Member function templates instantiation in gcc

[MiKom]

Hi,

I have a problem with class member function templates. It seems that gcc is trying to analyze the code semantically too early, without explicit template instantiation.

Let the code speak

A.h:

#ifndef __A_H__
#define __A_H__
class A {
public:
	A* ptr;

	template<class T>
	void method() {
		((B*) ptr)->invoke();
	}
};
#endif 

B.h:

#ifndef __B_H__
#define __B_H__
#include<iostream>
class B : public A {
        public:
        void invoke() {
                std::cout << "lol" << std::endl;
        }
};
#endif 

C.h:

#ifndef __C_H__
#define __C_H__

class C : public A {
	
};
#endif 

C.cpp:

#include"A.h"
#include"B.h"
#include"C.h" 

main.cpp:

#include "A.h"
#include "B.h"
#include "C.h"

int main()
{
	A a;
	a.ptr = new B();
	a.method<int>();
	return 0;
}

In Visual studio 2010 this code comples fine. However, g++ seems to make too eager template instantiation because even when compiling just C.cpp:

g++ -o c.cpp.o -c C.cpp

I get the error about unknown identifier B:

In file included from C.cpp:1:0:
A.h: In member function ‘void A::method()’:
A.h:10:5: error: ‘B’ was not declared in this scope
A.h:10:7: error: expected primary-expression before ‘)’ token
A.h:10:9: error: expected ‘)’ before ‘ptr’

It seems, that g++ tries to analyse templated function member not only syntactically but also semantically, even though I didn't told it to (I never tried to instantiate the member function template).

Am I missing something? Which compiler is behaving correctly according to the standard?

[jsmith]

Visual Studio has the bug, not gcc. All symbols must be declared before being used, therefore B.h must include A.h and C.h must include A.h.

[MiKom]

Includes are not the problem here. As You can see in C.cpp all needed symbols are present. My concern is however about A::method(). Why gcc even tries to compile this method if I never requested template instantiation? (I'm talking about compiling only C.cpp). If gcc is tring to compile A::method() when compiling C.cpp, what does it use as template parameter T?

In main.cpp however, all symbols are also present in line 9 where I request template instantiation. Doesn't C++ standard state that templates are instantiated on request? If so, class B is present, declared and defined.

[jsmith]

gcc isn't trying to compile it. It is performing some simple syntactic checks. For all the compiler knows, A could be a typedef for int or the constant value 42 or a string or... in which case there is a syntax error at the declaration of class B.

EDIT: I'm barking up the wrong tree slightly. After looking at your error message more closely, the compiler is complaining about this: ((B*) ptr)->invoke(); not because of the derivation. But the problem is the same: A.h is not self-standing because it does not include a definition of B. B's definition may not come after A's definition ever, even when it is inside a template function. This happens because you include A.h before B.h, but you well can't change that order since B depends on A. The answer is to forward declare B in A.h.

[ne555]

Forward declare is not enough, (invalid use of incomplete type ‘struct B’)

//A.h
class A{
	template<class T>
	void method();  //declaration of methods
}

#include "B.h"
template<class T>
void A::method(){ //definition of methods that uses B 

[MiKom]

Wow, nice solution. Thanks.