Re: Possible bug in gdk_window_get_device_position ?

[salsaman <salsaman gmail com>]

The device value was not NULL when the crash occured, I checked.

There are also similar crashes in gdk_device_warp(),
gdk_device_get_position(), and gdk_device_get_window_at_position().


I also tested with code like:

GdkDeviceManager *devman=gdk_display_get_device_manager(disp);
GList *devlist=gdk_device_manager_list_devices(devman,GDK_DEVICE_TYPE_MASTER);

...

        for (k=0;k<g_list_length(devlist);k++) {
          GdkDevice *device=(GdkDevice *)g_list_nth_data(devlist,k);
          if (gdk_device_get_display(device)==disp&&
              gdk_device_get_source(device)==GDK_SOURCE_MOUSE)
            mouse_device=device;
        }

If I use mouse_device then it also works.



Regards,
Salsaman.


http://lives.sourceforge.net
https://www.ohloh.net/accounts/salsaman


On Sat, Feb 16, 2013 at 2:53 AM, Lex Trotman <elextr gmail com> wrote:
> On 16 February 2013 16:07, salsaman <salsaman gmail com> wrote:
> > Hi,
> > with gtk 3.4.4 I get a crash with code like:
> >
> > gboolean button_callback (GtkWidget *widget, GdkEventButton *event,
> > gpointer user_data) {
> >   gint x,y;
> >   GdkWindow *xwin=gtk_widget_get_window(widget);
> >   GdkDevice *device=gdk_event_get_source_device((GdkEvent *)event);
>
> Which IIRC can return NULL, so you will crash in those situations,
> test your return values :)
>
> >   gdk_window_get_device_position (xwin,device,&x,&y,NULL);  // crash !!
> >
> > }
> >
> > There is not even a stack trace, it just instantly quits the app.
> >
> >
> > However, if I replace:
> >
> > GdkDevice *device=gdk_event_get_source_device((GdkEvent *)event);
> >
> > with:
> >
> > GdkDeviceManager *device_manager =
> > gdk_display_get_device_manager(gdk_window_get_display(xwin));
> > GdkDevice *device = gdk_device_manager_get_client_pointer (device_manager);
>
> This doesn't return NULL because it always can find a display.
>
> > then there is no crash. Is this a bug ?
> >
> > In the documentation for gdk_device_manager_get_client_pointer(), it says:
> > "You should use this function sheldomly (sic), only in code that isn't
> > triggered by a GdkEvent..."
>
> Well, it works in the situation you happened to use it in, but that
> doesn't guarantee it works in all situations.  For the documentation
> to warn you not to use it, there only needs to be one situation in
> which it fails.
>
> > http://developer.gnome.org/gdk3/stable/GdkDeviceManager.html#gdk-device-manager-get-client-pointer
> >
> >
> > If it is not a bug then the documentation is somewhat misleading.
>
>
> It would be a bug and the documentation misleading if it worked in all
> circumstances and was guaranteed by the developers that it would keep
> doing so.  Neither of those conditions is known to be true.
>
> Cheers
> Lex
>
> > Regards,
> > Salsaman.
> >
> >
> > http://lives.sourceforge.net
> > https://www.ohloh.net/accounts/salsaman
> > _______________________________________________
> > gtk-list mailing list
> > gtk-list gnome org
> > https://mail.gnome.org/mailman/listinfo/gtk-list