Re: Fly-by-night clandestine removal of refcount check! Aroooga!

[Miguel de Icaza <miguel gnu org>]

> I just had another idea.  We could add a flag to BonoboObjects like
> this:
> 
>     gboolean i_have_been_qid_before;
> 
> Once an object has been qid, we make bonobo_object_add_interface
> fail.  This would be guaranteed to work correctly.
> 
>     Reactions?  Impressions?  Does anybody just want to talk about
> their childhood?

Well, there is a genuine use for aggregate at runtime.

The only actual case we worry about is replying "no" and then replying
"yes" to a query interface request.   But being able to grow the
supported interfaces of the object at runtime is a good thing,
speciallly if you can do this dynamically.

So we only want to make sure that we are consistent in our "no"
returns (hence the "query_interface" signal), but we need to have
stronger detection of broken code in our QI handler code.

Miguel.