unistd::crypt Boundry Case

FROM : Jesse Thompson
DATE : Wed Oct 13 23:53:37 2004

I've found a problem with Mac's implementation of crypt. Given the code:

#define _XOPEN_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

int main( int argc, const char **argv )
{
        int index = 0;
        char c[2] = { 32, 89 };
        FILE    *outFile = fopen( "crypt.out", "w" );

        for( index = 32; index <= 126; index++ ) {
                c[0] = index;
                fprintf( outFile, "%s\t%c\n", crypt( "cook", c ), c[0] );
        }

        fclose( outFile );

        return 0;
}

Which permutes through the range of printable characters for the first
character of salt, crypt is returning constant results for non [ 0-9,
a-z, A-Z ] characters. The same code on a Linux or Unix box returns
proper hashes. My CS professor states that the crypt function should
conform to NSB's implementation of DES, and this obviously doesn't.
Here's a snip-it of the output of this program:

YE.6Yy26visQ
!YE.6Yy26visQ  !
"YE.6Yy26visQ  "
#YE.6Yy26visQ  #
$YE.6Yy26visQ  $
%YE.6Yy26visQ  %
&YE.6Yy26visQ  &
'YE.6Yy26visQ  '
(YE.6Yy26visQ  (
)YE.6Yy26visQ  )
*YE.6Yy26visQ  *
+YE.6Yy26visQ  +
,YE.6Yy26visQ  ,
-YE.6Yy26visQ  -
.YE.6Yy26visQ  .
/YevSCZ/FT75w  /
0YZZJ3k9.p0dc  0
1YhuSC9o37LLM  1
2Y/fR9L/ylgEk  2
3YVZmoR2ioo8k  3
4YsrD5ASSxdfY  4
5YpX7aBI.FRHM  5
6YWn73aYQNjZk  6
7YGi5LS/mQoyI  7
8Y9fdm4BU9D8k  8
9Y6M4Uw3W2vWI  9

- Jesse