Re: Bonobo Unique Application v4

A Ter, 2003-04-29 às 16:21, Michael Meeks escreveu:
> Hi Gustavo,
> 
> On Tue, 2003-04-29 at 14:15, Gustavo J A M Carneiro wrote:
> > 	I changed the API a bit, to allow for registration of messages with
> > name, return type, argument types.  I try to map GType's  to
> > CORBA_TypeCode's, to make a more GLib-ish API.  You think it's better to
> > deal with typecodes directly?  It would give us more power (all CORBA
> > types available), but...
> 
> 	It's fine to stick with GLib types for this I guess; we can always use
> the CORBA interface directly if necessary.

   Yes, however I'm trying to use BONOBO_TYPE_CORBA_ANY with no
success.  I think bonobo-arg.c does not handle this GType.

> 
> 	I didn't read in depth; but a few things to think on:

  Good.  That means you didn't catch the bugs I discovered later :)

> 
> 	va_arg (args, GType):
> 
> 		I _think_ you have to have a builtin type as the 2nd arg 
> 	on a number of compilers / platforms; how does G_VALUE_COLLECT do
> 	this ?
> 

  But GType is a builtin type, after the typedefs are dereferenced.  Is
it still a problem?  See this in gtype.h:

#if	GLIB_SIZEOF_LONG == GLIB_SIZEOF_SIZE_T
typedef gulong                          GType;
#else	/* hm, shouldn't happen? */
typedef gsize                           GType;
#endif


> 	Should we not add a closure (and have a convenience G_CALLBACK addition
> helper) along with the 'message description' ? that would seem necessary
> to have the method to callback into - of course; it's nice having the
> type-safe argument lists - that's really good.
> 
> 	Some examples of the BonoboClosure stuff are in
> libbonoboui/bonobo/bonobo-ui-component.c - along with the helpers FWIW.

  I have a problem with closures.  I never could understand them
completely, that's why I try to avoid them :)  But I'll look at it some
more.

> 
> 	Looking nicer anyhow.

  Thanks.  I was about to try to setup a 'new-instance' message (and
also as signal of BonoboApplication), which used CORBA_any to send the
argv, but I'm struggling with CORBA_any as GValue problems.  I hope to
get it fixed soon..

> 
> 	One of the things I want to do is to fix bonobo-activation, so that the
> component's capabilities are defined by the .server file - instead of a
> (hapazard) set of (not thread-safe) properties that people will always
> forget to set ;-) That'll take a little re-structuring I think, but it
> has to be the right way to go; then registering our unqiue application
> we'd pass a fragment of XML to say how 'unique' it is ;-)
> per-screen/per-display/per-lang etc. ;-)
> 
> 	Great work though,
> 
> 	Regards,
> 
> 		Michael.
-- 
Gustavo J. A. M. Carneiro
<[email protected]> <[email protected]>