views:

681

answers:

3

So I have the following program:

int main(){
  char* one = "computer";
  char two[] = "another";
  two[1]='b';
  one[1]='b';
  return 0;
}

It segfaults on the line "one[1]='b'" which makes sense because the memory that the pointer "one" points to must be in read only memory. However, the question is why doesn't the line "two[1]='b'" segfault? Looking at the assembly output from gcc:

.file   "one.c"
        .section        .rodata
.LC0:
        .string "computer"
.LC1:
        .string "another"
        .text
.globl main
        .type   main, @function
main:

We see that both strings are in the rodata section so they are readonly. So then how come the line "two[1]='b' does not segfault?

+23  A: 

one points directly to the string located in a read-only page. On the other hand, two is an array allocated on the stack and is initialized with some constant data. At run time, the string in the read only section of the executable will be copied to the stack. What you are modifying is the copy of that string on the stack, not the read-only memory page.

At a higher level perspective, from the language point of view, "abcd" is an expression of type const char* and not char*. Thus, modifying the value pointed by such an expression results in undefined behavior. The statement char* one = "something"; merely stores the pointer to the string in a variable (unsafely, since it's casting away const modifier). The char two[] = "something"; is totally different. It's actually declaring an array and initializing it, much like int a[] = {1,2,3};. The string in quotes here is the initialization expression.

Mehrdad Afshari
Wow... I've been programming in C for about 5 years now and I was not aware that `char[]` made a copy of constant data. I always thought `[]` was just a fancier way of writing `*`. Thank you! +1
Earlz
I often write `char str[] = {"Something"};` in an attempt to make the association clearer.
LnxPrgr3
+1  A: 

The "another" you see in the rodata section is what will be copied in the array two when it will be initialized. On the other hand the address of the string "computer" will be assigned to one.

So, one is pointing to a read only segment (and hence the segfault on write) while two will be allocated on the stack and then "another" will be copied into it.

Remo.D
A: 

The second form creates an array by copying the literal string.

It is equivalent to:

char two[] = {'a', 'n', 'o', 't', 'h'. 'e', r', '\0'};

You can initialize a character array with variables, such as

char c = 'a';
char two[] = {'a', 'n', c, '\0'};
Sherwood Hu