integer math errors with large numbers and no errors or warnings given

[archivist]

mysql> select version();
+-----------------------------+
| version() |
+-----------------------------+
| 5.1.39-maria-beta-debug-log |
+-----------------------------+
1 row in set (0.00 sec)

mysql> SELECT 4294967296 * 4294967296;
+-------------------------+
| 4294967296 * 4294967296 |
+-------------------------+
| 0 |
+-------------------------+
1 row in set (0.00 sec)

mysql> show warnings;
Empty set (0.00 sec)

mysql> SELECT 3037000500 * 3037000500;
+-------------------------+
| 3037000500 * 3037000500 |
+-------------------------+
| -9223372036709301616 |
+-------------------------+
1 row in set (0.00 sec)

mysql> show warnings;
Empty set (0.00 sec)

mysql> SELECT 4294967296 * 4294967297;
+-------------------------+
| 4294967296 * 4294967297 |
+-------------------------+
| 4294967296 |
+-------------------------+
1 row in set (0.00 sec)

mysql> show warnings;
Empty set (0.00 sec)

mysql> SELECT 4294967296 * 4294967298;
+-------------------------+
| 4294967296 * 4294967298 |
+-------------------------+
| 8589934592 |
+-------------------------+
1 row in set (0.00 sec)

mysql> show warnings;
Empty set (0.00 sec)

mysql> SELECT 4294967296.0 * 4294967298.0;
+-----------------------------+
| 4294967296.0 * 4294967298.0 |
+-----------------------------+
| 18446744082299486208.00 |
+-----------------------------+
1 row in set (0.08 sec)

mysql> show warnings;
Empty set (0.00 sec)

mysql> quit;
Bye
archivist@cnc:~/archivist-cnc/archivist-cnc/build/mysql-test$ uname -a
Linux cnc 2.6.24-16-rtai #1 Tue Sep 30 22:54:33 EEST 2008 i686 GNU/Linux

adding a .0 to the large number gets it to use float/double maths, the integer maths does not seem to check its range or overflow conditions correctly thus giving incorrect results.

# Bug#xxxxx: Reproducible MariaDB server integer math fault
#
# this returns 0 incorrectly it appears to be the left side * 0
--exec $MYSQL test -e "SELECT 4294967296 * 4294967296\g" 2>&1
#
# this gives an incorrect negative number
--exec $MYSQL test -e "SELECT 3037000500 * 3037000500\g" 2>&1
#
# this returns 4294967296 the number on the left times 1
--exec $MYSQL test -e "SELECT 4294967296 * 4294967297\g" 2>&1
#
# this returns 8589934592 the number on the left times 2
--exec $MYSQL test -e "SELECT 4294967296 * 4294967298\g" 2>&1
#
# this seems to be correct it has cast to a float/double
--exec $MYSQL test -e "SELECT 4294967296.0 * 4294967298.0;\g" 2>&1

[Michael Widenius (monty)]

The issue is that you are using integer arithmetic, which is described at:
http://dev.mysql.com/doc/refman/5.1/en/arithmetic-functions.html

Not much we can do with this, as we are trying to be compatible with MySQL.

We should however consider to find a way to create a warning in cases when you get overflow.

I will create a worklog entry for this.