integer overflow yields incorrect results and buffer overflow on 64-bit machines

Torbjorn Granlund tg at
Mon Mar 3 23:24:34 CET 2008

Vincent Lefevre <vincent at> writes:

  On 2008-02-27 18:19:39 +0100, Torbjorn Granlund wrote:
  > What do you think of this change?
  Seems OK.
On second thought, something trickier will be needed for 32-bit
machines (or actually any machine where sizeof(int) = sizeof(long))
since we don't support operands greater than whose bit count can be
stored in an unsigned long int.

  However there's a possible overflow in n_pow_ui.c.
And it's not easy to handle that without a umul_ppmm in the general
case.  Using umul_ppmm is no disaster, but I prefer to stay away from
it when possible, since it is expensive on certain machines.


More information about the gmp-bugs mailing list