Question

What is the difference between

 char cur_byte=*((char *)(buf+i));

and

char *b=(char *)(buf);
char cur_byte=*(b+i);

Assume:
buf is a pointer to void// void *buf; and
i is used as an iterator in a for loop
I found this code(the first line) in a c source code which generates rabin fingerprints and because VC2010 express reported it as an error I had to replace it with the second two lines. And I am not sure if it can do the intended purpose. Plus I would be grateful if anyone can give me a hint where to get a working C++ source code for content defined chunking and fingerprint generating.

Answer 1

In your first statement you add an integer (i is an integer type, right?) to a void*, casting to char* afterwards. Pointer arithmetic with void pointers is not defined by the C standard, because the compiler has no way to know of how much it should increment the pointer. Some compilers, however, define sizeof(void) == 1. In this case, your two snippets are equivalent, which explains why this code may have worked with another compiler (thanks Steve Jessop for pointing this).

What you meant in your first snippet was probably
char cur_byte=*(((char *) buf) + i);, the character pointed by the address located i characters after buf.

In the following schema, where i==4, cur_byte would be assigned the value r.

Memory: |a| |w|o|r|d
         ^       ^
        buf     buf+i

---

In your second statement:

char *b=(char *)(buf);
char cur_byte=*(b+i);

you first assign buf to b, and then assign the content of b + i to cur_byte. b has type char* so adding i will give the address i characters after b.

Memory: |a| |w|o|r|d|
         ^       ^
        buf         
         b      b+i

In the end these two statements are equivalent (except for the assignment of b).