Re: Single threading and CORBA_Object return types

[Elliot Lee <sopwith redhat com>]

On Thu, 28 Jun 2001, Bill Haneman wrote:

> I am trying to understand some behavior I am seeing and would like
> confirmation that it is a consequence of ORBit's single threading, and
> solicit suggestions for dealing with it.
>
> I am obtaining an instance of a registry object from OAF, and then
> need to query it (in a client).  It appears to me that if the client
> calls any functions that return CORBA_Object references, the return is
> accomplished via a push-IIOP stream from the server to the caller.
>
> This works fine if I make one such call; and I get the reply without
> entering the corba event loop (which I do via bonobo_main()).  However
> if I make two such calls in my client without entering bonobo_main the
> second call blocks in
>
> __select()
> __DTOR_END__ ()
> giop_main_next_message_2 (blocking=1, ...)
> giop_recv_reply_buffer_use_multiple_2 (..., block_for_reply=1)
> giop_recv_buffer_use_2 (..., block_for_reply=1)
> <my client-side CORBA call is here in the stack>
>
> Is this a single-threading issue, as I expect?  If so, what is the
> recommended workaround so that my client can programmatically query
> the service?  The problem only arises with calls that return
> CORBA_Object derivatives, not calls that return simple data types.
>
> Believe me, I have been wading through gdb for awhile, my client-side
> call is not calling back into the requestor.  The block seems to occur
> when the service tries to return the CORBA_Object to the caller, but
> only the second time.

It sounds like Dietmar and Michael are both chasing red herrings - you
should be able to do

myobj = object_method(obj, ev);
myobj2 = object_method2(obj, ev);

Just fine. It's probably a bug.

-- Elliot
"It's a well-kept secret that KDE games are no longer toys but maturing
entertainment applications." - dot.kde.org