llgo: back in business

I’ve been hacking away on llgo, on evenings and weekends when I’ve had the chance. It’s now roughly equivalent in functionality to where it was before I upgraded LLVM to trunk (3.1) and broke everything. There’s a couple of added bonuses too, like proper arbitrary precision constants, and partial support for untyped constants.

Now that the basics are working again, I’ll get back to working on the import/export mechanism. I expect this will expose more design flaws, and will take a while. I still plan to make use of debug metadata, which I am not altogether familiar with. I’ll also need to decide how the linker and the runtime library are going to work.

In other news, I’ve moved Pushy to GitHub. I’m not actively developing it at the moment, but I wanted to consolidate the services I’m consuming. I do have an addition to the Java API in the works: a sort of remote classloader, that will communicate over a Pushy connection to fetch classes/resources. The idea is to make it really quick and easy to run a bit of Java code on a remote machine, without having to deploy the application remotely. I’ll hopefully get around to pushing this change within the coming few weeks.

Posted February 11, 2012. Tags: go, llgo, llvm, pushy.

Java Pushy API

It’s been some time since I’ve spruiked Pushy, so here we go. One of my colleagues was talking the other day about how he had implemented a netcat-like service for STAF, an automation framework aimed at testing. This got me thinking about how this could be done relatively easily, using the pushy.net package, something I haven’t written about much, primarily because it’s still a work in progress.

Back in version 0.3, I introduced a Java API to Pushy, as I mentioned in an earlier post. I briefly mentioned the incorporation of packages which mimic, and in many cases are interface compatible with, packages in the Java standard library such as java.io and java.net.

The pushy.net package currently contains three main classes:
  • RemoteInetSocketAddress (extends java.net.InetSocketAddress)
  • RemoteSocket (extends java.net.Socket), and
  • RemoteServerSocket (extends java.net.ServerSocket).
RemoteInetSocketAddress simply provides a means of creating an InetSocketAddress whose address is resolved at the remote host. RemoteSocket is a wrapper around a remote Python socket object, but extends java.net.Socket to provide a familiar interface to Java developers. Similarly, RemoteServerSocket extends java.net.ServerSocket, and wraps a remote Python socket object.

So how about netcat emulation? Well, I won’t cover the whole implementation of a netcat clone, as that would be a non-trivial undertaking. But I will show you one of the fundamental requirements: to bind a server socket, accept a client connection, and print out the data received from that client.

Step 1. Bind a socket, and listen for connections.

import java.net.ServerSocket;
import java.net.Socket;
import pushy.Client;

public class Test {
public static void main(String[] args) throws Exception {
Client conn = new Client(“local:“);
try {
ServerSocket server = new pushy.net.RemoteServerSocket(conn, 0);
try {
} finally {
} finally {

In this code snippet, we’re creating a RemoteServerSocket, and assigning it to a java.net.ServerSocket, to illustrate interface compatibility. The first argument to the constructor is the pushy.Client object we previously created, and the second argument is the port to bind to. Specifying a port of zero, means that we want to bind to an ephemeral port.

The creation of the RemoteServerSocket involves creating a Python socket object on the remote host, and performing the bind and listen methods on it.

Step 2. Accept a client connection.

Socket client = server.accept();
try {
} finally {

Here we can see that accepting a client connection is exactly as we would do with a standard java.net.ServerSocket. This probably isn’t surprising, since we’ve upcasted our RemoteServerSocket to a plain old ServerSocket. One thing of interest here is that the Server object returned is in fact a RemoteServerSocket, wrapping a remote Python socket object.

Step 3. Read data from the client connection.

InputStream in = client.getInputStream();
byte[] buf = new byte[1024];
int nread = in.read(buf, 0, buf.length);
while (nread != -1) {
System.out.write(buf, 0, nread);
nread = in.read(buf, 0, buf.length);

Et voila! We can read the remote socket’s output via a java.io.InputStream object, returned by the getInputStream method, which is overridden by RemoteSocket. One thing you may have noticed: to run this all on the local host, sans Pushy, you could substitute the right-hand side of the initial ServerSocket construction with a standard ServerSocket, and the rest of the code would remain unchanged.

There are a few defects in the 0.3 release related to the pushy.net package, which will prevent these examples from working. I have rectified them in the process of writing this post. If you grab the trunk, it should all work nicely. There is one defect remaining: the InputStream returned by RemoteSocket is based on a a file returned by Python’s socket.makefile method. This differs from the InputStream returned by Java’s standard Socket, in that a request for N bytes will not return until all N bytes, or EOF, are received. I hope to have this fixed in the coming days.

Posted February 10, 2011. Tags: java, pushy.