Gtk_container_cell_widget_set should chain up to its parent's set
function, not its parent's unset function. This was resulting in
accessibles being erroneously marked defunct after being created.
Old code assumed that AT-SPI would keep track of references and
therefore tried to only hold weak references. On the other hand it also
tried to keep objects alive so it referenced objects very randomly. All
of that lead to cycles and leaking.
As AT-SPI does not keep track of objects at all, the treeview now does.
The refcounting looks as follows:
GtkTreeViewAccessible
=> creates per row/column
GtkTreeViewAccessibleCellInfo
=> which references 1
GtkCellAccessible
If there is only one cell, this accessible is a
GtkRendererCellAccessible, otherwise a GtkContainerCellAccessible is
created and that accessible holds references to the
GtkRendererCellAccessibles that are created for every cell renderer.
https://bugzilla.gnome.org/show_bug.cgi?id=554618
This commit exposes the get_type() functions and standard
headers for accessible implementations. This makes it possible
to derive from the GTK accessible implementations without
GType magic tricks. This is necessary, because we require the
a11y type hierarchy to be parallel to the widget type hierarchy.
So, if you derive a widget and need to adjust its a11y implementation,
you have to be able to derive its accessible implementation.
This commit probably exposes more than is absolutely necessary,
it also exposes accessibles of widgets that are unlikely candidates
for deriving from.
Move instance fields to a private struct, in preparation
for installing a11y headers.
This also required removing access to GtkContainerCellAccessible
innards from the GtkCellAccessible implementation.
I could have tried to make GtkContainerCellAccessible implement
GtkCellAccessibleParent, but the current implementation of that
interface doesn't make sense for it.
