problem with strlen

[happyefforts]

void show1(char *s) {
  int i;
  for (i = 0; i < strlen(s); i++) printf("%c", s[i]);
  printf("\n");
}

void show2(char *s) {
  int i;
  for (i = 0; i <= strlen(s) - 1; i++) printf("%c", s[i]);
  printf("\n");
}

When I do show1(""), it has no problem. But show2("") gives me a error. Anyone know why?

[Markus]

The string length (strlen) of "" is 0 (strlen = 0)
strlen - 1 = -1
s[-1] is accessing data to which your not permitted access. This is a fundamental part of working with memory in C.

[Banfa]

Close Markus, it never tries to access s[-1], although you are right about the condition causing the problem.

strlen(s) - 1 for s == "" == -1

Therefore the condition is i <= -1

Since i is initialised to 0 on a 32 bit system that allows i to have the range 0 - 2147483647 until i final wraps and causes undefined behaviour (for wrapping a signed integer).

However since s has length 0 as soon as i >= 1 you are accessing memory outside the bounds of the array and at some point in the 2147483646 out of bounds addresses it tries I expect it hits something that really causes a problem.


For this reason if you have a size (especially if it can be 0) for something SIZE and are range testing you should always use

range < SIZE

and not

range <= SIZE - 1

I have seen this in a production product cause the same problem.

[donbock]

What error do you get for show2?

Banfa: if strlen(s)-1 evaluated to -1 then execution would skip over the loop. There would be no output or run-time error.

strlen returns size_t, which is typically an unsigned type.

Is the expression (strlen(s) - 1) evaluated for type size_t or type int? If size_t, then -1 is considered to be a very large positive number that is too large to fit in an int (for a 2's-complement architecture). If int, then the result is truly -1.

The comparison of i to the expression is performed using int arithmetic. If the expression is of type size_t then the very large positive number is converted to int. The result of this conversion may be a positive number or a negative number, depending on implemention-dependent properties of the compiler.

At least that's what I think might be going on.

[Markus]

Of course, now I see it.

Mark (shouldn't rush to answer things).