Re: MIME decoding confused by non-MIME characters

On Tue, Feb 27, 2007 at 09:20:56AM +0000, Brian Candler wrote:
> irb(main):001:0> RUBY_VERSION
> => "1.8.5"
> irb(main):002:0> a = "b2s="
> => "b2s="
> irb(main):003:0> b = "\xef\xbb\xbf" + a
> => "\357\273\277b2s="
> irb(main):004:0> a.unpack("m")
> => ["ok"]
> irb(main):005:0> b.unpack("m")
> => ["$\000\e\332"]
> 
> That is, non-printable characters (here a UTF8-encoded BOM) are causing MIME
> unpack to return garbage.

I had a quick look in pack.c in 1.8.6-preview2. I think the problem is here:

    char *s;
    ...
                    if ((a = b64_xtable[(int)s[0]]) == -1) break;
                    if ((b = b64_xtable[(int)s[1]]) == -1) break;
                    if ((c = b64_xtable[(int)s[2]]) == -1) break;
                    if ((d = b64_xtable[(int)s[3]]) == -1) break;

A negative signed char will still be negative after casting to int, as can
be demonstrated like this:

---- 8< ----------------------------------------
#include <stdio.h>
int main(void)
{
  char *s = "\xff";
  printf("%d\n", (int)s[0]);   /* prints -1 */
  return 0;
}
---- 8< ----------------------------------------

So I think those entries need to be changed to

    b64_xtable[((unsigned char *)s)[0]]  ... etc

Regards,

Brian.