Question

I have a class (A) which has to include two files, class X and class Y. unfortunately class Y also needs to include class X in it’s header file because the constructor takes a pointer to a class X type as an argument.

The potential problem is that there would be a linker error? as class A now has two copies of class X in it, one that it needs to use, and one from class Y. This is the situation where header guards are of no use. My question is – is this purely a structural problem or is there a way around this?

I really would prefer to not include anything in class Y’s header file, in case I want to include THAT in anything else, but is it necessary because of the function prototype?

Answer 1

If you have the following:

X.h

#ifndef X_H__
#define X_H__

class X
{
public:
    int foo() { return 1; }
};

#endif

Y.h

#ifndef Y_H__
#define Y_H__

#include "X.h"

class Y
{
public:
    Y(X *pX) { myval = pX->foo(); }
    int myval;
};

#endif

something.cpp

#include "X.h"

...

something_else.cpp

#include "Y.h"

...

Then there should be no problem.

However, if X.h instead looks like this:

#ifndef X_H__
#define X_H__

class X
{
public:
    int foo();
};

int X::foo() { return 1; }

#endif

then you will indeed get a linker error when you try to link something.cpp and something_else.cpp. X::foo will have been defined non-inline into two separate translation units.