Timeouts still occur, causing slow connects with RUN_NOWAIT

[bagder]

As mentioned in https://github.com/NEAT-project/neat/pull/147#issuecomment-252902539: if we build and run the client_http_run_once example, it will always trigger (what seems to be) a 1000 milliseconds timeout for the connect operation. Even though everything else is working fine.

[bagder]

If a timeout is necessary for driving neat, the neat_get_backend_timeout() really ought to return a correct (very short) value instead of 1000 milliseconds.

[karlgrin]

@bagder I am not sure what you are referring to. As far as I can tell, the timeout doesn't drive client_http_run_once. Instead, it seems to be driven by events on the file descriptors of the respective libuv event loop.

[karlgrin]

@bagder Did some further test runs, and running "./client_http_run_once -n 200 bsd10.fh-muenster.de" took less than 900 ms. I am not in a position to say whether on not that is fast or slow.