Description:
------------
When I want to calculate integer summ of two integers with overflow, I get unexpected result, becouse result converts to float, then converts to integer.

Important: I've tested this example on two systems with same versions of php (5.1.4) and different versions of freebsd (4.11 and 6.1).
On last operation [$b = (int)($a + $a)], freebsd 4.11 produce expected result: int(0), but freebsd 6.1 produce minimal integer (-2147483648).

How can I implement algorithm with integer mathematics, that requires overflows? i have it successfully working on freebsd 4.11, but it don't work on freebsd 6.1 ((

Reproduce code:
---------------
$a = 0x80000000; //i'd like it would integer, but it float ;)
var_dump($a, dechex($a));
$a = (int)0x80000000; //ok, i'll make it integer myself
var_dump($a, dechex($a));
$b = $a + $a; //int + int = int ?
var_dump($b, dechex($b));
$b = (int)($a + $a);
var_dump($b, dechex($b));


Expected result:
----------------
int(-2147483648) string(8) "80000000"
int(-2147483648) string(8) "80000000"
int(0) string(8) "0" //expected, int + int = int with overflow
int(0) string(8) "0" //expected, conversion with overflow

Actual result:
--------------
float(2147483648) string(8) "80000000" // expected int
int(-2147483648) string(8) "80000000" // when I convert to int myself
float(-4294967296) string(8) "80000000" //expected int, but it overflows and converts to float
int(-2147483648) string(8) "80000000"