Garbage collection on dynamic source destruction ?

[Russsgithub]

Hello,

I'm using a altered version of this script (from the book, apart from the storing of shutdown callbacks)

shutdown_callbacks = ref([])

def play(pl)
  s = mksafe(playlist("playlists/#{pl}"))
  toggle = interactive.bool("switch", true)
  s1 = input.http("url")
  s1 = switch(track_sensitive=false, [(toggle, s1), ({true}, s)])
  s = output.icecast(%mp3, mount="#{pl}", s1)
  def shutdown()
    s.shutdown()
    s1.shutdown()
    toggle.remove()
    garbage_collect()
  end
  shutdown_callbacks := list.add((pl, shutdown), !shutdown_callbacks)
  "Playing #{pl}."
end

def stop(pl)
  shutdown = list.assoc(pl, !shutdown_callbacks)
  shutdown_callbacks := list.assoc.remove(pl, !shutdown_callbacks)
  shutdown()
  "Stopped #{pl}."
end

server.register("play", play)
server.register("stop", stop)

Before a source is started htop shows a small list of liquidsoap processes all at approx 0.0% cpu usage, and telnet list returns []

When I start a few sources all is fine a few more liquidsoap processes show up in htop with the bottom process showing higher cpu usage than the rest, telnet list shows the added sources.

but have noticed that when a source is stopped and shutdown the processes remains (htop) , cpu usage on all but the bottom process in htop drop to approx 0.0%, but the bottom one remains active at say 20% (on my machine), and telnet list still shows the removed sources (and in my case a switch).

The log for the script shows sources get down for all sources, and telnet reports source finished on each source

When I start another few sources the liquidsoap processes in htop increase again, some of which sit at 0.0% . This seems to continue. I tried adding garbage_collect() hoping to destroy these unused processes but they still hang around.

Is this intended , I'm a bit worried that if the script runs for a long time that it's going to turn into a resource eating monster :).

As always thanks for doing what you do @toots and @smimram .

I have had 50 sources with security fallovers and switches running , and when they are all removed I'm still left with the bottom liquidsoap process in htop using 70% cpu (split over all cores) while the script is outputting nothing.