Re: c++/6006: problem with pure virtual functions and destructord

[Nathan Sidwell <nathan at acm dot org>]

jody@ifi.unizh.ch wrote:
> 
> Hi
> Is there a particular reason why a a virtual function called
> in a destructor is the function of this class and not the derived class?
because the standard says so.

> (There are implementations of C++ where the virtual
> function of the derived class is called)
those implementations are buggy.

> In my example, however, the function g() calls the pure virtual
> function f(). But g() can be called from the destructor of A,
> without causing a problem.
neither call 'works'. Drat, someone's borrowed my c++ std, so this is
from memory. in the dtor of A, the B object is not complete (B's
dtor has been executed, modulo the bits pertaining to its bases).
We shouldn't call (non-static) member functions of B, because
there's no constructed B object for their this pointer to point to.
The std says, that in this case, virtual functions dispatch to the
function seen in the definition of the base being ctor'd or dtor'd
(i.e. what the final overrider would be if that base was the complete
object).

IIRC, calling a pure virtual function is undefined, so anything can
happen. g++ 3.0 reports
6006.cc:37: abstract virtual `virtual void A::f()' called from destructor
because it can prove that that resolved to a pure virtual. It cannot
do so for the one embedded in g, but if you executed such a program
it would fail.

nathan

-- 
Dr Nathan Sidwell :: Computer Science Department :: Bristol University
           The voices in my head told me to say this
nathan@acm.org  http://www.cs.bris.ac.uk/~nathan/  nathan@cs.bris.ac.uk