🚀 go-pugleaf

Joi Ellis [TeamB] <joi@aravox.com> wrote in message news:3ca86f83_1@dnews...
> Owen Green <owen@begonefoulmachine.kennulz.demon.co.uk> wrote:
> > Hi,
>
> > I'm having a bit of trouble reliably detecting when a socket has been
> > disconnected from the other end. I've seen various opinions around the
place
> > that a read or write should result in an exception when this has
happened,
> > but the code below fails to produce one (at least on my system which is,
> > gack, w95).
>
> As I understand it, when a remote socket shuts down normally, your IO
streams
> will see EOF markers, not exceptions.  You'll get an exception if the
> socket dies without being properly closed, IE the remote program crashed
> and didn't clean up, the host crashed, or some device tossing packets
between
> the two hosts has stopped cooperating.  That's when you get a socket
> exception.
That makes sense.

> Don't you have to read once before you can get an EOF indication?
It appears not.
> What happens if you alter the test as follows?  (Look for //### comments.)
It doesn't compile (println not being a method of OutputStream) <g>

It behaves as expected (i.e. reads those bytes out and then returns -1s).

I suppose I could use some sort of latch so EOFs will only have attention
paid to them after the first successful read, but it's a bit hacky. If the
far end is suppossed to send a TCP FIN or somesuch when it closes gracefully
it seems a bit silly that one can't reliably detect it.

Cheers for your help,
Owen