[Olivier Abad <oabad@xxxxxxx>]

Le mar 23/09/2003 à 06:30, Guy Harris a écrit :
> That patch has a debugging printf in it - should that be checked in?

No. I checked in a temporary patch until I find the right solution.

> > My only problem is that it doesn't work on my Debian sid system (which
> > uses gtk+ v2.2.4), because GTK_CLIST(w)->clist_window has a bad value
> > (0x3ee) so the test always fails.
> 
> Are you using a binary built entirely on your sid system, or is that a 
> binary built on a system using an earlier version of GTK+?

It's built entirely on my sid system. An I reproduced this behaviour
yesterday on a RedHat 7.3 system.

> The clist_window structure member doesn't appear to be documented in 
> the GTK+ API documentation, either for 1.2[.x] or 2.x; unless there's a 
> documented macro that refers to it, that might mean the GTK+ developers 
> consider it something applications shouldn't access, and therefore 
> don't consider it a binary compatibility problem if they move it.

It isn't documented, but I found this (old) mail on gtk-devel-list :
http://mail.gnome.org/archives/gtk-devel-list/1998-October/msg00031.html
and this comment in gtkclist.h (from gtk+ v2.2.4) :
/* return the row column corresponding to the x and y coordinates,
 * the returned values are only valid if the x and y coordinates
 * are respectively to a window == clist->clist_window
 */
gint gtk_clist_get_selection_info (GtkCList *clist,

So it seems we have to make the window == clist->clist_window test. BTW,
the test is used and works fine in our gtk v1.2 code.

I'll ask on the gtk lists for this.

Olivier

P.S. : sorry for the duplicate mail, Guy, I forgot to Cc the list in my
first reply.
-- 
For perfect happiness, remember two things:
(1) Be content with what you've got.
(2) Be sure you've got plenty.