Re: DnD through in-proc plug/sockets

[Owen Taylor <otaylor redhat com>]

On Thu, 2004-01-22 at 05:24, Alexander Larsson wrote:
> On Wed, 2004-01-21 at 19:28, Owen Taylor wrote:
> > On Wed, 2004-01-21 at 12:22, Alexander Larsson wrote:
> > > I have a problem with dnd in nautilus. It seems that drag_motion in the
> > > icon and list views are called with the wrong coordinates. Tracing
> > > through the recursive calls in gtk_drag_find_widget() it seems
> > > everything goes wrong in the BonoboPlug (which is a GtkPlug
> > > essentially). The coordinates returned from gdk_window_get_position for
> > > it seems to be the absolute position of the plug.
> > > 
> > > I don't know what the right thing to do here is? Why is it returning the
> > > absolute position? Why did this work before? Should we just check for
> > > !GTK_IS_PLUG in addition to !GTK_WIDGET_NO_WINDOW in
> > > gtk_drag_find_widget? 
> > 
> > This sounds like some sort of GDK bug ... gdk_window_get_position()
> > should certainly be returning relative coordinates for that window,
> > which when is reparented into a local window, is basically just
> > a GDK_WINDOW_CHILD.
> 
> That was the problem. It wasn't a GDK_WINDOW_CHILD, it was a
> GDK_WINDOW_TOPLEVEL, so the configure notify event handling set its
> position. The reason it was a toplevel and not a child was that gtkplug
> creates it thus, and then reparents it into the socket, and the
> reparenting just isn't turning it into a child. The cause of this is the
> patch from bug #117579. 
> 
> I'm not sure of all the details about toplevels, but I'm pretty sure
> this is the right patch:
> 
> Index: x11/gdkwindow-x11.c
> ===================================================================
> RCS file: /cvs/gnome/gtk+/gdk/x11/gdkwindow-x11.c,v
> retrieving revision 1.198
> diff -u -p -r1.198 gdkwindow-x11.c
> --- x11/gdkwindow-x11.c	21 Dec 2003 16:37:43 -0000	1.198
> +++ x11/gdkwindow-x11.c	22 Jan 2004 10:20:00 -0000
> @@ -103,7 +103,7 @@ static gpointer parent_class = NULL;
>  
>  #define WINDOW_IS_TOPLEVEL(window)		   \
>    (GDK_WINDOW_TYPE (window) != GDK_WINDOW_CHILD && \
> -   GDK_WINDOW_TYPE (window) != GDK_WINDOW_TOPLEVEL)
> +   GDK_WINDOW_TYPE (window) != GDK_WINDOW_FOREIGN)
>  
>  GType
>  gdk_window_impl_x11_get_type (void)

Looks like a fine fix to me.

Regards,
							Owen