> - if (tolower(s[count]) == 'k') { value *= 1024; count++; }
> - else if (tolower(s[count]) == 'm') { value *= 1024*1024; count++; }
> + {
> + int ov = value;
> + if (tolower(s[count]) == 'k') { value *= 1024; count++; }
> + else if (tolower(s[count]) == 'm') { value *= 1024*1024; count++; }
>
> - if (value < 0) log_write(0, LOG_PANIC_DIE|LOG_CONFIG_IN,
> - "integer \"%s\" is too large (overflow)", s);
> + if (value < 0 && ov > 0) log_write(0, LOG_PANIC_DIE|LOG_CONFIG_IN,
> + "integer \"%s\" is too large (overflow)", s);
> + }
Just thinking about it: Does the assumption of value < 0 indicating
overflow always hold?
I don't know any systems where signed integer overflows cause exceptions,
but I think C allows that to happen, so how about this:
if (tolower(s[count]) == 'k')
{
if (ok = (value>0 ? value>INT_MAX/1024 : value <INT_MIN/1024))
value *= 1024;
}
count++;
Not tested, just a thought. In case negative values are not allowed,
things are even easier.
Michael
