dns-async raises exception "writer fd unexpectedly closed".

[1yefuwang1]

After cloning, I type 'make test' on command line, and the following is what I get:

2017-10-16 14:19:16.434280+08:00 5.153.225.51
2017-10-16 14:19:16.440170+08:00 Error ("Exception raised to [Monitor.try_with] that already returned.""This error was captured by a default handler in [Async.Log]."(exn(monitor.ml.Error("Writer error from inner_monitor""writer fd unexpectedly closed "(writer((id 2)(fd((file_descr 5)(info(replaced(("connected to"8.8.8.8:53)(previously_was(socket((family((family PF_INET)(address_of_sockaddr_exn )(sexp_of_address )))(socket_type SOCK_DGRAM)))))))(kind(Socket Active))(supports_nonblock true)(have_set_nonblock true)(state Closed)(watching((read Not_watching)(write Not_watching)))(watching_has_changed false)(num_active_syscalls 0)(close_finished Empty)))(monitor(((name(id 16))(here())(id 16)(has_seen_error false)(is_detached false))((name try_with_or_error)(here())(id 9)(has_seen_error false)(is_detached true))))(inner_monitor(((name(id 17))(here())(id 17)(has_seen_error true)(is_detached true))((name try_with_or_error)(here())(id 9)(has_seen_error false)(is_detached true))))(background_writer_state Stopped_permanently)(background_writer_stopped(Full()))(syscall Per_cycle)(bytes_received 28)(bytes_written 0)(scheduled_bytes 0)(scheduled_back 0)(back 0)(close_state Open)(close_finished Empty)(close_startedEmpty)(num_producers_to_flush_at_close 0)(flush_at_shutdown_elt())(check_buffer_age(((writer )(maximum_age2m)(bytes_received_at_now_minus_maximum_age 0)(bytes_received_queue())(times_received_queue())(bytes_seen 0)(too_old Empty))))(consumer_left Empty)(raise_when_consumer_leaves true)(open_flags(Full(Ok(rdwr nonblock))))(line_ending Unix))))("Caught by monitor (id 16)"))))

Sorry for the bad formatting, I don't know how to pretty-print s-expressions. The message which stands out is ''writer fd unexpectedly closed" . So I did some digging and my guessing is that it is caused by closing fd prematurely before the corresponding writer and reader are closed. The problematic code is:

(* async/async_dns_resolver_unix.ml *)
let cleanfn ?log sock () =
  begin
    try Socket.shutdown sock `Both
    with exn -> Option.iter log ~f:(fun log -> Log.error log "%s" @@ Exn.to_string exn)
  end;
  Monitor.try_with_or_error (fun () -> Unix.close @@ Socket.fd sock) >>| function
  | Error err -> Option.iter log ~f:(fun log -> Log.error log "%s" @@ Error.to_string_hum err)
  | Ok () -> ()

where only the fd is closed, not the corresponding writer nor reader.

[1yefuwang1]

I have just submitted a pull request to fix this issue. #147
Hope someone review it.

[avsm]

the implementation of the 2.0 branch has totally changed in #159, and we need a new async implementation in #161. Closing this in favour of that issue