[boost] Re: change to linked_ptr release

From: Mark Borgerding (markab_at_[hidden])
Date: 2000-02-07 14:30:01

"greg colvin" <gcolvi-_at_[hidden]> wrote: 
original article:http://www.egroups.com/group/boost/?start=2125
> I still think linked_ptr::release is a dangerous invitation
> to dangling pointers.
> 

Yes, I tend to agree with you, but I think that release is important to
have.

Maybe we can all be happy ...

I have an idea for defining the behavior of a ref-counted smart
pointer's release().  It would give the user the ability to claim
responsibility for deletion of the pointer, but only if no other
references share the object. i.e. unique() returns true.

so the definition for release becomes

T* linked_ptr::release()
{
  T * origPtr = 0;
  if ( unique() ) {
   origPtr = ptr;
   ptr = 0;// just give it up without deletion
  }else 
   reset(0);// take *this out of list, leaving 
   // responsibility for deletion to other nodes
  return origPtr;// null if not unique
}

This approach tends to err on the side of double deletion, as opposed
to dangling pointers.  This should be much easier to catch in
debugging, and possibly less dangerous.

returning 0 if the object is not unique allows safe code of the form 
delete mySmartPtr.release();

This approach could also be implemented for the existing shared_ptr.

And yet another bonus, this approach would get rid of the bool
"released" member that increases the size of linked_ptr.

--Mark