[omniORB] omniORB 3/4 create_POA deadlocks

[Teemu Torma]

On Tuesday 22 October 2002 16:59, Duncan Grisby wrote:
> On Monday 14 October, Teemu Torma wrote:
> > There is actually one scenario that might fail--if we are waiting
> > for a child to destruct and someone destroys us.  Since we do not
> > hold any locks at the time, we should probably also increment our
> > ref count to avoid any nasty problems.
>
> I don't think that can happen unless the application is broken can
> it? The application ought to be holding a reference to us until the
> destroy() call has returned, so our reference count should be at
> least one.

No, I can not think any situation when this could happen in real life.  
It was more of an thought what might happen, but the program is broken 
in the first place if the caller does not guarantee that the poa really 
is there.

Teemu