The crash always happen due to to nmap-6.40-allresolve.patch
but can be made to crash without the patch by passing "-6" in
the command line, so the patch is not really wrong, just exposes
the issue.

  Steps to reproduce:
1. Boot with "ipv6.disable=1" as kernel parameter
2. Run "nc -6 localhost"

Adding -6 just for it to not try ipv4 before.

  Expectd output would be the same as:

$ nc -4 localhost
Ncat: Connection refused.

but what happens is:

$ nc -6 localhost
Socket troubles: Address family not supported by protocol
nc: engine_select.c:353: get_evmask: Assertion `sd >= 0' failed.
Aborted (core dumped)

  The fix probably should involve not adding to the event list
in nsock/src/nsock_connect.c:nsp_add_event() if it failed to
connect, or, nsock/src/engine_select.c:iterate_through_event_lists()
check the event and not call process_event() if the (connect_event)
event has event_done set, but would need to check too many states
for that.

  Likely fix would be in pseudo patch:

       if (!nse->event_done) {
         assert(nse->iod->sd >= 0);
         socket_count_read_inc(nse->iod);
         socket_count_write_inc(nse->iod);
         update_events(nse->iod, nsp, EV_READ|EV_WRITE|EV_EXCEPT, EV_NONE);
       }
+      else if (nse->iod->sd < 0) {
+          /* do something else, update logs, etc */
+          msevent_delete(nsp, nse);
+          /* do something else, update logs, etc */
+      }
+      else
       iod_add_event(nse->iod, nse);