Broken vala bindings (again)

[ricotz]

due to b52858b

There was a lengthy discussion regarding those annotations, see #214 and #208

If python is actually iterating over the elements of a GPtrArray to free them then this is wrong. It should be advised to use the available GLib API.

[tintou]

I raised this issue to GObject Introspection and there is a lack of annotation for this. The current way of dealing with this is to use g_ptr_array_set_free_func (ptrarray, NULL); before returning the GPtrArray…
See how GNOME Builder is doing this when returning a GPtrArray

[ximion]

I can't remove the free function, as this would break the C API.
Wow, this sucks... Can't either Vala, Python or GIR be fixed to address this?

[ricotz]

As a mitigation you can add a metadata file for vala to tweak the generated vapi.

[ximion]

That may be the best temporary fix...
At the moment it seems like I can choose between breaking all C users of libappstream, breaking Appcenter via Vala bindings, or breaking Python applications with the previous change that made Vala work (as reported in https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=974107 ).

[ximion]

How would I have to change the Vala metadata to fix this for Vala? I would like to go this route for the next release (scheduled for Monday), and maybe find a more permanent solution later.
Unfortunately, there doesn't seem to be a workaround for Python.