Yet to mention: I didn't include a NULL-pointer check

your solution doesn't deal with the case that one string is the prefix of the other.

>>return *(--s1) - *(--s2);
That may give you the wrong result. It should be return *s1 - *s2; There are four possible return values

1. If *s1 is '\0' and *s2 is not, then the return value will be some negative value.
2. If *s1 is NOT '\0' but s2 is, then the return value will be some positive value
3. If both *s1 and *s2 are '\0' then the return result will be 0.
4. If neither *s1 nor *s2 is '\0' then the return value could be either positive or negative, depending on their values.

In either event, decrementing s1 and s2 before the subtraction may (and most probably will) produce the wrong result because s1 and s2 will not be pointer to the characters that cause the previous loop to terminate.

@Ancient Dragon:
Thanks for pointing out this issue, I didn't notice it when I was writing the code. However, the solution you suggest (unfortunately) won't fix the problem, since it will create another one.
If I replace return *(--s1) - *(--s2); by return *s1 - *s2; ,
then strcmp("a", "b") would yield the result 0, and not -1, because both s1 and s2 are incremented by one in the while's expression, s1 and s2 will point to the
nul-terminators by the time the return statement is encountered, and thus
return *s1 - *s2; will result in returning 0.

correction

Here's a better idea for the return value: return an explicit value instead of allowing the possibility of signed integer overflow?

@Dave: how can *s1 - *s2 cause an integer overflow? The maximum values would be -255 (0 - 255). That's hardly even close to an integer overflow value.

@AD: I'll leave that as an exercise for the programmer.

You don't really like portable programming or the design process involved, do you? If it happens to run on a particular compiler at a given moment in time, it appears "correct" to you?

Do you remember that qsort comparison function with the integer overflow issue?



No. I don't wantto continue. You are the lord of this roost. I am exhausted trying to get people to write portable code -- especially since you always come along and tell people to write implementation-specific stuff. Again, this is why I think of you as Daniweb's Herb Schildt. But he is a good author who has influenced many C and C++ programmers. It just sucks that it took him like 10 years to clean up his code. I think I need a break from all of this stupid bickering. Please take some time to wander into c.l.c and learn some new things.

@Dave: I didn't quite get you when you said:
Also, you seem to do exactly the same as I and AD did in your code snippet:
http://www.daniweb.com/code/snippet216567.html .
So, could you please clarify what you meant with returning an explicit value to avoid the possibility of a signed integer overflow ?

[EDIT]
@Dave:
Do you perhaps mean: return (int) *s1 - (int) *s2; ?
[/EDIT]