On Wed, Jan 02, 2013 at 01:12:15PM -0600, H Hartley Sweeten wrote:
> On Wednesday, January 02, 2013 12:03 PM, Ian Abbott wrote:
> > On 02/01/2013 18:10, H Hartley Sweeten wrote:
> >> Will the munge_count always be larger, or equal, to the buf_read_alloc_count?
> >> The original code had the (int) cast so I wasn't sure. If so, then yes the variable
> >> should be an unsigned int.
> >
> > Well the value it is being compared to (nbytes) is an unsigned int, so
> > the int value (available) will be converted to unsigned int before the
> > comparison anyway. So making 'available' an unsigned int to begin with
> > shouldn't break anything that isn't already broke!
>
> Not so. Simple test:
>
Uh... Ian is correct. In the new code it doesn't matter if
"available" is signed or unsigned because the comparison is always
promoted to unsigned.
I think you are assuming that 99 is an unsigned int but actually it
defaults to int. The test should be something like:
int available = -40;
unsigned int nbytes = 5;
if (nbytes > available)
printf("Not promoted\n");
else
printf("Negative 'availabled' treated as large unsigned int.\n");
I think the rules are that comparisons are promoted to the type with
the most bits for positive integers. That's how I treat them in
Smatch.
regards,
dan carpenter
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
