One line: First code you are dereferencing uninitialized pointer which exhibits undefined behaviour, and in the second code you are dereferencing initialized pointer which will give access to the value at the address.
A bit of explanation:
First you need to realize that a pointer is nothing but an integer, an with the *var
we tell the compiler that we will be using the content of the variable var
(the integer in it) as an address to fetch the value in that address. If there is **var
similarly we tell the compiler that we will first use the stored value of the variable var
to fetch the value at the address and again use this fetched value as an address and fetch the value stored in it.
Therefore in your first declaration it is:
+----------+ +----------+
| garbage | | garbage |
+----------+ +----------+
| a | | b |
+----------+ +----------+
| addr1 | | addr2 |
+----------+ +----------+
Then you try to use the value stored in a
as an address. a
contains garbage, it can be any value, but you do not have access to any address location. Therefore the next moment when you do *a
it will use the stored value in a
as an address. Because the stored value can be anything, anything can happen.
If you have permission to access the location , the code will continue to execute without a segmentation fault. If the address happens to be an address from the heap book-keeping structure, or other memory area which your code allocated from heap or stack then when you do *a = 10
it will simply wipe off the existing value with 10
in that location. This can lead to undefined behaviour as now you have changed something without the knowledge of the context having the actual authority of the memory. If you don't have permission to the memory, you simply get a segmentation fault. This is called dereferencing of an uninitialized pointer.
Next statement you do a = &b
which just assigns the address of b
in a
. This doesn't help, because the previous line has dereferenced an uninitialized pointer.
Next code you have something like this after the third statement:
+----------+ +----------+
| addr2 |---+ | garbage |
+----------+ | +----------+
| a | +--> | b |
+----------+ +----------+
| addr1 | | addr2 |
+----------+ +----------+
The third statement assigns the address of b
into a
. Before that a
is not dereferenced, therefore the garbage value stored in a
before the initialization is never used as an address. Now when you assign a valid address of your knowledge into a
, dereferencing a
now will give you access to the value pointed to by a
.
Extending the answer, you need to keep an eye that, even if you have assigned a valid address to a pointer, you have to make sure at the time of dereferencing the pointer the lifetime of the address pointed to by the pointer has not expired. For example returning local variable.
int foo (void)
{
int a = 50;
return &a; //Address is valid
}//After this `a' is destroyed (lifetime finishes), accessing this address
//results in undefined behaviour
int main (void)
{
int *v = foo ();
*v = 50; //Incorrect, contents of `v' has expired lifetime.
return 0;
}
Same in the case of accessing freed memory location from heap.
int main (void)
{
char *a = malloc (1);
*a = 'A'; //Works fine, because we have allocated memory
free (a); //Freeing allocated memory
*a = 'B'; //Undefined behaviour, we have already freed
//memory, it's not for us now.
return 0;
}