1. Introduction

This document describes how to build the LinuxCNC software and documentation from source. This is primarily useful if you are a developer who is modifying LinuxCNC. It can also be useful if you’re a user who is testing developer branches, though then you also have the option of just installing Debian packages from the buildbot: http://buildbot.linuxcnc.org

1.1. Quick Start

For the impatient, try this:

> git clone git://github.com/linuxcnc/linuxcnc.git linuxcnc-dev
> cd linuxcnc-dev/src
> ./autogen.sh
> ./configure --with-realtime=uspace
> make

That will probably fail! That doesn’t make you a bad person, it just means you should read this whole document to find out how to fix your problems. Especially the section on Satisfying Build Dependencies.

If you are running on a realtime-capable system (such as an install from the LinuxCNC Live/Install Image, see the Realtime section below), one extra build step is needed at this time:

> sudo make setuid

After you’ve successfully built LinuxCNC it’s time to run the tests:

> source ../scripts/rip-environment
> runtests

This might fail too! Read this whole document, but especially the section on Setting up the test environment.

2. Supported Platforms

The LinuxCNC project targets modern Debian-based distributions, including Debian, Ubuntu, and Mint.

We continuously test on the platforms listed at http://buildbot.linuxcnc.org.

LinuxCNC builds on most other Linux distributions, though dependency management will be more manual and less automatic. Patches to improve portability to new platforms are always welcome.

2.1. Realtime

LinuxCNC is a machine tool controller, and it requires a realtime platform to do this job. This version of LinuxCNC supports three realtime platforms

RTAI

From https://www.rtai.org. A Linux kernel with the RTAI patch is available from the Debian archive at http://www.linuxcnc.org. See Getting LinuxCNC for installation instructions.

Xenomai

From https://xenomai.org. You will have to compile or obtain a Xenomai kernel yourself.

Preempt-RT

From https://rt.wiki.kernel.org. A Linux kernel with the Preempt-RT patch is occasionally available from the Debian archive at https://www.debian.org, and from the wayback machine at https://snapshot.debian.org.

To make use of the realtime capabilities of LinuxCNC, certain parts of LinuxCNC need to run with root privileges. To enable root for these parts, run this extra command after the make that builds LinuxCNC:

> sudo make setuid

2.2. Non-realtime

LinuxCNC can also be built and run on non-realtime platforms, such as a regular install of Debian or Ubuntu without any special realtime kernel.

In this mode LinuxCNC is not useful for controlling machine tools, but it is useful for simulating the execution of G-code and for testing the non-realtime parts of the system (such as the user interfaces, and some kinds of components and device drivers).

3. Build modes

There are two ways to build LinuxCNC: the developer-friendly "run in place" mode and the user-friendly Debian packaging mode.

3.1. Building for Run In Place

In a Run-In-Place build, the LinuxCNC programs are compiled from source and then run directly from within the build directory. Nothing is installed outside the build directory.

This is quick and easy, and suitable for rapid iteration of changes.

The LinuxCNC test suite runs only in a Run-In-Place build.

Most LinuxCNC developers primarily build using this mode.

Building for Run-In-Place follows the steps in the Quick Start section at the top of this document, possibly with different arguments to src/configure and make.

3.1.1. src/configure arguments

The src/configure script configures how the source code will be compiled. It takes many optional arguments.

List all arguments to src/configure by running this:

> cd linuxcnc-dev/src
> ./configure --help

The most commonly used arguments are:

--with-realtime=uspace

Build for any realtime platform, or for non-realtime. The resulting LinuxCNC executables will run on both a Linux kernel with Preempt-RT patches (providing realtime machine control) and on a vanilla (un-patched) Linux kernel (providing G-code simulation but no realtime machine control). If development files are installed for Xenomai (typically from package libxenomai-dev) or RTAI (typically from a package with a name starting "rtai-modules"), support for these real-time kernels will also be enabled.

--with-realtime=/usr/realtime-$VERSION

Build for the RTAI realtime platform using the older "kernel realtime" model. This requires that you have an RTAI kernel and the RTAI modules installed in /usr/realtime-$VERSION. The resulting LinuxCNC executables will only run on the specified RTAI kernel. As of LinuxCNC 2.7, this produces the best realtime performance.

--enable-build-documentation

Build the documentation, in addition to the executables. This option adds significantly to the time required for compilation, as building the docs is quite time consuming. If you are not actively working on the documentation you may want to omit this argument.

3.1.2. make arguments

The make command takes two useful optional arguments.

Parallel compilation

make takes an optional argument -jN (where N is a number). This enables parallel compilation with N simultaneous processes, which can significantly speed up your build.

A useful value for N is the number of CPUs in your build system. You can discover the number of CPUs by running nproc.

Building just a specific target

If you want to build just a specific part of LinuxCNC, you can name the thing you want to build on the make command line. For example, if you are working on a component named froboz, you can build its executable by running: 

> cd linuxcnc-dev/src
> make ../bin/froboz

3.2. Building Debian Packages

When building Debian packages, the LinuxCNC programs are compiled from source and then stored in a Debian package, complete with dependency information. This takes more time, and the programs can’t be used until the Debian package is installed on a target machine.

This build mode is primarily useful when packaging the software for delivery to end users, and when building the software for a machine that doesn’t have the build environment installed, or that doesn’t have internet access.

Building Debian packages requires the dpkg-buildpackage tool, from the dpkg-dev package:

> sudo apt-get install dpkg-dev

Building Debian packages also requires that all build dependencies are installed, as described in the section Satisfying Build Dependencies.

Once those prerequisites are met, building the Debian packages consists of two steps.

The first step is generating the Debian package scripts and meta-data from the git repo by running this:

> cd linuxcnc-dev/debian
> ./configure uspace
> cd ..
Note

The debian/configure script is different from the src/configure script!

The debian/configure script needs different arguments depending on the platform you’re building on/for, see the debian/configure arguments section.

Once the Debian package scripts and meta-data are configured, build the package by running dpkg-buildpackage (note that it needs to run from the linuxcnc-dev directory, not from linuxcnc-dev/debian):

> dpkg-buildpackage -b -uc

3.2.1. debian/configure arguments

The debian/configure script configures the Debian packaging. It must be run before dpkg-checkbuilddeps and dpkg-buildpackage can be run.

It takes a single argument which specifies the realtime or non-realtime platform to build for. The normal values for this argument are:

uspace

Configure the Debian package for Preempt-RT realtime or for non-realtime (these two are compatible).

noauto
rtai
xenomai

Normally, the lists of RTOSes for uspace realtime to support is detected automatically. However, if you wish, you may specify one or more of these after uspace to enable support for these RTOSes. Or, to disable autodetection, specify noauto. 

If you want just the traditional RTAI "kernel module" realtime, use
`-r` or `$KERNEL_VERSION` instead.
rtai=<package name>

If the development package for rtai lxrt does not start with "rtai-modules", or if the first such package listed by apt-cache search is not the desired one, then explicitly specify the package name.

-r

Configure the Debian package for the currently running RTAI kernel. You must be running an RTAI kernel on your build machine for this to work!

$KERNEL_VERSION

Configure the debian package for the specified RTAI kernel version (for example "3.4.9-rtai-686-pae"). The matching kernel headers debian package must be installed on your build machine (for example "linux-headers-3.4.9-rtai-686-pae"). Note that you can build LinuxCNC in this configuration, but if you are not running the matching RTAI kernel you will not be able to run LinuxCNC, including the test suite.

4. Satisfying Build Dependencies

On Debian-based platforms we provide packaging meta-data that knows what external software packages need to be installed in order to build LinuxCNC. This is called the Build Dependencies of LinuxCNC. You can use this meta-data to easily list the required packages missing from your build system.

Debian systems provide a program called dpkg-checkbuilddeps that parses the package meta-data and compares the packages listed as build dependencies against the list of installed packages, and tells you what’s missing.

First, install the dpkg-checkbuilddeps program by running:

> sudo apt-get install dpkg-dev

Then ask your LinuxCNC git checkout to generate its Debian package meta-data:

> cd linuxcnc-dev/debian
> ./configure uspace
> cd ..

Finally ask dpkg-checkbuilddeps to do its job (note that it needs to run from the linuxcnc-dev directory, not from linuxcnc-dev/debian):

> dpkg-checkbuilddeps

It will emit a list of packages that are required to build LinuxCNC on your system, but that are not installed yet. Install them all with sudo apt-get install, followed by the package names.

You can rerun dpkg-checkbuilddeps any time you want, to list any missing packages.

5. Setting up the environment

This section describes the special steps needed to set up a machine to run the LinuxCNC programs, including the tests.

5.1. Increase the locked memory limit

LinuxCNC tries to improve its realtime latency by locking the memory it uses into RAM. It does this in order to prevent the operating system from swapping LinuxCNC out to disk, which would have bad effects on latency.

Normally locking memory into RAM is frowned upon, and the operating system places a strict limit on how much memory a user is allowed to have locked.

When using the Preempt-RT realtime platform LinuxCNC runs with enough privilege to raise its memory lock limit itself. When using the RTAI realtime platform it does not have enough privilege, and the user must raise the memory lock limit.

If LinuxCNC displays the following message on startup, the problem is your system’s configured limit on locked memory:

RTAPI: ERROR: failed to map shmem
RTAPI: Locked memory limit is 32KiB, recommended at least 20480KiB.

To fix this problem, add a file named /etc/security/limits.d/linuxcnc.conf (as root) with your favorite text editor (e.g., sudo gedit /etc/security/limits.d/linuxcnc.conf). The file should contain the following line:

* - memlock 20480

Log out and log back in to make the changes take effect. Verify that the memory lock limit is raised using the following command:

> ulimit -l

6. Options for checking out the git repo

The Quick Start instructions at the top of this document said to make an anonymous local clone from our git repo at http://github.com/linuxcnc/linuxcnc.git. This is the quickest, easiest way to get started. However, there are other options to consider.

6.1. Fork us on Github

The LinuxCNC project git repo is at http://github.com/LinuxCNC/linuxcnc. github is a popular git hosting service and code sharing website. You can easily (and for no cost) create a fork of our git repo at github, and use that to track and publish your changes.

After creating your own github fork of LinuxCNC, clone it to your development machine and proceed with your hacking as usual.

We of the LinuxCNC project hope that you will share your changes with us, so that the community can benefit from your work. Github makes this sharing very easy: after you polish your changes and push them to your github fork, send us a Pull Request.