Hello,

I posted this question months ago, with no answer. Now I'm back on this same task. I need to be able to start a graphics application from either a rsh or a rlogin with nobody logged into the console. We are using Red Hat 9, init 5, on all our systems. How can I invoke a graphics application that is passed data via the network to display graphics on the console screen without someone having to physically log on to the console first? I know we can log on and run xhost +, but we need to be able to fire up an application on many systems that use the console graphics without someone being logged in. This is on about twenty systems that are rack mounted, and we just need to use the graphics outputs.
Any possible solutions are welcome! This is a high priority for me to resolve soon. These systems are used just to output graphics displays. Oh, all have Nvidia cards, and we're using KDE on most if not all.

Thanks in advance for any help!!
Merlin

I'm not 100% sure I understand what you want. If I try to give an example, tell me if it matches what you want to do...

You are logged into a machine, lets call it bertie. You have other machines, lets call one of them jeeves.

jeeves is running a local X server with KDE as the desktop environment.

You want to rsh from bertie to jeeves, and run a program which displays on jeeves' X server? For example

...and you want to do this for lots of machines which are set up like jeeves?

You're basically correct. I have a graphics application that executes on system "jeeves" and gets all it's inputs over the network. I have a interface system "bertie" that I have another application that gathers data, computes eyepoint direction and position/orientation and passes that over the network to system "jeeves".
I would like "jeeves" to be sitting there idle with no physical log-on, and accept a signal (ssh/rsh/etc.) from "bertie" to start up his own local copy of the application, and output to his "normal" console output screen which would be his physical output from his graphics card.
I can do this now, but I have to physically log into "jeeves" to start the window manager session, and then execute "xhost +" to allow a graphics application to be invoked and gain access to the screen.
This may not seem like much extra work, but we have 16 systems to log into, and soon 32. Each time I get up to leave the room, I have to log out of each system. When I come back I have to log back into each system. My goal is to make these "jeeves" systems stand-alone turn-key display servers. The problem is getting access to the display without a user being physically logged on the console.

Any Thoughts??

Thanks,
Merlin

When you log into jeeves using rsh/ssh/whatever, are all the command executed on the command-line? e.g.

If so, there are a few approaches.

1. Using public key authentication with ssh, you can execute remote commands without needing to enter a password. This means you can script commands from bertie, like this:
   Code:

   #!/bin/bash
   ssh user@jeeves command1
   ssh user@jeeves command2
   ssh user@jeeves2 command1
   ssh user@jeeves2 command2

2. You can use expect to automate logins, execute commands and conditionally perform tasks according to the output of those commands.

If however, you need to do stuff with the GUI, you'd need an X progam which does something like winrunner. I'm not sure about that. It's not a certain cause for depression though - many KDE applications can be instructed to do GUI things from the command line via the dcop interface. If course this requires that the app you want to control has a dcop interface and that it is designed to do what you want in a script-able manner.