Re: the callback situation (fwd)

[Owen Taylor <otaylor redhat com>]

ERDI Gergo <cactus cactus rulez org> writes:

> On Sun, 29 Jul 2001, Bill Haneman wrote:
> 
> > > was the deadlock situation of proc 1 calling a corba method in proc 2 that
> > > calls back corba methods in proc 1 ever resolved?
> > 
> > I think [hope!] so
> 
> Martin told me on #gnome that this issue is solved a long time ago by
> making ORBit2 multi-thread.
> However, my application crashes at the first time the out-of-proc
> component it operates on tries to call back to a bonobo object in the
> original process.
> I couldn't find a GNOME 2 test program that uses out-of-proc components
> and also callbacks, so I am testing it with
> bonobo-clipboard/tests/singleton-test. Please see if this works for you.

 ORBit1: when calls out are made, the main loop is run, so any
   event processing and incoming calls will be handled.
 ORBit2: when calls out are made, only the CORBA file descriptors
   are handled, not other types of event processing.

Even with the ORBit2 version, this allows for all sorts of
nasty reentrancy problems that people are most likely not
handling properly 98% of the time. But there should be no
problems with incoming callbacks being processed.

AFAIK, ORBit2 has no support for the multi-threaded POA modes,
so threading isn't involved in the solution here.

Regards,
                                        Owen