[06:31] <anpok> bschaefer: you are probably on the flight already..
[06:31] <anpok> bschaefer: btw there is a way in libinput already to inject an fd.. the actual problem with that approach will be dealing with revoking the fds and applying stuff like relative coordinates..
[06:32] <anpok> RAOF: ^..
[06:32] <anpok> bschaefer: libinput gets a function pointer struct on init.. that can be used to inject fds..
[06:33] <anpok> so we could totally load the libinput platform in nested and non-nested mode.. and I guess apart from the device monitor everything stays the same..
[06:33] <anpok> i am off..
[06:34] <anpok> RAOF: but how would we ensure the raw device fd is revoked on focus lost..
[23:11] <RAOF> anpok: EVIOCREVOKE!
[23:12] <RAOF> anpok: Specifically - there's an evdev ioctl to revoke access to the evdev fd; we'd need to hook focus up to that.
[23:12] <RAOF> And we don't want relative coordinates, anyway.