From: Brian Candler <B.Candler@...>
Date: 2008-09-08T22:05:10+09:00
Subject: Re: Telnet client

> I use ruby wrote a Telnet client application. It works fine until users
> want to change a user name or telnet to other host. I found all these
> failed operations are related with "su -" command or login. How can I
> let ruby deal with these? Below is my core code for this telnet client:
> 
> while 1 == 1
>   c=gets
>   tn.cmd(c){|c| print c}
>   if c=="exit"
>     break
>   end
> end
> 
> It simply get what user input and put it on the telnet server.

I am having to guess here, but I imagine that tn is an instance of a
Net::Telnet object?

You haven't described exactly what your problem is. In what way do your
"failed operations" fail? What exactly do the users type, what do they see
in response, and what do you expect them to see?

I can see a few problems with the above code though.

* 'gets' receives a whole line at a time. If you try to use an interactive
  program like vi, it will fail. If that's not what you want, have a look
  at Curses, HighLine, or ruby-termios.
  http://codeidol.com/other/rubyckbk/User-Interface/Getting-Input-One-Character-at-a-Time/

* In the third line above, under ruby 1.8 the block parameter 'c' is bound
  to the same variable as the 'c' outside. This means that the value of 'c'
  will have changed after the block has executed. This means that if c
  was equal to 'exit' before, it may not be equal afterwards if the block
  has been invoked one or more times.

* c will never equal "exit" because you are reading whole lines with their
  terminator. It will be "exit\n".

So a cleaned-up version of your code might be:

  loop do
    c = gets
    tn.cmd(c) { |resp| print resp }
    break if c == "exit\n"
  end

Having said that, it would be better to detect if the far end drops the
connection rather than seeing if the user typed "exit". The user could
legitimately type "exit" for reasons other than logging out; and they could
logout by typing "logout".

I'm not sure the best way to do that with Net::Telnet. If you're waiting for
a specific response then you can test whether the value returned from 'cmd'
matches the string you were expecting. Otherwise, you can try sending data
and rescuing the Errno::EPIPE you get if the other end has closed.

I'd say that Net::Telnet is really designed for scripting telnet sessions,
not for writing your own interactive telnet client.

HTH,

Brian.