[Vm-dev] Integer overflow with BitBlt rule 20 and depth 32
David T. Lewis
lewis at mail.msen.com
Fri Oct 30 10:12:09 UTC 2009
On Fri, Oct 30, 2009 at 10:09:36AM +0100, Henrik Johansen wrote:
>
> Not that it matters on 32bit architecture, but don't you in theory
> have to use unsigned long to ensure an integer of at least 32 bits is
> used?
>
> Other than that I see no problems, until someone decides to add depths
> > 32. (And then you'd have to modify the methods anyways).
unsigned int is 32 bits on all current Squeak platforms, while
unsigned long may be 64 bits. Thus unsigned int produces the same
behavior on all current platforms, and Juan added an overflow check
in the one method for which overflow was is a problem.
Dave
More information about the Vm-dev
mailing list