Re: Wrap GtkShortcutsWindow? ABI break in gtk+ 3.20?

[Murray Cumming <murrayc murrayc com>]

On Fri, 2016-02-26 at 09:10 +0100, Kjell Ahlstedt wrote:
> When I regenerated gtk_signals.defs and tried to rebuild gtkmm, I
> noticed something worrying.
> The gtk+ folks have added a new parameter to the
> GtkPlacesSidebar::show-other-locations signal. This signal was added
> in gtk+ 3.18 and gtkmm 3.18. Now a new parameter will be required in
> the signal handlers in gtk+ 3.20. Isn't that an ABI break both in
> gtk+ and gtkmm? Signal handlers made for gtk+ 3.18 will not find the
> user_data parameter where they expect it. It's admitted in the patch
> that it's in API break, but I think it's also an ABI break in gtk+.
> Am I right or wrong?
>
> See https://bugzilla.gnome.org/show_bug.cgi?id=754743 and 
> https://git.gnome.org/browse/gtk%2B/commit/?id=63a204e

Possibly, yes. I guess that the vfunc is not a symbol that is
identified by the linker, so that's OK. But I wonder what would happen
when the function pointer is called with the now-wrong number of
parameters. I asked about that:
https://bugzilla.gnome.org/show_bug.cgi?id=754743

-- 
Murray Cumming
murrayc murrayc com
www.murrayc.com