 VALUES
+(1, 0, 1, 0),
+(2, 1, 1, 1),
+(3, 1, 1, 1),
+(4, 0, 1, 1);
+EXPLAIN
+SELECT a
+FROM t1
+WHERE c = 1 AND b = 1 AND d = 1;
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1	ref	c,bd	bd	10	const,const	2	Using where
+CREATE TABLE t2 ( a INT )
+SELECT a
+FROM t1
+WHERE c = 1 AND b = 1 AND d = 1;
+SELECT * FROM t2;
+a
+2
+3
+DROP TABLE t1, t2;
+CREATE TABLE t1( a INT, b INT, KEY(a), KEY(b) );
+INSERT INTO t1 VALUES (1, 2), (1, 2), (1, 2), (1, 2);
+SELECT * FROM t1 FORCE INDEX(a, b) WHERE a = 1 AND b = 2;
+a	b
+1	2
+1	2
+1	2
+1	2
+DROP TABLE t1;
+# Code coverage of fix.
+CREATE TABLE t1 ( a INT NOT NULL AUTO_INCREMENT PRIMARY KEY, b INT);
+INSERT INTO t1 (b) VALUES (1);
+UPDATE t1 SET b = 2 WHERE a = 1;
+SELECT * FROM t1;
+a	b
+1	2
+CREATE TABLE t2 ( a INT NOT NULL AUTO_INCREMENT PRIMARY KEY, b VARCHAR(1) );
+INSERT INTO t2 (b) VALUES ('a');
+UPDATE t2 SET b = 'b' WHERE a = 1;
+SELECT * FROM t2;
+a	b
+1	b
+DROP TABLE t1, t2;
 #---------------- 2-sweeps read Index merge test 2 -------------------------------
 SET SESSION STORAGE_ENGINE = MyISAM;
 drop table if exists t1;

sql/mysql_priv.h

@@ -1047,7 +1047,8 @@ bool dispatch_command(enum enum_server_command command, THD *thd,
 		      char* packet, uint packet_length);
 void log_slow_statement(THD *thd);
 bool check_dup(const char *db, const char *name, TABLE_LIST *tables);
-bool compare_record(TABLE *table);
+bool records_are_comparable(const TABLE *table);
+bool compare_records(const TABLE *table);
 bool append_file_to_dir(THD *thd, const char **filename_ptr, 
                         const char *table_name);
 void wait_while_table_is_used(THD *thd, TABLE *table,

sql/sql_insert.cc

@@ -1483,9 +1483,7 @@ int write_record(THD *thd, TABLE *table,COPY_INFO *info)
           table->file->adjust_next_insert_id_after_explicit_value(
             table->next_number_field->val_int());
         info->touched++;
-        if ((table->file->ha_table_flags() & HA_PARTIAL_COLUMN_READ &&
-             !bitmap_is_subset(table->write_set, table->read_set)) ||
-            compare_record(table))
+        if (!records_are_comparable(table) || compare_records(table))
         {
           if ((error=table->file->ha_update_row(table->record[1],
                                                 table->record[0])) &&

sql/sql_update.cc

@@ -25,11 +25,68 @@
 #include "sql_trigger.h"
 #include "debug_sync.h"
 
-/* Return 0 if row hasn't changed */
 
-bool compare_record(TABLE *table)
+/**
+   True if the table's input and output record buffers are comparable using
+   compare_records(TABLE*).
+ */
+bool records_are_comparable(const TABLE *table) {
+  return ((table->file->ha_table_flags() & HA_PARTIAL_COLUMN_READ) == 0) ||
+    bitmap_is_subset(table->write_set, table->read_set);
+}
+
+
+/**
+   Compares the input and outbut record buffers of the table to see if a row
+   has changed. The algorithm iterates over updated columns and if they are
+   nullable compares NULL bits in the buffer before comparing actual
+   data. Special care must be taken to compare only the relevant NULL bits and
+   mask out all others as they may be undefined. The storage engine will not
+   and should not touch them.
+
+   @param table The table to evaluate.
+
+   @return true if row has changed.
+   @return false otherwise.
+*/
+bool compare_records(const TABLE *table)
 {
+  DBUG_ASSERT(records_are_comparable(table));
+
+  if ((table->file->ha_table_flags() & HA_PARTIAL_COLUMN_READ) != 0)
+  {
+    /*
+      Storage engine may not have read all columns of the record.  Fields
+      (including NULL bits) not in the write_set may not have been read and
+      can therefore not be compared.
+    */ 
+    for (Field **ptr= table->field ; *ptr != NULL; ptr++)
+    {
+      Field *field= *ptr;
+      if (bitmap_is_set(table->write_set, field->field_index))
+      {
+        if (field->real_maybe_null())
+        {
+          uchar null_byte_index= field->null_ptr - table->record[0];
+          
+          if (((table->record[0][null_byte_index]) & field->null_bit) !=
+              ((table->record[1][null_byte_index]) & field->null_bit))
+            return TRUE;
+        }
+        if (field->cmp_binary_offset(table->s->rec_buff_length))
+          return TRUE;
+      }
+    }
+    return FALSE;
+  }
+  
+  /* 
+     The storage engine has read all columns, so it's safe to compare all bits
+     including those not in the write_set. This is cheaper than the field-by-field
+     comparison done above.
+  */ 
   if (table->s->blob_fields + table->s->varchar_fields == 0)
+    // Fixed-size record: do bitwise comparison of the records 
     return cmp_record(table,record[1]);
   /* Compare null bits */
   if (memcmp(table->null_flags,
@@ -186,7 +243,6 @@ int mysql_update(THD *thd,
   bool		using_limit= limit != HA_POS_ERROR;
   bool		safe_update= test(thd->options & OPTION_SAFE_UPDATES);
   bool		used_key_is_modified, transactional_table, will_batch;
-  bool		can_compare_record;
   int           res;
   int		error, loc_error;
   uint		used_index= MAX_KEY, dup_key_found;
@@ -575,15 +631,6 @@ int mysql_update(THD *thd,
   if (table->file->ha_table_flags() & HA_PARTIAL_COLUMN_READ)
     table->prepare_for_position();
 
-  /*
-    We can use compare_record() to optimize away updates if
-    the table handler is returning all columns OR if
-    if all updated columns are read
-  */
-  can_compare_record= (!(table->file->ha_table_flags() &
-                         HA_PARTIAL_COLUMN_READ) ||
-                       bitmap_is_subset(table->write_set, table->read_set));
-
   while (!(error=info.read_record(&info)) && !thd->killed)
   {
     thd->examined_row_count++;
@@ -601,7 +648,7 @@ int mysql_update(THD *thd,
 
       found++;
 
-      if (!can_compare_record || compare_record(table))
+      if (!records_are_comparable(table) || compare_records(table))
       {
         if ((res= table_list->view_check_option(thd, ignore)) !=
             VIEW_CHECK_OK)
@@ -1695,18 +1742,8 @@ bool multi_update::send_data(List<Item> &not_used_values)
     if (table->status & (STATUS_NULL_ROW | STATUS_UPDATED))
       continue;
 
-    /*
-      We can use compare_record() to optimize away updates if
-      the table handler is returning all columns OR if
-      if all updated columns are read
-    */
     if (table == table_to_update)
     {
-      bool can_compare_record;
-      can_compare_record= (!(table->file->ha_table_flags() &
-                             HA_PARTIAL_COLUMN_READ) ||
-                           bitmap_is_subset(table->write_set,
-                                            table->read_set));
       table->status|= STATUS_UPDATED;
       store_record(table,record[1]);
       if (fill_record_n_invoke_before_triggers(thd, *fields_for_table[offset],
@@ -1721,7 +1758,7 @@ bool multi_update::send_data(List<Item> &not_used_values)
       */
       table->auto_increment_field_not_null= FALSE;
       found++;
-      if (!can_compare_record || compare_record(table))
+      if (!records_are_comparable(table) || compare_records(table))
       {
 	int error;
         if ((error= cur_table->view_check_option(thd, ignore)) !=
@@ -1908,7 +1945,6 @@ int multi_update::do_updates()
     DBUG_RETURN(0);
   for (cur_table= update_tables; cur_table; cur_table= cur_table->next_local)
   {
-    bool can_compare_record;
     uint offset= cur_table->shared;
 
     table = cur_table->table;
@@ -1945,11 +1981,6 @@ int multi_update::do_updates()
     if ((local_error = tmp_table->file->ha_rnd_init(1)))
       goto err;
 
-    can_compare_record= (!(table->file->ha_table_flags() &
-                           HA_PARTIAL_COLUMN_READ) ||
-                         bitmap_is_subset(table->write_set,
-                                          table->read_set));
-
     for (;;)
     {
       if (thd->killed && trans_safe)
@@ -1990,7 +2021,7 @@ int multi_update::do_updates()
                                             TRG_ACTION_BEFORE, TRUE))
         goto err2;
 
-      if (!can_compare_record || compare_record(table))
+      if (!records_are_comparable(table) || compare_records(table))
       {
         int error;
         if ((error= cur_table->view_check_option(thd, ignore)) !=

storage/innobase/row/row0sel.c

@@ -2621,12 +2621,6 @@ row_sel_store_mysql_rec(
 		prebuilt->blob_heap = NULL;
 	}
 
-	/* init null bytes with default values as they might be
-	left uninitialized in some cases and this uninited bytes
-	might be copied into mysql record buffer that leads to
-	valgrind warnings */
-	memcpy(mysql_rec, prebuilt->default_rec, prebuilt->null_bitmap_len);
-
 	for (i = 0; i < prebuilt->n_template; i++) {
 
 		templ = prebuilt->mysql_template + i;

storage/innodb_plugin/row/row0sel.c

@@ -2696,12 +2696,6 @@ row_sel_store_mysql_rec(
 		prebuilt->blob_heap = NULL;
 	}
 
-	/* init null bytes with default values as they might be
-	left uninitialized in some cases and these uninited bytes
-	might be copied into mysql record buffer that leads to
-	valgrind warnings */
-	memcpy(mysql_rec, prebuilt->default_rec, prebuilt->null_bitmap_len);
-
 	for (i = 0; i < prebuilt->n_template; i++) {
 
 		templ = prebuilt->mysql_template + i;