Re: Glib::wrap with copy=true semantics?

[Murray Cumming <murrayc murrayc com>]

On Fri, 2005-03-04 at 10:03 -0800, Philip Langdale wrote:
> Murray Cumming wrote:
> > 
> > Like I said, we don't refcount GtkAdjustments, and I don't recommend
> > that you do.
> 
> The whole point is that passing copy=true to the Glib::wrap call is
> that it does increment the refcount which is then not decremented
> when the wrapper is deleted.

And I explained why you shouldn't do that.

>  So it's doing something with the refcount.
> 
> I'm just asking whether this is the intended semantics of Glib::wrap
> or not. It is definitely unexpected for the wrapper to come and go
> and leave the ref count at +1.

The take_copy parameter is only for use with RefPtr<>.

> I have no problem passing copy=false in which case it won't touch
> the refcount but the documented semantics for copy=false are that
> the wrapper takes ownership of the object which is not what is going
> on here.

OK. A documentation patch would be welcome.

-- 
Murray Cumming
murrayc murrayc com
www.murrayc.com
www.openismus.com