Alan Bateman wrote:
> Sent: Thursday, September 29, 2011 16:30
> To: Jeroen Frijters
> Cc: net-dev@openjdk.java.net
> Subject: Re: non-blocking SocketChannel connect broken on Windows 8
>
> Jeroen Frijters wrote:
> > Hi,
> >
> > While testing on Windows 8 I found a bug in
> SocketChannelImpl.connect() when configured non-blocking.
> >
> > Here's repro:
> >
> > import java.net.*;
> > import java.nio.channels.*;
> >
> > public class test
> > {
> > public static void main(String[] args) throws Exception
> > {
> > SocketChannel sc = SocketChannel.open();
> > sc.configureBlocking(false);
> > sc.connect(new InetSocketAddress("google.com", 12345));
> > }
> > }
> >
> > When run on the Windows 8 Developer Preview this throws the following
> exception:
> >
> > Exception in thread "main" java.net.SocketException: Invalid argument:
> getsockname
> > at sun.nio.ch.Net.localInetAddress(Native Method)
> > at sun.nio.ch.Net.localAddress(Net.java:389)
> > at sun.nio.ch.SocketChannelImpl.connect(SocketChannelImpl.java:639)
> > at test.main(test.java:10)
> >
> > When you look at the code the problem is obvious. The connect method
> tries to get the localAddress while the connection is still pending, it
> should only do this after the connection attempt has finished.
> >
> This must be behavior change in Windows 8 as I don't think we've seen
> this before (as we have several tests that do non-blocking connects).
> The usual thing when you attempt to connect is that the operating system
> should first bind to a local address (if not bound already) before it
> attempts to establish the connection. I've no doubt that this will break
> other things too.
It's definitely a behavior change in Windows 8 and I agree that it probably has
some compatibility impact, but fundamentally I think that the OpenJDK code is
wrong, unless there is documentation somewhere that claims that the socket will
be bound when the async connect is initiated.
Regards,
Jeroen
