Re: ORBit2 source clarification, please?

[Michael Meeks <michael ximian com>]

Hi Diego,

On Fri, 2002-05-10 at 02:45, Diego Sevilla Ruiz wrote:
> 	I've been playing around with ORBit2-cpp (orbit-cpp-cactus
> branch) and I encountered some weird error when the client code of a C++
> CORBA program calls orb->shutdown. I have seen that in the corba-orb.c
> file, the following test is applied:
> 
> if (((ORBit_RootObject)orb)->refs != 2 + leaked_adaptors) {
> 
> (line 1116 of current HEAD), so my question is what does this "2" means?
> 
> 	As a fact, all the C++ programs come to this line with a "refs"
> value of "1", which causes the following message to be printed:

	Fascinating; the 2 is a magic number indeed - I think it's probably the
combination of the CORBA_ORB reference [ on itself ] and perhaps thee
RootPOA reference, or something else that's still left around at that
stage. Essentially if you wrap CORBA_ORB_init / shutdown you shouldn't
change the way the referencing works - I believe.

> ** (process:22791): WARNING **: CORBA_ORB_destroy: ORB still has 0 refs.
> 
> 	So, is there any known patter (bad pattern, of course) which
> leads to this situation? If I had a better idea of what could be the
> cause, I could fix the c++ branch in this regard.

	Um; no it's a bit of a mystery; start with an extremely simple program
that just inits the orb and shuts down, and see what's up. Alternatively
get that exit method to always dump the live references [ export
ORBIT2_DEBUG=refs AFAIR ], and see which 2 we normally leave and find
why with C++ we don't have them around.

	Oh, and a comment there would be great to explain the 2.

	Regards,

		Michael.

-- 
 mmeeks@gnu.org  <><, Pseudo Engineer, itinerant idiot