Re: How to simulate window close button (what I'm really trying to do...)

[Toralf Lund <toralf procaptura com>]

Murray Cumming wrote:
> On Mon, 2007-12-17 at 11:37 +0100, Toralf Lund wrote:
>   
> > OK, maybe I should tell people what I'm really trying to do - see my 
> > other post. [ ... ]
> >
> > In any case, I'm on to something that has sort of bugged me for years; 
> > in all my Gtk/Gtkmm window I generally have a button called "Close" or 
> > "Cancel", or a "Close" menu item, depending on the context, that is 
> > supposed to do *exactly* what a window manager close or "window delete" 
> > will do. It seems to me that this isn't exactly uncommon, so why is 
> > there no way of telling Gtk(mm) exactly what I want? I mean, why can't I 
> > just say, "please let this button do a window delete", and forget about 
> > the rest? Or can I do this? [ ... ]
> >
> >
> >     
>
> You already need to connect a signal handler for delete_event so you
> can, for instance, show your do-you-really-want-to-close dialog. If you
> also have a [Close] button then I'd just connect that to the same signal
> handler.
>
> The only reason that this is not quite so simple is that delete_event
> uses the nasty return-true/false-to-do-something system. But you can
> still abstract that out quite easily to a separate function, so you
> have 
>
> void on_button_close()
> {
>   if(ask_for_confirmation())
>     hide();
> }
>
> bool on_delete_event()
> {
>   return ask_for_confirmation();
> }
>
> or similar.
>   
Yeah, I know, but like I said, it's always bugged me that I have to do 
something like this. It seems to me there ought to be a simple and 
direct way of saying "this button should delete the window", and/or "do 
a window delete now"...

Or maybe I don't care that much, but it has at least surprised me that 
I've never been able to find a ready-made setup for the close behaviour,
and I've been thinking I must have overlooked something. So thought I 
might ask after this came up again recently because I had made a mistake 
and called hide() directly for a Close button where the delete event 
setup had other stuff...
> Emmitting signals directly, for a class that you didn't implement, seems
> like messing with the internals of that class.
>   
Well, perhaps it's not so much the signal emit as such I want, but 
rather simulation of a user action, i.e. a way of saying something like 
"pretend this button was pressed" or "execute the button press logic 
now". Which is not the same thing as duplicating the calls that will 
actually be done...

- Toralf