Bug#426000: it's a problem with errno...

[Frédéric BOITEUX]

  I've investigated in this annoying bug, and found that it was a bad
interaction between sscanf() code and errno : the problem arises in
sscanf() when it's called with errno set to EINTR. I've updated the
testcase accordingly, it's joined to this e-mail.

  The problem doesn't exist in Woody, appeared in Sarge, exists in Etch
but seems fixed in Lenny/Sid, I've tested it on a Sid with libc6 2.6-2.

        Fred.


/*
  Test case for a possible glibc bug.
  Frédéric Boiteux <[EMAIL PROTECTED]>

  Scan two times the same word with a sscanf(). Between them,
  set the errno to EINTR (instead of a real fonction, to simplify testcase) ;
  The second scan fails, the '%n' converter isn't honoured !

  Tested on i386 architecture :
  O.K.    on Debian GNU/Linux 3.0 (Woody), libc6 version 2.2.5-11.8
  Problem on Debian GNU/Linux 3.1 (Sarge), libc6 version 2.3.2.ds1-22sarge6
  Problem on Debian GNU/Linux 3.0 (Etch),  libc6 version 2.3.6.ds1-13


  Sample output :

string='Hello', res=1, word='Hello', n=5
string='Hello', res=1, word='Hello', n=-1  *Problem!*

*/


#include <errno.h>
#include <stdio.h>


void scan(char *string)
{
        int n, res;
        char word[100];

        n = -1;
        res = sscanf(string, "%s %n", word, &n);
        printf("string='%s', res=%d, word='%s', n=%d", string, res, word, n);
        if ((res == 1) && (n < 0))
                printf("  *Problem!*\n");
        else
                putchar('\n');
}

int main(void)
{
        scan("Hello"); // success
        // success, expected result :
        // string='Hello', res=1, word='Hello', n=5

        // a function call would set the errno to EINTR...
        errno = EINTR;

        // uncomment following line to get correct behaviour :
        // errno = 0;

        scan("Hello"); // fails, 'n' isn't set...

        return 0;
}