[omniORB] _CORBA_ObjRef_Element Self-assignment

[Felix Nawothnig]

Let's hope this I'm reporting a real bug this time. :-)

The assignment operator in _CORBA_ObjRef_Element don't check for
self-assignment - so I currently have a crash in the duplicate call() in

inline T_element& operator= (const T_element& p) {
if( pd_rel ) {
T_Helper::release(pd_data);
T_Helper::duplicate(p.pd_data);
pd_data = p.pd_data;
}
else
pd_data = p.pd_data;
return *this;
}

The _CORBA_ObjRef_Member::T_member& operator= (T_ptr p) implementation
looks suspicious too... I wouldn't expect that by assigning the actually
stored _ptr it now contains a freed (invalid) reference - but that might
actually be okay.

Cheers,

Felix

[Duncan Grisby]

Post by Felix Nawothnig
The assignment operator in _CORBA_ObjRef_Element don't check for
self-assignment - so I currently have a crash in the duplicate call() in

Yes, you're right. The code should check if the object reference in the
element is the same as the existing one, and not release it if so. I've
checked in the fix to svn.

[...]

Post by Felix Nawothnig
The _CORBA_ObjRef_Member::T_member& operator= (T_ptr p) implementation
looks suspicious too... I wouldn't expect that by assigning the actually
stored _ptr it now contains a freed (invalid) reference - but that might
actually be okay.

There's nothing wrong with that. The member is required to take
ownership of the pointer it's given. If that leads to the object
reference being released, it's a bug in the application.

Cheers,

Duncan.