Re: strange deadlock ...

[Darin Adler <darin bentspoon com>]

On Tuesday, July 17, 2001, at 04:58  PM, Michael Meeks wrote:

> 	I appreciate that :-) the thing is, that this is not some whacked
> out, unusual race condition - caused by some obscure use of threading,
> that will not affect anyone.
>
> 	On the contrary, it will cripple anyone using threads and Gtk+ eg.
> try adding a g_thread_init to gtk-demo, and then pressing "Dialog and
> Message Boxes" and then any button [ to cause a re-entering of the glib
> mainloop through gtk_dialog_run ].
>
> 	Bang, deadlock.

But this same syndrome existed in ORBit/gtk+/Bonobo all along. Programs 
like Nautilus had to defer work until idle from all incoming CORBA calls 
to avoid this kind of problem -- see 
nautilus/libnautilus/nautilus-idle-queue.c, 
nautilus/libnautilus/nautilus-view.c, and 
nautilus/src/nautilus-view-frame-corba.c for specifics. We never did solve 
the problem of the incoming unref call from the Bonobo side (like unref) 
that we have no way of deferring and other incoming calls from the Control 
interface.

I'll be happy to give more details on request. The fact that incoming 
CORBA can happen anywhere outgoing CORBA calls are made leads to trouble.

    -- Darin