VC++ bug or not?

[snorkelman]

This code works fine with gcc 3.x and 4.0, but generates odd compile
time errors with VC++ 2003 .NET. Is there something subtle that's wrong
with it, or is it just a compiler bug?

struct A {
static const char c = 'A';
};

struct B : public A {
static const char c = 'B';
typedef A PARENT;
};

struct C : public B {
static const char c = 'C';
typedef B PARENT;
};

main()
{
cout << C::c;
cout << C::PARENT::c;
cout << C::PARENT::PARENT::c;
return 0;
}

For some reason, the compiler is unable to resolve the line of code
that attempts to root back through two typedefs to the base class A. It
tells me that "__ctor is not a member of B".

[John Carson]

<sn********@gmail.com> wrote in message
news:11**********************@g49g2000cwa.googlegr oups.com

This code works fine with gcc 3.x and 4.0, but generates odd compile
time errors with VC++ 2003 .NET. Is there something subtle that's
wrong with it, or is it just a compiler bug?

struct A {
static const char c = 'A';
};

struct B : public A {
static const char c = 'B';
typedef A PARENT;
};

struct C : public B {
static const char c = 'C';
typedef B PARENT;
};

main()
{
cout << C::c;
cout << C::PARENT::c;
cout << C::PARENT::PARENT::c;
return 0;
}

For some reason, the compiler is unable to resolve the line of code
that attempts to root back through two typedefs to the base class A.
It tells me that "__ctor is not a member of B".

VC++2005 beta 2 compiles and executes the code without a problem. Comeau
also accepts it. Appears to be a bug.

--
John Carson

[Victor Bazarov]

sn********@gmail.com wrote:

This code works fine with gcc 3.x and 4.0, but generates odd compile
time errors with VC++ 2003 .NET. Is there something subtle that's wrong
with it, or is it just a compiler bug?

struct A {
static const char c = 'A';
};

struct B : public A {
static const char c = 'B';
typedef A PARENT;
};

struct C : public B {
static const char c = 'C';
typedef B PARENT;
};

main()
Aside from the missing #include and 'using', and the return value type
of the 'main', there is nothing wrong with it.
{
cout << C::c;
cout << C::PARENT::c;
cout << C::PARENT::PARENT::c;
return 0;
}

For some reason, the compiler is unable to resolve the line of code
that attempts to root back through two typedefs to the base class A. It
tells me that "__ctor is not a member of B".

It's a bug. You might want to check with 'microsoft.public.vc.language'
to see if folks who use those compilers the most, know anything about it
or to see how to check if Microsoft knows about it. FWIW, VC++ 2005 Beta
doesn't compile it either.

BTW, all 'c' static consts have to be defined outside the respective class
definitions since they are used elsewhere in the program (in 'main').

V

[John Carson]

"Victor Bazarov" <v.********@comAcast.net> wrote in message
news:Ij******************@newsread1.mlpsca01.us.to .verio.net

sn********@gmail.com wrote:

It's a bug. You might want to check with

'microsoft.public.vc.language' to see if folks who use those
compilers the most, know anything about it or to see how to check if
Microsoft knows about it. FWIW, VC++ 2005 Beta doesn't compile it
either.

Beta 1? Beta 2 handles it without a problem.

--
John Carson

[Victor Bazarov]

John Carson wrote:

"Victor Bazarov" <v.********@comAcast.net> wrote in message
news:Ij******************@newsread1.mlpsca01.us.to .verio.net

sn********@gmail.com wrote:

It's a bug. You might want to check with
'microsoft.public.vc.language' to see if folks who use those
compilers the most, know anything about it or to see how to check if
Microsoft knows about it. FWIW, VC++ 2005 Beta doesn't compile it
either.

Beta 1?

Yes.
Beta 2 handles it without a problem.

Good. I keep forgetting to mention '1'.

V

[snorkelman]

Thanks all.. If it's fixed in Beta 2 I suspect the chances of any fix
in 2003 .NET are nil. I was mostly concerned that I was up against
personal ignorance of some standard scoping rules and was making a
gaffe.

I'll check the vc newsgroup for vendor-specific info.

|
| Mark Nelson markn at ieee.org
|