Re: GtkSocket ...

[Owen Taylor <otaylor redhat com>]

Michael Meeks <michael@helixcode.com> writes:

> Hi all,
> 
> 	This bug was recently fixed in Bonobo Plug/Socket, it
> caused fatal BadMatch errors; be good to fix Gtk too. Problem
> solved by Peter Williams and Ettore Perazolli.
> 
> This will fix the intermitent BadMatch bug that crops up every so
> often. The problem was that we were trying to set the focus to the
> parent window, which was not guaranteed to be mapped and viewable.
> 
> The first few lines of our focus_out_event are now something like
> this:
> 
> {
> 	GtkWidget *toplevel;
> 	BonoboSocket *socket;
> 	XWindowAttributes attr;
> 
> 	g_return_val_if_fail (BONOBO_IS_SOCKET (widget), FALSE);
> 	socket = BONOBO_SOCKET (widget);
> 
> 	toplevel = gtk_widget_get_ancestor (widget,
> gtk_window_get_type());
> 	XGetWindowAttributes (GDK_DISPLAY (), GDK_WINDOW_XWINDOW
> (toplevel->window), &attr);
> 
> 	/* FIXME: can we just check the return value of 
> 	 * XGetWindowAttributes? */
> 	if (toplevel && attr.map_state == IsViewable)

Why create a race condition? The window manager could unmap the parent
of the window between the check and the unmap.

The correct thing to do in a case like this is do
gdk_error_trap_push() OPERATATION gdk_flush() gdk_error_trap_pop()
around the operation.

Regards,
                                        Owen