🚀 go-pugleaf

John Arisco wrote:
> 
> Ranko Orlic <rorlic@eps.agfa.be> wrote:
> 
> >Using MSVC 4.0 compiler I get the following:
> 
> >File test.cpp: ---------------------------------
> 
> >class A {
> >public:
> >       class L {
> >       public:
> >               L() {}
> >               virtual void f() {
> >                       int dummy = 0;
> >               }
> >       };
> >       A();
> >};
> 
> >class B : public A {
> >       class L : public A::L {
> >       public:
> >               L() {}
> >               virtual void f() {
> >                       A::L::f();
> >               }
> >       };
> >       B();
> >};
> 
> >Compilation result: ----------------------------
> 
> >test.cpp(18) : error C2352: 'A::L::f' : illegal call of
> >nonstatic member function
> >Error executing cl.exe.
> >test.obj - 1 error(s), 1 warning(s)
> [snip]
> You cannot call a non-static member function without specifying
> an object.  The way you've written this, there is no way for B::L::f()
> to set its "this" pointer.  Of course, if this is really what you want
> 
> to do, the declaration of B::L::f() is unnecessary, since B::L
>  inherits the A::L::f() virtual method from its base class.
> 

That's nonsensical.  We're talking class declarations here.  
B::L::f is a non-static member function of B::L, so it has a 
this pointer.

I moved A::L out of A, making it just L, and I changed B::L to 
B::BL, and it compiled fine.  Then I thought, could it be the 
names were causing a problem?  So, I moved L back into A, but 
left BL as is.  I got the original error.  I also moved BL out 
of B, leaving L nested in A.  I got the original error.  
Finally, I made both functions f non-inline.  I got the 
original error.  Clearly, the problem has to do with deriving 
from a nested class.

I don't think that should be a problem, but clearly VC doesn't 
handle it.  So, I decided to try another compiler.  Compiling 
under Silicon Graphics' C++ (NCC) v4.0.  This isn't enough to 
justify my interpretation that it should work, but it is 
comforting.

BTW, I did a cursory examination of the Jan DWP and found 
nothing to suggest that it shouldn't work.

-- 
Robert Stewart		| My opinions are usually my own.
Datalytics, Inc.	| stew@datalytics.com