Re: GET_SORT_COLUMN_ID segfault

[muppet <scott asofyet org>]

On Dec 27, 2006, at 6:19 PM, Jens Luedicke wrote:

> On 12/27/06, muppet <scott asofyet org> wrote:
>
> > sort_column_id and order are return parameters, pointers to the
> > location at which the values should be stored.  The code in
> > gtk2perl_tree_sortable_get_sort_column_id() expects them never to be
> > NULL (because it doesn't make sense to call the function if you
> > aren't going to get them), and is segfaulting when trying to
> > dereference a null pointer.  However, we should be more robust than
> > that...  In gtk+ HEAD, the code in gtk_tree_view_column_update_button
>
> I understand the C part of your explanation, but the column and order
> are predefined in my custom ListStore and not undefined.

Whether they are defined in your custom store is irrelevant;  
gtk_tree_view_column_update_button() (a private function inside  
GtkTreeViewColumn) is calling the iface's vfunc with a NULL because  
it doesn't care about the order value.  So, you're returning it  
properly, but when the binding goes to store the value you returned,  
it's attempting to dereference an invalid pointer, passed in by other  
C code (not your perl code).

So, basically, just a binding bug.