Hello NG,


I am curious why the resolution doesn't work in the following case.
$ g++ --version
g++ (GCC) 4.2.0


#include <iostream>
#include <cstdlib>

using namespace std;

// building double diamond
struct A { int x; };
struct L : A {};
struct R : A {};
struct C : L, R { int p; };
struct LL : C { int q; };
struct RR : C {};
struct D : LL, RR {};

void working() {
D d, * pd = &d; // d has 4 x-int's
C * pc[2]; // d consists of 2 C's
A * pa[4]; // and 4 A's

pc[0] = (LL *)pd; // LL branch to C
pc[1] = (RR *)pd; // RR branch to C
cout << pc[0] << endl << pc[1] << endl;

pa[0] = (L *)pc[0]; // L branch to A
pa[1] = (R *)pc[0]; // R branch to A
pa[2] = (L *)pc[1]; // L branch to A
pa[3] = (R *)pc[1]; // R branch to A

cout << pa[0] << endl << pa[1] << endl;
cout << pa[2] << endl << pa[3] << endl;

// assign all integers
for(int i=0; i<4; i++)
pa[i]->x = i;
}

void notWorking() {
D d;
//d.D::LL:C::L::A::x = 0; // <= this line
d.D::LL::q = 0;
//d.D::LL::C::p = 0; // and this line
}

int main() {
working();
notWorking();
return EXIT_SUCCESS;
}


g++ says:
error: ‘C’ is an ambiguous base of ‘D’
of cause it has 2 C bases, but the "path" to the p member is
no more ambiguous.

Is there something in the standard about this situation?
How do other compilers handle this code?

Regards, Daniel

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

Schüle Daniel ha scritto:
>
> // building double diamond
> struct A { int x; };
> struct L : A {};
> struct R : A {};
> struct C : L, R { int p; };
> struct LL : C { int q; };
> struct RR : C {};
> struct D : LL, RR {};
>
> void notWorking() {
> D d;
> //d.D::LL:C::L::A::x = 0; // <= this line
> d.D::LL::q = 0;
> //d.D::LL::C::p = 0; // and this line
> }
>
> g++ says:
> error: ‘C’ is an ambiguous base of ‘D’
> of cause it has 2 C bases, but the "path" to the p member is
> no more ambiguous.

The "path" information is discarded. When you write D::LL::C::L::A the
compiler replaces all that with A. So d.D::LL::C::L::A::x is the same as
d.A::x and that is ambiguous.

> Is there something in the standard about this situation?

It's in 3.4.5 in the standard.

> How do other compilers handle this code?

They should behave the same way.

Ganesh


--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

Schüle Daniel schrieb:
> Hello NG,
>
>
> I am curious why the resolution doesn't work in the following case.
> $ g++ --version
> g++ (GCC) 4.2.0
>
>
> #include <iostream>
> #include <cstdlib>
>
> using namespace std;
>
> // building double diamond
> struct A { int x; };
> struct L : A {};
> struct R : A {};
> struct C : L, R { int p; };
> struct LL : C { int q; };
> struct RR : C {};
> struct D : LL, RR {};

Are you sure that this is your structure? There is *no* diamond here at
all. Each D has two C's, and four A's. Do you probably mean "virtual
base class"?

So long,
Thomas


--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

> The "path" information is discarded. When you write D::LL::C::L::A the
> compiler replaces all that with A. So d.D::LL::C::L::A::x is the same as
> d.A::x and that is ambiguous.

thank you for explanation

Is there rationale behind that, such as simplification of the compiler
or language?

>> Is there something in the standard about this situation?
>
> It's in 3.4.5 in the standard.
>
>> How do other compilers handle this code?
>
> They should behave the same way.

Regards, Daniel

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

> Are you sure that this is your structure? There is *no* diamond here at
> all. Each D has two C's, and four A's. Do you probably mean "virtual
> base class"?

I am aware of the "virtual" inheritance, but it would reduce common data
members into one. I was trying to build something complex to see how
to reach all the contained members.

Regards, Daniel

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]