Re: parent while in INIT_INSTANCE

[Torsten Schoenfeld <kaffeetisch gmx de>]

On 19.08.2013 03:12, Kevin Ryde wrote:
> With recent debian i386 glib 2.36.3 and gtk 2.24.20 the foo.pl below
> dies with
>
>   label Gtk2::Label=HASH(0x9fe6cf8)
>   label now packed
>   parent MyContainer=HASH(0x9e5d770)
>   return from INIT_INSTANCE
>   GLib-GObject-CRITICAL **: g_object_notify_queue_thaw: assertion `g_atomic_int_get(&object->ref_count) > 
> 0' failed at foo.el line 26.
>   GLib-GObject-CRITICAL **: g_object_notify_queue_thaw: assertion `g_atomic_int_get(&object->ref_count) > 
> 0' failed at foo.el line 26.
>   object 9e61008 is not really a GObject at foo.el line 26.
>
> It seems to be due to taking $label->parent() within INIT_INSTANCE.  Or
> at least if you don't do that it's ok.  Is that supposed to work?  It
> worked in the past.

I would need to think longer about this to find out whether this is
supposed to work.  But in meantime, it seems likely that the change in
behavior is due to
<https://git.gnome.org/browse/perl-Glib/commit/?id=eb6ff1b1e6e7fb5dcef4aa131cae9d95d2049460>.
 Can you verify that (perhaps with "git bisect")?