Thread: friend declarations and scope

Code:

class myClass;

class otherClass {
public:
    otherClass(int x): valor(x) {}

    inline int sum(myClass &obj);
private:
    int valor;
};

class myClass {
public:
    myClass(int x): val(x) {}

    friend otherClass::sum(myClass&);
private:
    int val;
};

inline int otherClass::sum(myClass &obj) {
    return obj.val + valor;
}

The code serves only to ilustrate my question.

I believe C++ Primer (Lippman and others) fails miserably when explaining friend function declarations. A small glitch, I'm sure, on an otherwise awsome book. Or maybe I'm not following this rather small section correctly. But while it explains to:

- Define the function only after the class granting friendship is defined.
- Define the friend class before the class granting friendship is defined.

However, it didn't mention that obviously the function declaration in otherClass presents myClass. And that name can't possibly be in scope yet because I'm forced to only declare it later.

Previous chapters taught me well enough and so it wasn't hard for me to understand all I needed to do was to declare myClass (not define it) and thus introduce it to the scope. The code now compiles and executes correctly when tested.

My question is though (since i'm not used to C++ Primer containing flaws), is this the correct way? Or am I missing something here?

I believe the use of a forward declaration here makes sense, though there's a syntax error in the friend statement since otherClass::sum's return type is omitted.

You are quite right laserlight. I didn't realize the error before you mentioned it. Quite curious the fact VC++ 2005 doesn't even issue a warning.

minGW, on the other hand, was quick to catch it.

Quite curious the fact VC++ 2005 doesn't even issue a warning.

Could be backwards compatibility (combatibility?) with earlier versions of the compiler, such that an omitted return type defaults to int.

Could be. Got the irony

However, VC++ 2005 does not accept default int anymore. At least it does not on any other circumstance so far (function and object declarations, both). Well... truth the matter is that it throws a Compiler Warning C4430 that is treated as an error (combatibility, I know). However, I do have the compiler configured for level 4 warnings. And on this particular case that warning is not being thrown.

I'm getting inclined to believe this to be a bug.

For anyone interested, this seems to be a bug. I presented the problem on the MSDN Forums and one of the MVPs apparently "acknowledged" something is askew.

http://cboard.cprogramming.com/showthread.php?t=78889

I'm fully aware of that thread