Created attachment 121834 [details] [review]
Patch

> --- gobject/pygtype.c	(revision 974)
> +++ gobject/pygtype.c	(working copy)
> @@ -83,6 +83,23 @@
>      return py_type;
>  }
>  
> +int
> +_wrap_g_type_wrapper__set_pytype(PyGTypeWrapper *self, PyObject* value, void *closure)
> +{

Shouldn't that be a static function?

I'd really like to see this get into pygobject.

We have a getter already, so it make sense to add a setter as well.

Created attachment 131142 [details] [review]
updated patch

Here's an updated version of John's patch with the following changes:

* ChangeLog entry
* consistent formatting
* made the setter a static function
* setter typechecks for None/type objects (throws a TypeError if invalid)
* added two testcases and associated helper methods

Comment on attachment 131142 [details] [review]
updated patch

I *think* that the reference counting is wrong here, if value is non NULL it should be Py_INCREF:ed and g_type_set_qdata_full's destroy notifier should Py_DECREF it.
Shouldn't matter too much in practice though, since the GTypes are likely to survive anyway.

Committed in revision 1030.