[pechtcha: Bug in scanf &Co (Was Re: Error message from antiword since upgrade to cygwin 1.5.10)]

[Christopher Faylor <me at cgf dot cx>]

----- Forwarded message from Igor Pechtchanski <pechtcha> -----

From: Igor Pechtchanski
To: cygwin
Cc: "Gerrit P. Haase"
Subject: Bug in scanf &Co (Was Re: Error message from antiword since upgrade to cygwin 1.5.10)
Date: Wed, 2 Jun 2004 14:04:33 -0400 (EDT)
In-Reply-To: <Pine.GSO.4.58.0406021327090.18478@slinky.cs.nyu.edu>

On Wed, 2 Jun 2004, Igor Pechtchanski wrote:

> On Wed, 2 Jun 2004, Gerrit P. Haase wrote:
>
> [snip]
> > Everything works ok with cygwin-1.5.9.  I'll try to recompile with
> > 1.5.10 later, maybe that helps...
> >
> > Gerrit
>
> Try the attached patch.  FWIW, it seems like there's a bug here, but
> sscanf used to mask it.  Was there a change in the sscanf implementation
> since 1.5.9?
>         Igor

Never mind.  This is a bug in the newlib scanf family implementation
(actually, in __svfscanf_r in libc/stdio/vfscanf.c) that causes it to read
strings of the form 0x0+ incorrectly.  Attached is a testcase that
demonstrates the bug.  I suspect the following change:

2004-04-21  J"orn Rennecke

        * libc/stdio/vfscanf.c (NNZDIGITS): New define.
        (__svfscanf_r): In integer conversions, leave out leading zeroes
        which are not part of a base prefix.
        Keep track of width truncation to fit into buf, not counting left-out
        zeroes against width till the truncation has been compensated for.

#include <stdlib.h>
#include <stdio.h>

int main(char ac, char *av[]) {
  const char *str1 = "0x1 0x0 0x3";
  const char *str2 = "0x1 0x2 0x3";
  unsigned int i,j,k;
  sscanf(str1, "%x %x %x", &i, &j, &k);
  printf("Values are: %x %x %x\n", i, j, k);
  sscanf(str2, "%x %x %x", &i, &j, &k);
  printf("Values are: %x %x %x\n", i, j, k);
}

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/

----- End forwarded message -----