Simulating console activity in Java

Question

Hello,
I have a small Java program that basically uses telnet to connect to another server. Currently, the program basically acts like the regular telnet client. When I run it, I get the basic login screen that I normally get in the standard command line. I can input my username and password and login normally through the console of this program.
I want to automate this. Basically, I want the program to detect that there is a prompt and input the correct information accordingly. So when it sees the word "login:" it puts in a username etc.
I am using the apache.commons.net package to do this (as I was unable to find something else). The code was taken from http://www.java2s.com/Code/Java/Network-Protocol/ExampleofuseofTelnetClient.htm
The program uses a thread to display the text onto the screen. Here is the function that takes care of this:
    public void run()
        InputStream instr = tc.getInputStream();
        try
            byte[] buff = new byte[1024];
            int ret_read = 0;
            do
                ret_read = instr.read(buff);
                if(ret_read > 0)
                    System.out.print(new String(buff, 0, ret_read));
            while (ret_read >= 0);
        catch (Exception e)
            System.err.println("Exception while reading socket:" + e.getMessage());
        try
            tc.disconnect();
        catch (Exception e)
            System.err.println("Exception while closing telnet:" + e.getMessage());
    }Something needs to be changed in here I'm guessing, but I have no clue what to use to achieve this. I already have it implemented using a perl script but the organization I'm with wants to standardize everything and eliminate the dependencies for this application. Any help would be greatly appreciated.
Cheers
P.S. I wasn't sure if this is the correct forum to post this stuff into. Please move the thread if it belongs elsewhere

Answer 1

I have a bot that run inside of our subnet and monitor all our Solaris servers via telnet. The gist of the solution is to read from the stream and buffer the input until you see something you can parse. You then interpret the result and send the corresponding commands though the wire. The reason for the buffering is that the read can be discontinuous, so you might see "log" first and then follow by "in:" on the next read. The parsing won't be trivial, but you can just treat it like a simple state machine. Line length can also be problematic, as some commands could wrap its output arbitrarily, so a newline is not a reliable delimiting token.