[vtkusers] vtk51 cvs memory corruption in vtkInteractorObserver derived classes

[Mathieu Malaterre]

Andrew J. Dolgert wrote:
> Hi,
> 
> I'm running a cvs snapshot from 19 March under Visual Studio 2005 on
> Windows XP. I found a problem in how vtkInteractorObserver receives
> callbacks that leads to memory corruption. I think most compilers don't
> notice it, but Microsoft Visual Studio does throw exceptions under
> certain circumstances.
> 
> In its constructor, vtkInteractorObserver initializes its callback
> object as follows:
> 
> vtkInteractorObserver::vtkInteractorObserver()
> {
>   this->Enabled = 0;
>   this->Interactor = NULL;
>   this->EventCallbackCommand = vtkCallbackCommand::New();
>   this->EventCallbackCommand->SetClientData(this);
> 
>   <more stuff>
> }
> 
> Then a derived class, like vtkBoxWidget, registers its callback in its
> constructor:
> 
> vtkBoxWidget::vtkBoxWidget()
> {
>   this->State = vtkBoxWidget::Start;
>   this->EventCallbackCommand->SetCallback(vtkBoxWidget::ProcessEvents);
> 
>   <more stuff>
> }
> 
> In ProcessEvents, the pointer to vtkInteractorObserver (its "this"), is
> cast to a vtkBoxWidget. That's a problem.
> 
>   vtkBoxWidget* self = reinterpret_cast<vtkBoxWidget*>(clientdata);
>   self->OnLeftButtonDown();
> 
> The "this" pointer of a base class is not guaranteed to be the same as
> that of a derived class and often is shifted by a few bytes. Visual
> Studio is throwing a System.AccessViolationException. If the derived
> class were to call SetClientData(), then all would be well.

Andrew,

	Can you open a bug report for that at:

http://vtk.org/Bug

	Possibly could you add a small script/c++ code to reproduce the bug.

Thanks
Mathieu