well of course that would help if disorderfs helps ;e) we can merge this and get reproducible builds. i'd hold off upstreaming though, because this doesn't address the underlying question of how a file gets a different bytecode based on its compilation order