Re: link pointer access problem



vichy.kuo@xxxxxxxxx wrote:
Dear all:
Below is my program:
#include<stdio.h>
typedef struct{
int b_value;
}B;

typedef struct{
B* next;
int a_value;
}A;

int main(void)
{
A a;

At this point, a is uninitialized; all of it's members have
indeterminate values. That means, in particular, that a.next may have
a trap representation.

//B* pb;
a.a_value=12;
//pb=a.next;
//pb->b_value=23;

You don't have any objects of type B in your program, and you have not
set a.next to point to any such object. You haven't even set a.next to
be a null pointer (though that would not help, in this case). At this
point, a.next still has an indeterminate value. Therefore, the
behavior of the next statement is undefined:

a.next->b_value=23;

The same is true of this statement, for the same reason:

printf("b_value=%dvn",a.next->b_value);

//printf("b_value=%dvn",pb->b_value);
return 0;
}

The program will get segment fault or

That is an entirely reasonable and plausible result from attempting to
dereference a pointer with an indeterminate value.

... compiler will say "request for
member 'b_value' in something not a structure or union".

However, I see no justification for such a message. A smart compiler
could have recognized at compile time that your code dereferences an
uninitialized pointer; but if so, this message seems like a very odd
way to describe that fact.

The program will be fine if I unmark "//" part, which is the way that
people usually to access link list.

It will not be fine. Even simply copying the value of a.next to pb has
undefined behavior, if a.next contains a trap representation. If
a.next doesn't contain a trap representation, it's indeterminate value
will be copied to pb, with no guarantee that it points at any location
that can be safely dereference. However, since the behavior is
undefined, one of the possible results is that, this time, the
uninitialized value of a.next happens to point at memory that you can
safely read and write to.

"next" is an element of A, so a.next should be fine.
"a.next" is a pointer which point to B,

Correct.

... so a.next->b_value seems ok.

Not until a.next has been given a value that points at an actual B
object.
.

Relevant Pages

  • Re: Neatest way to get the end pointer?
    ... And you invoke undefined behavior. ... as a pointer 'index' compared against an int * iterator. ... One byte beyond is fine (though you can't dereference it), ...
    (comp.lang.c)
  • Re: Pointer assigning.
    ... element) to the object pointed to by ptr2. ... int *ptr2 = NULL; ... want to - we were mistaken in adding a dereference operator). ... All we want is to assign a pointer value to ptr2. ...
    (comp.lang.c)
  • Re: The difference between 0 and NULL?
    ... There's a real distinction here between assigning a null pointer ... and assigning all-bits-zero where a null pointer is NOT all-bits-zero. ... If you take *ANY* integer constant, assign it to an int variable, ... and then cast it to a pointer and dereference it, ...
    (comp.unix.programmer)
  • Re: "float* f" vs. "float *f"
    ... int i, *j; ... because `*j' is how I'd dereference `j' in an expression. ... which is not that different (function pointer syntax as well as arrays is a special case anyway). ... unsigned char *const *bibble; ...
    (comp.lang.c.moderated)
  • CreateMapFile e LPCWSTR
    ... int g_Length; ... from the same pointer i can read the values stored. ... and set another pointer to my structure by MapViewOfFile, ...
    (microsoft.public.pocketpc.developer)