Unhandled socket timeout errors

[marcelotto]

In the Panpipe library (an Elixir wrapper for Pandoc that uses exile), users reported crashes under high concurrency: marcelotto/panpipe#8

[error] GenServer #PID<0.517.0> terminating
** (MatchError) no match of right hand side value: {:error, :timeout}
    (exile 0.12.0) lib/exile/process/exec.ex:78: Exile.Process.Exec.receive_fds/2
    (exile 0.12.0) lib/exile/process/exec.ex:40: Exile.Process.Exec.start/2
    (exile 0.12.0) lib/exile/process.ex:906: Exile.Process.exec/1
    (exile 0.12.0) lib/exile/process.ex:628: Exile.Process.handle_continue/2
    (stdlib 6.0) gen_server.erl:2163: :gen_server.try_handle_continue/3
    (stdlib 6.0) gen_server.erl:2072: :gen_server.loop/7
    (stdlib 6.0) proc_lib.erl:329: :proc_lib.init_p_do_apply/3

Looking at the code in lib/exile/process/exec.ex:78 the issue appears to be an unhandled timeout:

  {:ok, sock} = :socket.accept(lsock, @socket_timeout)

[akash-akya]

Hey @marcelotto thanks for reporting the issue. I'll check as soon as I get some free time.

[akash-akya]

@marcelotto @fmv1992 timeout should not happen actually since we are the one using it to send the data from the other end. It might happen if the new process spawn is taking too much time under a heavy load.

I can increase the socket accept timeout form 2s to 5s but there is nothing to handle as such.

@fmv1992 if possible, can you help to test the dev branch changes (timeout)?

{:exile, github: "akash-akya/exile", branch: "dev"}

[akash-akya]

I there a way I can reproduce this issue? a simple test case/script would be a great help!