07:35 <kchibisov> How https://wayland.app/protocols/pointer-constraints-unstable-v1#zwp_locked_pointer_v1:request:set_cursor_position_hint supposed to work? Should compositor place cursor in that particular position?

20:16 <daniels> pq: errrr, untangling mappedness involves pulling apart a lot more of libweston-desktop than I'm really comfortable with right now

20:16 <daniels> I'm happy to start doing it gradually, but would like to push the fix to make it not crash atm

20:17 <daniels> pq: basically, subsurfaces will map themselves when they have a valid buffer but never unmap themselves

20:17 <daniels> all other shells will mostly just be mapped on the first commit call, whether or not they should be

20:18 <daniels> I'm having a go at a couple of approaches to fix this, but none of them seem like they'll be easy or quick - I'd rather try to deal with it bit by bit in a more easily bisectable & reviewable way

20:34 <daniels> weston_surface_unmap() does indeed unmap the views, but there's nothing really to ensure that surfaces are only mapped when we have buffers, and there's also no surface-mapped signal to let users only map views when the surface is mapped

20:34 <daniels> we could of course just handle that in ->committed(), but in terms of API design I think there's nothing wrong with having views mapped when surfaces are not mapped, as long as they're consistently ignored