Failures in logging have catastrophic consequences for Actors

[obrok]

Running the following code in irb produces a fatal: No live threads left. Deadlock? on 2.1.2. On jruby the process just hangs indefinitely.

require "concurrent"
Concurrent.configuration.logger = lambda do |level, progname, message = nil, &block|
  fail
end

class Bob < Concurrent::Actor::RestartingContext
  def on_message(msg)
    p :hi
  end
end

bob = Bob.spawn(:bob)

I also encountered this when the process was finishing and the logger used inside Concurrent.configuration.logger got closed before concurrent-ruby finished its cleanup, but I couldn't reproduce it in a code snippet.

[pitr-ch]

Thanks @obrok for letting us know! I'll add a rescue.

[obrok]

Thanks for the quick reaction