Bad argument in arithmetic expression

[slashdotdash]

Environment

• Elixir: v1.6.6
• Operating system: Docker container running alpine-elixir

Current behavior

We've encountered the following exception numerous times in production error reports when using the pigeon library for sending large batches of FCM push notifications (1,000 notifications):

Elixir.ArithmeticError: bad argument in arithmetic expression
  File "lib/connection/processor.ex", line 131, in Kadabra.Connection.Processor.process/2
  File "lib/connection.ex", line 155, in Kadabra.Connection.handle_info/2
  File "gen_server.erl", line 616, in :gen_server.try_dispatch/4
  File "gen_server.erl", line 686, in :gen_server.handle_msg/6
  File "proc_lib.erl", line 247, in :proc_lib.init_p_do_apply/3
  Module "Elixir.Kadabra.Connection", in Kadabra.Connection.init/1

Unfortunately I haven't seen it happen locally, nor have I been able to reproduce it.

I've looked through Kadabra.Connection.Processor module and am wondering whether the issue is because the default max_concurrent_streams setting in the Kadabra.Connection.Settings module is :infinite but line 131 is assuming it will be a number:

to_ask =
  settings.max_concurrent_streams - flow.stream_set.active_stream_count

Could that be the culprit for the exception?

[hpopp]

Thats probably it exactly. If you notice in the def block directly above:

  # nil settings means use default
  def process(%Frame.Settings{ack: false, settings: nil}, state) do
    %{flow_control: flow, config: config} = state

    Egress.send_settings_ack(config.socket)

    case flow.stream_set.max_concurrent_streams do
      :infinite ->
        GenServer.cast(state.queue, {:ask, 2_000_000_000})

      max ->
        to_ask = max - flow.stream_set.active_stream_count
        GenServer.cast(state.queue, {:ask, to_ask})
    end

    {:ok, state}
  end

Weird that I didn't make the same validation. I'll have a patch up.

[hpopp]

Should be fixed with v0.4.4. Let me know if it crops up again.

[slashdotdash]

@hpopp Thanks for the rapid fix and release. I will upgrade to v0.4.4 today ready for deployment.