Re: Rounding of OS match percentages differs between architectures

["Michael Pattrick" <mpattrick () rhinovirus org>]

On Wed, Nov 19, 2008 at 8:26 PM, David Fifield <david () bamsoftware com> wrote:
> On Wed, Nov 19, 2008 at 07:54:14PM -0500, Michael Pattrick wrote:
> > On Wed, Nov 19, 2008 at 6:32 PM, David Fifield <david () bamsoftware com> wrote:
> > > Nmap currently uses printf("%d%%", (int) (frac * 100)) to show match
> > > percentages. Using printf("%.f%%", frac * 100) would remove this problem
> > > but would create another, namely that everything 99.5% and above would
> > > be rounded to 100%.
> > >
> > > Anyone have an opinion on this or see a good way to handle it?
> >
> > If I understand the problem properly - you want to strip the decimals,
> > why not just use floor()?
>
> Hey, thanks, that works! I guess I assumed that ((int) d) and floor(d)
> would have the same result when d is positive. Apparently floor has
> extra logic to know when its argument is less than but very close to an
> integer.
>
> (To see what I mean by "very close to," run python and type in "0.98". I
> see "0.97999999999999998" as output.)
>
> I wonder if this is really caused by the C library? Can anyone else
> reproduce the "97% 98%" output from the test program I posted?
I got the same results as you testing on my 64 and 32bit Linux computers.

I'm no expert but I would assume the the problem is an unhappy
combination of a bug in libc and floating point inaccuracies. Consider
the following

#include <iostream>
using namespace std;
int main(){
     for(float i=1.01;i<=5.05;i=i+1.01){
          cout << i << ", ";
     }
     cout << endl;
}

outputs "1.01, 2.02, 3.03, 4.04,"

_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://SecLists.Org