The feature from legacy/trac#3313 does a prctl or ptrace call if it's on during options_act(), and does a log_notice if it's off during options_act. So I can change it to off, hup, and get the log_notice saying attaching should work. But it doesn't actually do anything to make it work again. rransom points out quite rightly that our process isn't *supposed* to be able to do anything to make it work again. So the correct fix is to stop deceiving the user. Perhaps we should make it an invalid transition in options_transition_allowed()? Also, should we really be calling tor_disable_debugger_attach() on every hup? Seems like the sort of thing that we should have a static variable to remember. And finally (and related), if the option is off, you get your log_notice on every hup whether anything changed or no.