Created attachment 194957 [details] [review]
Avoid special handling of GObject

It turns out that most of the special-casing of GObject.Object inside scanner is not needed any more.  This patch removes most of the special-case handling of GObject and GInitiallyUnowned, and this also results in 'notify' signal being present again inside Object.

Created attachment 194958 [details] [review]
Diff of GObject-2.0.gir before and after patch

This is diff between GObject-2.0.gir generated before and after patch, illustrating the impact fo the patch.

Review of attachment 194957 [details] [review]:

This looks good to me. 
* The abstract flags appears to be wrong (gobject.c doesn't set them), so I guess it's good that they were removed
* The removal of the writable flag for g_type_instance is probably okay as well, as it should be private

As long as the tests pass we should commit this. Shouldn't break anything, but pushing it to master will get more testing

Review of attachment 194957 [details] [review]:

Committed.  Thanks for the review!

Tested in gjs and now it finds "notify" signal on the GISignalInfo of GObject. However, the type of the first arg is GI_TYPE_TAG_INTERFACE so it does 
            info = g_type_info_get_interface (type_info);
            info_type = g_base_info_get_type (info);
but that info_type is OBJECT so it will do a g_value_get_object() which asserts because the GValue has type "GParam".

given that using GObject::notify in gjs obviously works, is this bug still valid?

Closing this bug report as no further information has been provided. Please feel free to reopen this bug if you can provide the information asked for.
Thanks!

[Mass-moving gobject-introspection tickets to its own Bugzilla product - see bug 708029. Mass-filter your bugmail for this message: introspection20150207 ]