/dev/null for graphics

Have the following ever happened to you?

You are trying to automate some data processing step. The step in question involves some complex, GUI-driven application, for which there is no source available or that source is a little to complex do dive in and extract the required functionality. As a happy coincidence, the application includes batch mode - but it still creates and displays the gui when running in batch mode (even when started with "nogui" option).

(in my particular situation, the app in question was Brainstorm, a collection of Matlab scripts. The original author is using Windows exclusively, thus he probably couldn't even imagine this scenario)

All of the above could be acceptable, but there is a slight problem - batch processing happens on a mainframe-style Linux server, without keyboard, display, or even X server running. Thus the application starts up, tries to create it's cosy GUI, obviously fails, and crashes hard.

When applications output some unneeded stuff to stdout, we have a universal solution - we send it to /dev/null. Turns out that Linux is flexible enough to do the similar thing with graphics!

We'll still need some sort of X server. In my case, I didn't have full control of the target machine, so I used what was already installed - Xvnc. Most probably, Xvfb (virtual framebuffer) will work better, since it was designed almost exactly for this use-case, but I wasn't able to install it.

The following small script showcases the usage - we launch X server, launch our program, wait for it to finish, and then kill X server.
#!/bin/bash
Xvnc :17 &
xvnc_pid=$!

# Wait for X server to come online
# Usually, it starts up much faster, but there are occassional stalls
sleep 10

export DISPLAY=:17

matlab -nosplash -nodesktop -r "addpath('build/'); \
  try; $1; exit(0); catch e; disp(e.message); disp(e.stack); exit(1); end"
exit_value=$?

kill $xvnc_pid

exit $exit_value

Comments

Post a Comment

Popular posts from this blog

How to create your own simple 3D render engine in pure Java

Image
3D render engines that are nowdays used in games and multimedia production are breathtaking in complexity of mathematics and programming used. Results they produce are correspondingly stunning. Many developers may think that building even the simplest 3D application from scratch requires inhuman knowledge and effort, but thankfully that isn't always the case. Here I'd like to share with you how you can build your very own 3D render engine, fully capable of producing nice-looking 3D images. Why would you want to build a 3D engine? At the very least, it will really help understanding how real modern engines do their black magic. Also it is sometimes useful to add 3D rendering capabilities to your application without calling to huge external dependencies. In case of Java, that means that you can build 3D viewer app with zero dependencies (apart from Java...
Read more

Solving quadruple dependency injection problem in Angular

Angular comes with built-in dependency injection and module management system, but it can get a bit tedious to use when your project grows. Especially if you use RequireJS alongside it (a must for almost any project) and also want to be minification-safe with all those dependency injections. Let's look at a simple example with one module that provides a value and another module that prints that value: // answer.js define(["angular"], function(angular) { angular.module("the.answer") .value("TheAnswer", 42); }); // main.js define(["angular", "the.answer" ], function(angular) { angular.module("main", [ "the.answer" ]) .run([ "TheAnswer" , function( TheAnswer ) { console.log(TheAnswer); }]); }); As you can see, to provide a single dependency to my code, I had to specify it in four places (highlighted in red) - first as RequireJS dependency, then as Angular module dependency,...
Read more

Configuration objects in Scallop

It turned out that spreading type annotations throughout all the code in all places where options are requested is not a good idea. And, of course, it is prone to errors - who likes those type annotations? When using such option parsing libraries, users usually end up creating a big object with fields corresponding to each option: object Conf { var apples:Int = 0 var bananas:Int = 0 } // ... parsing options ... Conf.apples = parser.getOption("option name", ...) Conf.bananas = parser.getOption("other option name", ...) Sure, this works, and usually flexible enough - but why would anybody want to create such boilerplate? And by the way, nobody likes var's in their code :) So in Scallop , I needed to cut down that unnecessary repetition. There are several ways: Create a compiler plugin. The most powerful option. And the least enjoyable - even given the ease of installing a compiler plugin using sbt, I doubt that anybody will install compiler p...
Read more