wtf...? deriving from template

Code:

---

template <class T> class B {
protected:
  T *p;
  // even if replacing T with int doesnt help
};

template <class T> class D : public B<T> {
public:
  D()
  {
    p = NULL; // error ?!?!?!?!?!
  }
};

---

why does it say p is undefined?

(using gcc)
is it a syntatic problem?

Works for me using MS.Net 2003.

yes, that code works for me too on ms-vc6.0
but it fails to compile with g++ >_<
error:
test.cpp:18: error: `p' undeclared (first use this function)
test.cpp:18: error: (Each undeclared identifier is reported only once for each function it appears in.)

[edit]
so my first guess would be that its a syntatic problem and the ms compilers allow that syntax while the c++ standard does not... i might be wrong though

[edit2]
the funny thing is that when replacing B<T> with e.g. B<int> in line
template <class T> class D : public B<T>
it compiles...

oh well...

i found this "fix"
but would still be nice if anyone could tell me what the problem is,
btw: gcc - v yields
gcc version 3.4.3 20050227 (Red Hat 3.4.3-22.fc3)

Code:

---

template <class T>
class D : public B<T> {
public:
  using B<T>::p; // !!!

  D()
  {
    p = NULL;
  }
};

---

it could be a bug. I had a problem when overloading += operator recently in gcc
http://cboard.cprogramming.com/showthread.php?t=66655
but probaly you're getting that error because you're acessing the variable from a constructor, which shouldn't hold any problem. Make a method and try to use the variable in it, to se what happens

methods yield the same error.

so its either a bug, or in some documents it states that it has to be done that way... strange

well i could try a newer version of gcc (even though there are only 2-3 newer versions)

>why does it say p is undefined?
Because it is undefined. p doesn't depend on the template parameter T for class D, but class B does. The standard says that dependent base classes aren't searched when looking up non-dependent names. The solution is either to qualify p with B<T>:

Code:

---

template <class T> class B {
protected:
  T *p;
  // even if replacing T with int doesnt help
};

template <class T> class D : public B<T> {
public:
  D()
  {
    B<T>::p = NULL; // error ?!?!?!?!?!
  }
};

---

Or this:

Code:

---

template <class T> class B {
protected:
  T *p;
  // even if replacing T with int doesnt help
};

template <class T> class D : public B<T> {
public:
  D()
  {
    this->p = NULL; // error ?!?!?!?!?!
  }
};

---

Assuming you don't include a header that defines NULL, you also have an error there. :)

hehe, so
adding
using B<T>::p (posted earlier in the thread)
to my D class was just the right thing to do

thxalot! well my other guess was also right: there is some standard which states that this shouldnt compile
and it also shows that m$ doesnt follow it...

>there is some standard which states that this shouldnt compile
Yes

>and it also shows that m$ doesnt follow it...
Despite your apparent attempt to bash Microsoft, you won't find a compiler that completely conforms to the C++ standard, but Microsoft's is one of the few that come close.

now that i have fedora core 3 running, let me have the fun bashing ms :D
and ms needs to bashed for many reasons (like threatening asian governments who use linux)
and i don't wanna know what happens at the vote this wednesday when they decide about software patents in europe >_<

edit:
oops, this wednesday is tomorrow, i meant next wednesday

that code to a programmer isn't ambiguous, but if the standard says so, then we'll have to carry the book around whenever we want to use more complex syntax... :mad:

it's funny to see that the website for c/c++ standard, doesn't use standard html

What does the C++ standard have to do with the HTML standard?

Just remember that it's not a good idea to initialize base members in the derived constructor.