[nanomsg] Re: NN_ERRFD proposal

  • From: Martin Sustrik <sustrik@xxxxxxxxxx>
  • To: nanomsg@xxxxxxxxxxxxx
  • Date: Mon, 25 Feb 2013 07:02:11 +0100

On 24/02/13 10:57, Martin Sustrik wrote:

My proposal here is to remove NN_ERRFD socket option altogether and
signal both NN_RCVFD and NN_SNDFD in the case of error. That would
unblock any polling function, irrespective of whether user was polling
for IN or OUT. Subsequent call to nn_send() or nn_recv() would then
return error.

The behaviour is somewhat ugly, but, on the other hand, it saves up to 2
file descriptors per socket.

OK. New behaviour pushed to the master. Please note that it may break the bindings, in case they were already usign NN_ERRFD constant.

Martin

Other related posts: