Re: [avr-libc-dev] sscanf problem

[Joerg Wunsch]

As Kang Tin LAI wrote:

> sscanf(qstrg, "00=%u&01=%u&02=%u&03=%u&" \
>       "04=%u&05=%u&06=%u&07=%u&" \
>       "08=%u&09=%u&0a=%u&0b=%u&" \
>       "0c=%x&0d=%x&0e=%x&0f=%x&10=%x&11=%x", \

The %u format didn't work until a few days ago, i. e. all released
versions of avr-libc do still have an erroneous implementation of it.

>       aap, aap + 1, aap + 2, aap + 3, \
>       aap + 4, aap + 5, aap + 6, aap + 7, \
>       aap + 8, aap + 9, aap + 10, aap + 11, \
>       &mac[0], &mac[1], &mac[2], &mac[3], &mac[4], &mac[5]);

> The *(aap + 12), ie., addr[24] and addr[25], and mac[6] are changed by sscanf 
> unexpectly. So addr[24] and mac[6] will crash, use addr[26] and mac[7] to 
> bypass 
> this problem.

Can you try compiling the latest version from CVS, and see whether the
problem still exists with that version?  If it still fails, it must be
a general problem with the way the compiler handles long variadic
argument lists.  I guess this feature has only been rarely tested so
far, since it is not frequently used.

-- 
J"org Wunsch                                           Unix support engineer
address@hidden        http://www.interface-systems.de/~j/