Re: Calling gtk_widget_destroy()

[Alexey Dokuchaev <danfe nsu ru>]

On Sat, Sep 11, 2004 at 11:55:19AM -0400, Stuart Brorson wrote:
> As for your problem: when you click "delete_event", gtk_main calls
> gtk_main_quit. Gtk_main_quit kills all its windows, and then returns
> to gtk_main. Then, gtk_main returns,  and you proceed to the next
> statement gtk_window_destroy.  However, all the windows are already
> gone and the memory has already been freed!  Therefore, you get a
> segfault.    

May I suppose this is not quite right what you're saying.  If
gtk_main_quit() would destroy all widgets (windows) unconditionally,
like you described, clicking a button would lead to segfault as well.
On practice, this does not happen.

My current vision of pretty much entire situation is:

  Gtk_main_quit() does not really call gtk_widget_destroy().  However,
  since "delete_event" callback function (here, gtk_main_quit()) does
  not return TRUE, gtk_widget_destroy() gets called later, supposedly
  in "destroy" default handler, or "delete_event" itself (I'm not quite
  sure at this time).  When gtk_main_quit() hooked to some other event,
  like "clicked", no one really cares what it returned, and no one calls
  for gtk_widget_destroy().  That explains (at least for me) why closing
  a window by clicking a button works, but closing it via windowmanager
  explodes.  (Here we speak of a program that calls gtk_widget_destroy()
  after gtk_main().)

How does it sound to you?

Thanks!

./danfe