Thread: Stack corruption ?

Hi everyone,

Now every error seems weird, but this one _really_ does to me ;-)

I have to classes, A and B. B is an interface for let's say B_a and B_b.
'A' contains a pointer to 'B', pointing to an allocated object of either 'B_a' or 'B_b' (via some factory), and the 'B-object' contains a pointer to A (some pimpl implementation).
Class 'A' also has a derived class, therefore some of it's member functions are virtual (and here's the problem).

Code:

class A;

class B {
 public:
  B(A* p_a) : a_(p_a) {}
  virtual void read( 'file ') = 0;
 protected:
  A* a_;
};

class B_a : public B  // class implements 'read'
class B_b : public B  // class implements 'read' in a similar way as B_a (see below)


class A {
 public:
  A() : _b(new B_b(this)) {}  // or new B_a(this) , this is done in some factory method
  void read( 'file' ) { b_->read('file'); }
  virtual void f();
  virtual void g();
 protected:
  B* b_;
};


void B_a :: read( 'file' ) {
  // do a lot of things specific for B_a (or B_b)
  a_->f(); // someting weird happens here

  // do a lot of other things specific for B_a (or B_b)
  a_->g();
}

Now the weird thing is : the call 'a_->f()' does _not_ call 'A::f()' , but it call 'B::g()' !!

If I remove the keyword 'virtual' (temporarily) for 'A::f()' it works, but then it produces a similar error for the call 'a_->g()' . Unless I remove the keyword 'virtual' ....

** I have verified the adresses of the pointers, that's all correct. I am debugging the correct code, I verified with some 'asserts'.
** calling 'a_->f()' from another context just just just before 'B_a::read', does not produce the problem.

Is there somewhere a stack corruption ? If so, how could I detect it ?

Thanks a lot !!!

There is no function B::g() in what you posted.
Non-compileable bits like ('file') serve no useful purpose in this example. You may as well just leave out that parameter.
Please fix your example.