Bernie Maier wrote:
> The app I am working needs to occasionally show an arbitrary number
> of web pages. In the windows port we use that dreaded thing that is
> truly an operating system component not a transparent addon to gain
> world domination and crush competitors. That allows as to create
> browser components on the fly, destroy them when they're not needed,
> and later recreate them again.
>
> In the Linux port, I've now got Mozilla to successfully be created
> on the fly, but I'm not convinced it's cleaning up after itself when
> I destroy the GTK embedding widget. It's not the obvious problem
> document on the API pages where after calling NS_TermEmbedding you
> cannot call NS_InitEmbedding in the same process. I had this
> problem, but then added my own call to gtk_moz_embed_push_startup()
> which ensures that NS_TermEmbedding is not called when I destroy
> the widget.
>
> What happens now is that after I create an instance of the
> embedding widget, then after a while destroy it, I get GTK
> warnings, such as:
>
> Gtk-WARNING **: invalid cast from `(unknown)' to `GtkObject'
>
> Gtk-WARNING **: invalid cast from `(unknown)' to `GtkObject'
>
> Gtk-CRITICAL **: file gtksignal.c: line 541 (gtk_signal_emit): assertion
> `gtk_type_is_a(GTK_OBJECT_TYPE (object), signal->object_type)' failed.
>
> What appears to be happening is that gtk_mozarea_get_toplevel_focus() is
> being called, but since there is no longer any Mozilla widget, GTK's
> type casting is going awry. Eventually, since we are presumably somehow
> still referring to objects that have been destroyed, this results in a
> segmentation fault.
>
> So, is there any further cleanup required apart from just destroying the
> GTK embedding widget? Specifically, what is this "mozarea" GTK class, and
> why does it keep hanging around (assuming that's what's actually happening).
>
> Cheers,
>
> Bernie
>
I assume that you're using 0.9? That was a bug that was in that release
and it's definitely the problem that you're describing exactly.
There's a patch in:
http://bugzilla.mozilla.org/show_bug.cgi?id=79022
And my rpms for 0.9 on ftp.mozilla.org include the patch. It's also
been fixed in the trunk so 0.9.1 will include it. The trunk also
includes a lot of other fixes as well.
--Chris
