[Bug c++/16636] Unable to emit vtable info

["miko at stacken dot kth dot se" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From miko at stacken dot kth dot se  2004-07-22 12:52 -------
(In reply to comment #4)
> If you still think that this is a bug, you will have to present us with 
> a complete example where this is actually a problem. In your example, 
> you don't have any user of the vtable, so there doesn't seem to be a bug 
> with not emitting it. Please show us a complete example that actually 
> fails to link. 

Based on comments made by Andrew Pinski that the code isn't legal, I don't
think I can argue this is a bug, per se.

> If you can show us an example where this actually does lead to a problem,  
> we may reconsider. In general, however, the rule to emit the vtable with the 
> first non-inlined function seems to be OK with everyone, though. 

Ok, I did some more playing around. It turns out that the example given
(which was obviously a condensed version of what I'm running into) DOES
emit a vtable with -fno-default-inline. However, if class B 

class B : public A
{
public:

  int b;

  virtual void foo() { ++b; }

  virtual
  ~B () { }

  B() : A(), b(0) { }
};



becomes:

class B : public A
{
public:

  int b;

  virtual
  ~B () { }

  virtual void foo();

  B() : A(), b(0) { }
};

(note the reordering of the functions)

then no vtable is emitted with -fno-default-inline. This would appear to
be inconsistent behaviour. If -fno-default-inline is enough to force a vtable
emission when all fns are inline, you'd have thought it enough to force it when
the first is inline, since we're stating not to inline functions by default.

Yes/no?



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16636