\( \newcommand{\E}{\mathrm{E}} \) \( \newcommand{\A}{\mathrm{A}} \) \( \newcommand{\R}{\mathrm{R}} \) \( \newcommand{\N}{\mathrm{N}} \) \( \newcommand{\Q}{\mathrm{Q}} \) \( \newcommand{\Z}{\mathrm{Z}} \) \( \def\ccSum #1#2#3{ \sum_{#1}^{#2}{#3} } \def\ccProd #1#2#3{ \sum_{#1}^{#2}{#3} }\)
CGAL 4.14.2 - Manual
Installation

Authors
Eric Berberich, Joachim Reichel, and Fernando Cacciola

Introduction

This document describes how to install CGAL on Windows, Unix-like systems, and MacOS X.

Ideally, setting up CGAL amounts to:

cd CGAL-4.14.2 # go to CGAL directory
cmake . # configure CGAL
make # build the CGAL libraries

Compiling an example or a demo shipped with CGAL is similarly simple:

cd examples/Triangulation_2 # go to an example directory
cmake -DCGAL_DIR=$HOME/CGAL-4.14.2 . # configure the examples
make # build the examples 
cd demo/Triangulation_2 # go to a demo directory
cmake -DCGAL_DIR=$HOME/CGAL-4.14.2 . # configure the demos
make # build the demos 

Compiling an own non-shipped program is also close:

cd /path/to/program 
cgal_create_CMakeLists -s executable 
cmake -DCGAL_DIR=$HOME/CGAL-4.14.2 . 
make

where the second line creates a CMakeLists.txt file (check its options in Section cgal_create_CMakeLists for various details).

In a less ideal world, you probably have to install CMake, a makefile generator, and third party libraries. That is what this manual is about.

Prerequisites

Installing CGAL requires a few components to be installed ahead: a supported compiler (see Section Supported Compilers, CMake, Boost, and at least Gmp, and Mpfr; see Section Essential Third Party Libraries for more details on essential third party software.

OS Specific Installation

Some operating systems with package managers offer CGAL and its essential third party software through the manager, for instance, Mac OS X, or some Linux distribution (e.g. Debian). For Windows, an installer is provided.

CGAL on MAC OS X

For instance, use macports in the following way:

sudo port install cgal

or if Qt5 are desired

sudo port install cgal +qt5 +universal +demos

The setup is similar for homebrew .

CGAL on Linux

For instance in debian/Ubuntu, use apt-get in the following way:

sudo apt-get install libcgal-dev

To get the demos use

sudo apt-get install libcgal-demo

Check the CGAL-FAQ for source repository of newest releases.

On other distributions, please consult your package manager documentation.

CGAL on Windows

You can download and run CGAL-4.14.2-Setup.exe from https://www.cgal.org/download.html. It is a self extracting executable that installs the CGAL source, and that allows you to select and download some precompiled third party libraries. However, you will need to compile the library using your favorite compiler.

A tutorial is provided on how to proceed with Microsoft Visual Studio.

Downloading CGAL

You can obtain the CGAL library from https://www.cgal.org/download.html and install it yourself.

After you have downloaded the file CGAL-4.14.2.tar.gz containing the CGAL sources, you have to unpack it. Under a Unix-like shell, use the command:

tar xzf CGAL-4.14.2.tar.gz

In both cases the directory CGAL-4.14.2 will be created. This directory contains the following subdirectories:

Directory Contents
auxiliary precompiled Gmp and Mpfr for Windows
cmake/modules modules for finding and using libraries
config configuration files for install script
demo demo programs (most of them need Qt, geomview or other third-party products)
doc_html documentation (HTML)
examples example programs
include header files
scripts some useful scripts (e.g. for creating CMakeLists.txt files)
src source files

The directories include/CGAL/CORE and src/CGALCore contain a distribution of the Core libraryhttps://cs.nyu.edu/exact/ version 1.7 for dealing with algebraic numbers. Core is not part of CGAL and has its own license.

The directory include/CGAL/OpenNL contains a distribution of the Open Numerical Library which provides solvers for sparse linear systems, especially designed for the Computer Graphics community. OpenNL is not part of CGAL and has its own license.

The only documentation shipped with CGAL sources is the present installation manual. The CGAL manual must be downloaded separately from https://www.cgal.org/download.html.

Supported Compilers

In order to build the CGAL libraries, you need a C++ compiler. CGAL 4.14.2 is supported, that is continuously tested, for the following compilers/operating systems:

Compiler Operating System
Gnu g++ 4.1 or laterhttp://gcc.gnu.org/ Linux / MacOS X
MS Windows
MS Visual C++ 12.0, 14.0, 15.9 (Visual Studio 2013, 2015, and 2017)https://visualstudio.microsoft.com/ MS Windows
Clang http://clang.llvm.org/ compiler version 7.0.1 Linux
Apple Clang compiler version 7.0.2 MacOS X

It may work for older versions of the above listed compilers.

Attention
Recent versions of CMake are needed for recent versions of MS Visual C++. Please refer to CMake's documentation for further information.

Configuring CGAL with CMake

In order to configure, build, and install the CGAL libraries, examples and demos, you need CMake, a cross-platform "makefile generator". If CMake is not installed already you can obtain it from https://cmake.org/. CMake version 3.1 or higher is required. This manual explains only those features of CMake which are needed in order to build CGAL. Please refer to the CMake documentation at https://cmake.org/ for further details.

Before building CGAL you have to choose the compiler/linker, set compiler and linker flags, specify which third-party libraries you want to use and where they can be found, and which CGAL libraries you want to build. Gathering all this information is called configuration. The end of the process is marked by the generation of a makefile or a Visual C++ solution and project file that you can use to build CGAL.

Configuring CGAL with the CMake GUI

The simplest way to start the configuration is to run the graphical user interface of CMake. We recommend to use cmake-gui. You must pass as argument the root directory of CGAL. For example:

cd CGAL-4.14.2
cmake-gui . # Notice the dot to indicate the current directory.

After cmake-gui opens, press 'Configure'. A dialog will pop up and you will have to choose what shall gets generated. After you have made your choice and pressed 'Finish', you will see the output of configuration tests in the lower portion of the application. When these tests are done, you will see many red entries in the upper portion of the application. Just ignore them and press 'Configure'. By now CMake should have found many libraries and have initialized variables. If you still find red entries, you have to provide the necessary information. This typically happens if you have installed software at non-standard locations. Providing information and pressing 'Configure' goes on until all entries are grayed. You are now ready to press 'Generate'. Once this is done, you can quit cmake-gui.

Configuring CGAL with the cmake Command-Line Tool

Alternatively, you can run the command-line tool called cmake. You pass as argument the root directory of CGAL. For example:

cd CGAL-4.14.2
cmake . # Notice the dot to indicate the current directory.

The very first thing CMake does is to detect the compiler to use. This detection is performed by a special CMake module called a generator. A CMake generator understands the build requirements for a particular compiler/linker and generates the necessary files for that. For example, the UNIX Makefiles generator understands the GNU chain of tools (GCC, ld etc.) and produces makefiles, which can be used to build a target by a simple call to make. Likewise, the Visual Studio 2010 generator produces solution and project files and can be manually launched in the VS IDE to build the target.

Each platform has a default generator, so you only need to select one when the default is not what you want. For example, under Windows, it is possible to generate NMakefiles instead of Visual Studio project files in order to build the library with nmake. Running cmake with no parameters in a command-line prints the list of available generators supported by your platform and CMake version. If the generator you need is not listed there, you can try a newer CMake version, as generators are hardcoded into CMake, and additional generators are added with each release.

Since the choice of the generator determines the type of build files to generate, in some cases you choose a particular generator as a mean to choose a specific compiler (because they use different build files). For example, the following generates solution files for use in Visual C++ 15.0 on a 64bit machine:

cd CGAL-4.14.2
cmake -G"Visual Studio 15 2017 Win64" . 

In other cases, however, the generator doesn't directly identify a specific compiler but a chain of tools. For example, the UNIX Makefiles generator produces makefiles that call some auto-detected command-line compiler, like GCC. If you need the makefiles to use a different compiler, you need to specify the desired compiler in the call to CMake, as in this example:

cd CGAL-4.14.2
cmake -DCMAKE_CXX_COMPILER:FILEPATH=g++-4.7 . 

CMake maintains configuration parameters in so-called cmake variables, like the CMAKE_CXX_COMPILER in the example above. These variables are not environment variables but CMake variables. Some of the CMake variables represent user choices, such as WITH_examples or CMAKE_BUILD_TYPE=Release, while others indicate the details of a third-party library, such as Boost_INCLUDE_DIR or the compiler flags to use, such as CMAKE_CXX_FLAGS.

The command line tool cmake accepts CMake variables as arguments of the form -D<VAR>:<TYPE>=<VALUE>, as in the example above, but this is only useful if you already know which variables need to be explicitly defined.

Advanced

CMake keeps the variables that a user can manipulate in a so-called CMake cache, a simple text file named CMakeCache.txt, whose entries are of the form VARIABLE:TYPE=VALUE. Advanced users can manually edit this file, instead of going through the interactive configuration session.

The configuration process not only determines the location of the required dependencies, it also dynamically generates a compiler_config.h file, which encodes the properties of your system and a special file named CGALConfig.cmake, which is used to build programs using CGAL. The purpose of this file is explained below.

CGAL Libraries

CGAL is split into five libraries. During configuration, you can select the libraries that you would like to build by setting a CMake variable of the form WITH_<library>. By default all are switched ON. All activated libraries are build after configuration; see Building CGAL

We next list the libraries and essential 3rd party software (see Essential Third Party Libraries) for each library:

Library CMake Variable Functionality Dependencies
CGAL none Main library Gmp, Mpfr, Boost (headers), Boost.Thread and Boost.System (library) for compilers not supporting the keywords thread_local and the class std::mutex
CGAL_Core WITH_CGAL_Core The CORE library for algebraic numbers.CGAL_Core is not part of CGAL, but a custom version of the Core library distributed by CGAL for the user convenience and it has it's own license. Gmp and Mpfr
CGAL_ImageIO WITH_CGAL_ImageIO Utilities to read and write image files OpenGL, zlib, Vtk(optional)
CGAL_Qt5 WITH_CGAL_Qt5 QGraphicsView support for Qt5-based demos Qt5 and OpenGL

Debug vs. Release

The CMake variable CMAKE_BUILD_TYPE indicates how to build the libraries. It accepts the values Release or Debug. The default is Release and should be kept, unless you want to debug your program.

This is not an issue for solution/project files, as there the user selects the build type from within the IDE.

Static vs. Shared Libraries

Shared libraries, also called dynamic-link libraries, are built by default (.dll on Windows, .so on Linux, .dylib on MacOS). You can choose to produce static libraries instead by setting the CMake variable BUILD_SHARED_LIBS to FALSE. If you use cmake-gui, a tick box for that variable is available to set it.

Header-only Option

Header-only with CMake Configuration

Since CGAL 4.9, CGAL can be used in header-only mode, i.e. without compiling the CGAL libraries and linking with these libraries when compiling examples, tests and demos. This possibility can be enabled by setting the value of the CMake variable CGAL_HEADER_ONLY to ON. CMake version 3.0.0 or higher is required to use this option.

One advantage of using CGAL in header-only mode is that you do not need to compile and install CGAL libraries before compiling a given example or demo. Note that even in header-only mode we still need to run CMake on CGAL in order to generate different configuration files. So, setting up CGAL becomes now:

cd CGAL-4.14.2 # go to CGAL directory
cmake -DCGAL_HEADER_ONLY=ON . # configure CGAL

and we do not need to run make anymore.

Header-only without CMake Configuration

Since CGAL 4.12, CGAL can be used in header-only mode, without even configuring CGAL. Programs using CGAL (examples, tests, demos, etc.) must be directly configured using CMake. In this case, CGAL will be configured at the same time. The variable CGAL_DIR must point to the root directory of the CGAL source code (either the root of the unpacked release tarball, or the root of the Git working directory).

So, using CGAL becomes now:

cd /path/to/your/code # go to the directory of the code source using CGAL
cmake -DCGAL_DIR=<CGAL-root> .

CGAL Dependencies

CGAL can be used as a header-only library, though not all its dependencies are header-only. The libraries Gmp and Mpfr, for example, are not header-only.

Possible Problem on Windows

There is one possible problem when using CGAL in header-only mode on a Windows operating system when compiling a program using several modules (executable programs or dynamic-link libraries DLL). If two different modules use the same static variable, this variable is defined independently in each of these modules. If one module modifies the value of this variable, it will not be modified in the other module, which could induce an unexpected behavior. In CGAL, this concerns only a few specific variables: the default random, the failure behavior, IO mode. One example is the following: if you change the default random in one DLL, then if you use the default random in another DLL, you will not obtain the modified default random but the original one.

Essential Third Party Libraries

The focus of CGAL is on geometry, and we rely on other highly specialized libraries and software for non-geometric issues, for instance, for numeric solvers, or visualization. We first list software that is essential to build (all) libraries of CGAL, that is, this software must be found during the configuration of CGAL for an actived library of CGAL (i.e. WITH_<library>=ON); see sec3partysoftwareconfig to specify the location of 3rd party software.

The libraries STL (shipped with any compiler) and Boost are essential to all components (i.e. libCGAL, libCGAL_Core, libCGAL_ImageIO, and libCGAL_Qt5).

Standard Template Library (STL)

CGAL heavily uses the STL, and in particular adopted many of its design ideas. You can find online documentation for the STL at various web sites, for instance, http://www.cplusplus.com/reference/, or https://msdn.microsoft.com/en-us/library/1fe2x6kt(v=vs.140).aspx.

The STL comes with the compiler, so there is nothing to install.

Boost

The Boost libraries are a set of portable C++ source libraries. Most of Boost libraries are header-only, but a few of them need to be compiled or installed as binaries.

CGAL requires the Boost libraries. In particular the header files and the threading library (Boost.Thread and Boost.System binaries). Version 1.48 (or higher) are needed for compilers not supporting the keyword thread_local and the class std::mutex (This is supported for GCC 4.8 and later when using the C++ 11 standard, and for Visual C++ starting with 2015, that is VC14).

As an exception, because of a bug in the GCC compiler about the C++ 11 keyword thread_local, the CGAL_Core library always requires Boost.Thread if the GCC compiler is used.

On Windows, as auto-linking is used, you also need the binaries of Boost.Serialization and Boost.DateTime, but the dependency is artificial and used only at link-time: the CGAL libraries do not depend on the DLL's of those two libraries.

In CGAL some demos and examples depend on Boost.Program_options.

In case the Boost libraries are not installed on your system already, you can obtain them from https://www.boost.org/. For Visual C++ you can download precompiled libraries from https://sourceforge.net/projects/boost/files/boost-binaries/. For Visual C++ versions prior to 2015 Boost.Thread is required, so make sure to either install the precompiled libraries for your compiler or build libboost-thread and libboost-system.

As on Windows there is no canonical directory for where to find Boost, we recommend that you define the environment variable BOOST_ROOT and set it to where you have installed Boost, e.g., C:\boost\boost_1_41_0.

GMP and MPFR

The components libCGAL, libCGAL_Core, and libCGAL_Qt5 require Gmp and Mpfr which are libraries for multi precision integers and rational numbers, and for multi precision floating point numbers.

CGAL combines floating point arithmetic with exact arithmetic, in order to be efficient and reliable. CGAL has a built-in number type for that, but Gmp and Mpfr provide a faster solution, and we recommend to use them.

Having Gmp version 4.2 or higher and Mpfr version 2.2.1 or higher installed is recommended. These libraries can be obtained from https://gmplib.org/ and https://www.mpfr.org/, respectively.

As Visual C++ is not properly supported by the Gmp and Mpfr projects, we provide precompiled versions of Gmp and Mpfr, which can be downloaded with the installer CGAL-4.14.2-Setup.exe.

zlib

zlib is a data compression library, and is essential for the component libCGAL_ImageIO.

In CGAL this library is used in the examples of the 3D Surface Mesh Generation Reference package.

If it is not already on your system, for instance, on Windows, you can download it from http://www.zlib.net/.

OpenGL

OpenGL (Open Graphics Library) provides an API for applications that produce 2D and 3D computer graphics.

In CGAL the library is essential for the component libCGAL_Qt5, as well as libCGAL_ImageIO and for various demos.

Typically, OpenGL is preinstalled on systems; if not, it can be downloaded from http://www.opengl.org/.

Qt5

Qt is a cross-platform application and UI framework.

The component libCGAL_Qt5 requires Qt5 installed on your system. In case Qt is not yet installed on your system, you can download it from http://www.qt-project.org/.

The exhaustive list of Qt5 components used in demos is: Core, Gui, Help, OpenGL, Script, ScriptTools, Svg, Widgets, qcollectiongenerator (with sqlite driver plugin) and Xml.

Qt version 5.9.0 or later is required.

CGAL Examples and Demos

CGAL is distributed with a large collection of examples and demos. By default, these are not configured along with the CGAL libraries, unless you set the variables WITH_examples=ON and/or WITH_demos=ON.

Nevertheless, even when configured with CGAL, they are not automatically built along with the libraries. You must build the examples or demos targets (or IDE projects) explicitly.

If you do not plan to compile any demos, you might skip some of the essential libraries (as Qt or OpenGL), as the corresponding CGAL-libraries are not linked. But for your own demos you might need these CGAL-libraries.

Optional Third Party Libraries

Optional 3rd party software can be used by CGAL for various reasons: Usually certain optional libraries are required to build examples and demos shipped with CGAL or to build your own project using CGAL. Another reason is to speed up basic tasks. In order to support these goals, all optional libraries can be prepared to be used with CGAL while configuring CGAL, just in the same way as essential libraries are configured. Whenever building an example or a demo (or your own executable), these preconfigured libraries are available when using CGAL.

LEDA

LEDA is a library of efficient data structures and algorithms. Like Core, LEDA offers a real number data type.

In CGAL this library is optional, and its number types can be used as an alternative to Gmp, Mpfr, and Core.

Free and commercial editions of LEDA are available from https://www.algorithmic-solutions.com/leda/index.html.

MPFI

Mpfi provides arbitrary precision interval arithmetic with intervals represented using Mpfr reliable floating-point numbers. It is based on the libraries Gmp and Mpfr. In the setting of CGAL, this library is optional: it is used by some models of the Algebraic Kernel.

Mpfi can be downloaded from http://mpfi.gforge.inria.fr/. Version 1.4 or higher is recommended.

RS and RS3

Rs (Real Solutions) is devoted to the study of the real roots of polynomial systems with a finite number of complex roots (including univariate polynomials). In CGAL, Rs is used by one model of the Algebraic Kernel.

Rs is freely distributable for non-commercial use. You can download it from http://vegas.loria.fr/rs/. Actually, the Rs package also includes Rs3, the successor of Rs, which is used in conjunction with it.

The libraries Rs and Rs3 need Mpfi, which can be downloaded from http://mpfi.gforge.inria.fr/.

NTL

Ntl provides data structures and algorithms for signed, arbitrary length integers, and for vectors, matrices, and polynomials over the integers and over finite fields. The optional library Ntl is used by CGAL to speed up operations of the Polynomial package, such as GCDs. It is recommended to install Ntl with support from Gmp.

Ntl can be downloaded from http://www.shoup.net/ntl/. Version 5.1 or higher is recommended.

Eigen

Eigen is a C++ template library for linear algebra. Eigen supports all matrix sizes, various matrix decomposition methods and sparse linear solvers.

In CGAL, Eigen provides sparse linear solvers in the Poisson Surface Reconstruction Reference and the Triangulated Surface Mesh Parameterization Reference packages.

In addition, Eigen also provides singular value decomposition for the Estimation of Local Differential Properties of Point-Sampled Surfaces Reference and the Approximation of Ridges and Umbilics on Triangulated Surface Meshes Reference packages.

The Eigen web site is http://eigen.tuxfamily.org.

ESBTL

The Esbtl (Easy Structural Biology Template Library) is a library that allows the handling of Pdb data.

In CGAL the Esbtl is used in an example of the 3D Skin Surface Meshing Reference package.

It can be downloaded from http://esbtl.sourceforge.net/.

Intel TBB

Tbb (Threading Building Blocks) is a library developed by Intel Corporation for writing software programs that take advantage of multi-core processors.

In CGAL, Tbb is used by the packages that offer parallel code.

The Tbb web site is https://www.threadingbuildingblocks.org.

LASlib

LASlib is a C++ library for handling LIDAR data sets stored in the LAS format (or the compressed LAZ format).

In CGAL, LASlib is used to provide input and output functions in the Point Set Processing Reference package.

The LASlib web site is https://rapidlasso.com/lastools/. LASlib is usually distributed along with LAStools: for simplicity, CGAL provides a fork with a CMake based install procedure.

OpenCV

OpenCV (Open Computer Vision) is a library designed for computer vision, computer graphics and machine learning.

In CGAL, OpenCV is used by the Classification Reference package.

The OpenCV web site is https://opencv.org/.

TensorFlow

TensorFlow is a library designed for machine learning and deep learning.

In CGAL, the C++ API of TensorFlow is used by the Classification Reference package for neural network. The C++ API can be compiled using CMake: it is distributed as part of the official package and is located in tensorflow/contrib/cmake. Be sure to enable and compile the following targets:

  • tensorflow_BUILD_ALL_KERNELS
  • tensorflow_BUILD_PYTHON_BINDINGS
  • tensorflow_BUILD_SHARED_LIB.

The TensorFlow web site is https://www.tensorflow.org/.

METIS

METIS is a library developed by the Karypis Lab and designed to partition graphs and produce fill-reducing matrix orderings.

CGAL offers wrappers around some of the methods of the METIS library to allow the partitioning of graphs that are models of the concepts of the Boost Graph Library, and, by extension, of surface meshes (see Section Graph Partitioning of the package CGAL and the Boost Graph Library).

More information is available on the METIS library at http://glaros.dtc.umn.edu/gkhome/metis/metis/overview.

Building CGAL

The results of a successful configuration are build files that control the build step. The nature of the build files depends on the generator used during configuration, but in all cases they contain several targets, one per library, and a default global target corresponding to all the libraries.

For example, in a Unix-like environment the default generator produces makefiles. You can use the make command-line tool for the succeeding build step as follows:

cd CGAL-4.14.2
# build all the selected libraries at once
make

The resulting libraries are placed in the subdirectory lib under <CMAKE_BINARY_DIR> (which is CGAL-4.14.2 in case you run an in-source-configuration).

With generators other than UNIX Makefiles the resulting build files are solution and project files which should be launched in an Ide, such as Visual Studio or KDevelop3. They will contain the targets described above, which you can manually build as with any other solution/project within your Ide.

Alternatively, you can build it with the command line version of the Visual Studio Ide:

devenv CGAL.sln /Build Debug

The "Debug" argument is needed because CMake creates solution files for all four configurations, and you need to explicitly choose one when building (the other choices are Release, RelWithDebInfo and MinSizeRel).

Advanced

The build files produced by CMake are autoconfigured. That is, if you change any of the dependencies, the build step automatically goes all the way back to the configuration step. This way, once the target has been configured the very first time by invoking cmake, you don't necessarily need to invoke cmake again. Rebuilding will call itself cmake and re-generate the build file whenever needed. Keep this in mind if you configure CGAL for the Visual Studio IDE since a build could then change the solution/project file in-place and VS will prompt you to reload it.

If you have turned on the configuration of examples (-DWITH_examples=ON) and/or demos (-DWITH_demos=ON), there will be additional targets named examples and demos, plus one target for each example and each demo in the build files. None of these targets are included by default, so you need to build them explicitly after the CGAL libraries have been successfully built. The targets examples and demos include themselves all the targets for examples and demos respectively.

# build all examples at once
make examples
# build all demos at once
make demos

Advanced

When using UNIX Makefiles you can find out the exact name of the example or demo target of a particular package by typing make help | grep <package>.

Installing CGAL

On many platforms, library pieces such as headers, docs and binaries are expected to be placed in specific locations. A typical example being /usr/include and /usr/lib on Unix-like operating systems or C:/Program Files/ on Windows. The process of placing or copying the library elements into its standard location is sometimes referred to as Installation and it is a postprocessing step after the build step.

CMake carries out the installation by producing a build target named install. The following example shows a typical session from configuration to installation in a Unix-like environment:

cd CGAL-4.14.2
cmake . # configure
make # compile
make install # install

If you use a generator that produces IDE files (for Visual Studio for instance) there will be an optional INSTALL project, which you will be able to "build" to execute the installation step.

Advanced

The files are copied into a directory tree relative to the installation directory determined by the CMake variable CMAKE_INSTALL_PREFIX. This variable defaults to /usr/local under Unix-like operating systems and C:\Program Files under Windows. If you want to install to a different location, you must override that CMake variable explicitly at the configuration time and not when executing the install step.

The file CGALConfig.cmake is installed by default in $CMAKE_INSTALLED_PREFIX/lib/CGAL-4.14.2.

Multiple Variants of Makefiles (out-of-source build)

While you can choose between release or debug builds, and shared or static libraries, it is not possible to generate different variants during a single configuration. You need to run CMake in a different directory for each variant you are interested in, each with its own selection of configuration parameters.

CMake stores the resulting makefiles and project files, along with several temporary and auxiliary files such as the variables cache, in the directory where it is executed, called CMAKE_BINARY_DIR, but it takes the source files and configuration scripts from CMAKE_SOURCE_DIR.

The binary and source directories do not need to be the same. Thus, you can configure multiple variants by creating a distinct directory for each configuration and by running CMake from there. This is known in CMake terminology as out-of-source configuration, as opposite to an in-source configuration, as showed in the previous sections.

You can, for example, generate subdirectories CGAL-4.14.2/cmake/platforms/debug and CGAL-4.14.2/cmake/platforms/release for two configurations, respectively:

mkdir CGAL-4.14.2/cmake/platforms/debug
cd CGAL-4.14.2/cmake/platforms/debug
cmake -DCMAKE_BUILD_TYPE=Debug ../../..
mkdir CGAL-4.14.2/cmake/platforms/release
cd CGAL-4.14.2/cmake/platforms/release
cmake -DCMAKE_BUILD_TYPE=Release ../../..

Configuring and Building Programs Using CGAL

Ideally, configuring and compiling a demo/example/program amounts to

cd CGAL-4.14.2/examples/Triangulation_2
cmake -DCGAL_DIR=$HOME/CGAL-4.14.2 .
make

In this ideal world, as for all shipped examples and demos of CGAL, the required CMakeLists.txt is already provided.

CMake can also be used to configure and build user programs via such CMake-scripts. In this less ideal world, one has to provide the CMakeLists.txt script either manually, or with the help of a shell-script that is introduced below.

For a user program executable.cpp, the ideal world looks like this:

cd /path/to/program 
cgal_create_CMakeLists -s executable
cmake -DCGAL_DIR=$HOME/CGAL-4.14.2 . 
make

In both examples we specify the CGAL_DIR: During configuration of the CGAL libraries a file named CGALConfig.cmake is generated in CGAL's root directory (in contrast to CGAL's source directory that has been used for installation). This file contains the definitions of several CMake variable that summarize the configuration of CGAL. In order to configure a program, you need to indicate the location of that config file in the CMake variable CGAL_DIR (as indicated in the example above). CGAL_DIR can also be an environment variable. Setting CGAL_DIR makes particular sense if having multiple out-of-source builds of CGAL as in Section Multiple Variants of Makefiles (out-of-source build).

If you have installed CGAL, CGAL_DIR must afterwards be set to $CMAKE_INSTALLED_PREFIX/lib/CGAL. Note that CGAL is recommended to be installed in release mode when using it to build programs.

Creating a CMake Script for a Program Using CGAL

For compiling a non-shipped program, it is recommended, to also rely on a CMake-supported configuration using a CMakeLists.txt used for configuration.

Use the following Bourne-shell script for programs that are relatively simple to configure:

cgal_create_CMakeLists

The Bourne-shell script cgal_create_CMakeLists.txt resides in the CGAL-4.14.2/scripts directory. It can be used to create CMakeLists.txt files for compiling CGAL applications. Executing cgal_create_CMakeLists.txt in an application directory creates a CMakeLists.txt containing rules to build the contained application(s). Three command line options determine details of the configuration.

-s source

If this parameter is given the script will create a single executable for 'source' linked with compilations of all other source files (*.cc, *.cp, *.cxx, *.cpp, *.CPP, *.c++, or *.C). This behaviour is usually needed for (graphical) demos.

If the parameter is not given, the script creates one executable for each given source file.

-c com1:com2:...

Lists components ("com1", "com2") of CGAL to which the executable(s) should be linked. Valid components are CGAL's libraries (i.e. "Core", "ImageIO", and "Qt5") and all preconfigured 3rd party software, such as "MPFI", "RS3"). An example is -c Core:GMP:RS3:MPFI

-b boost1:boost2:...

Lists components ("boost1", "boost2") of Boost to which the executable(s) should be linked. Valid options are, for instance, "filesystem" or "program_options".

This options should suffice to create CMakeLists.txt script for most directories containing programs. However, in some special cases, it might still be required to create the script manually, for instance, if some source files/executables need a different linking than other source files.

cgal_create_cmake_script

Such a shell-script simply creates a CMake script. Processing it with CMake, searches for CGAL using find_package. If found, the variable CGAL_USE_FILE is set to a compilation environment CMake file. Including this file within a CMake script sets up include paths and libraries to link with CGAL and essential third party libraries. Beyond, find_package can demand for COMPONENTS of CGAL, that is, all CGAL libraries libCGAL_Core (Core), libCGAL_ImageIO (ImageIO), and libCGAL_Qt5 (Qt5) or optional 3rd party software such as MPFI, RS3. A user is free to create the CMakeLists.txt without calling the script (manual creation).

Custom Flags in the Programs Using CGAL

Normally, programs linked with CGAL must be compiled with the same flags used by the compilation of CGAL libraries. For this reason, the very first time a program is configured, all the flags given by the CMake variables CMAKE_*_FLAGS are locked in the sense that the values recorded in CGALConfig.cmake are used to override any values given by CMake itself or yourself.

This does not apply to the additional flags that can be given via CGAL_*_FLAGS.

Such inherited values are then recorded in the current CMake cache for the program. The flags are then unlocked in the sense that at any subsequent configuration you can provide your own flags and this time they will not be overridden.

When using the interactive cmake-gui the first press on Configure unlocks the flags, so that you can edit them as needed.

Advanced

The locking of flags is controlled by the variable CGAL_DONT_OVERRIDE_CMAKE_FLAGS which starts out FALSE and is toggled right after the flags have been loaded from CGALConfig.cmake.

If you use the command line tool you can specify flags directly by setting the controlling variable right up front:

cd CGAL-4.14.2
cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS=-g .
cd CGAL-4.14.2/examples/Triangulation_2
cmake -DCGAL_DIR=CGAL-4.14.2 -DCMAKE_BUILD_TYPE=Debug -DCMAKE_CXX_FLAGS=-O2 -DCGAL_DONT_OVERRIDE_CMAKE_FLAGS=TRUE . 

Summary of CGAL's Configuration Variables

Most configuration variables are not environment variables but CMake variables. They are given in the command line to CMake via the -D option, or passed from the interactive interface of cmake-gui. Unless indicated differently, all the variables summarized below are CMake variables.

Component Selection

The following boolean variables indicate which CGAL components to configure and build. Their values can be ON or OFF.

Variable Default Value
WITH_examples OFF
WITH_demos OFF
WITH_CGAL_Core ON
WITH_CGAL_Qt5 ON
WITH_CGAL_ImageIO ON

Compiler and Linker Flags

The following variables specify compiler and linker flags. Each variable holds a space-separated list of command-line switches for the compiler and linker and their default values are automatically defined by CMake based on the target platform.

Have in mind that these variables specify a list of flags, not just one single flag. If you provide your own definition for a variable, you will entirely override the list of flags chosen by CMake for that particular variable.

The variables that correspond to both debug and release builds are always used in conjunction with those for the specific build type.

Program Both Debug and Release Release Only Debug Only
C++ Compiler CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_RELEASE CMAKE_CXX_FLAGS_DEBUG
Linker (shared libs) CMAKE_SHARED_LINKER_FLAGS CMAKE_SHARED_LINKER_FLAGS_RELEASE CMAKE_SHARED_LINKER_FLAGS_DEBUG
Linker (static libs) CMAKE_MODULE_LINKER_FLAGS CMAKE_MODULE_LINKER_FLAGS_RELEASE CMAKE_MODULE_LINKER_FLAGS_DEBUG
Linker (programs) CMAKE_EXE_LINKER_FLAGS CMAKE_EXE_LINKER_FLAGS_RELEASE CMAKE_EXE_LINKER_FLAGS_DEBUG

Additional Compiler and Linker Flags

The following variables can be used to add flags without overriding the ones defined by cmake.

Program Both Debug and Release Release Only Debug Only
C++ Compiler CGAL_CXX_FLAGS CGAL_CXX_FLAGS_RELEASE CGAL_CXX_FLAGS_DEBUG
Linker (shared libs) CGAL_SHARED_LINKER_FLAGS CGAL_SHARED_LINKER_FLAGS_RELEASE CGAL_SHARED_LINKER_FLAGS_DEBUG
Linker (static libs) CGAL_MODULE_LINKER_FLAGS CGAL_MODULE_LINKER_FLAGS_RELEASE CGAL_MODULE_LINKER_FLAGS_DEBUG
Linker (programs) CGAL_EXE_LINKER_FLAGS CGAL_EXE_LINKER_FLAGS_RELEASE CGAL_EXE_LINKER_FLAGS_DEBUG

Miscellaneous Variables

Variable Description Type Default Value
CMAKE_BUILD_TYPE Indicates type of build. Possible values are 'Debug' or 'Release' CMake Release
CMAKE_CXX_COMPILER Full-path to the executable corresponding to the C++ compiler to use. CMake platform-dependent
CXX Idem Environment Idem

Variables Used Only When Building Programs (Such as Demos or Examples)

Variable Description Type Default Value
CGAL_DIR Full-path to the binary directory where CGAL was configured Either CMake or Environment none

Variables Providing Information About 3rd-Party Libraries

The following variables provide information about the availability and location of the 3rd party libraries used by CGAL. CMake automatically searches for dependencies so you need to specify these variables if CMake was unable to locate something. This is indicated by a value ending in NOTFOUND.

Since 3rd-party libraries are system wide, many of the CMake variables listed below can alternatively be given as similarly-named environment variables instead. Keep in mind that you must provide one or the other but never both.

Boost Libraries

In most cases, if Boost is not automatically found, setting the BOOST_ROOT variable is enough. If it is not, you can specify the header and library directories individually. You can also provide the full pathname to a specific compiled library if it cannot be found in the library directory or its name is non-standard.

By default, when Boost binary libraries are needed, the shared versions are used if present. You can set the variable CGAL_Boost_USE_STATIC_LIBS to ON if you want to link with static versions explicitly.

On Windows, if you link with Boost shared libraries, you must ensure that the .dll files are found by the dynamic linker, at run time. For example, you can add the path to the Boost .dll to the PATH environment variable.

Variable Description Type
BOOST_ROOTThe environment variable can be spelled either BOOST_ROOT or BOOSTROOT Root directory of your Boost installation Either CMake or Environment
Boost_INCLUDE_DIR Directory containing the boost/version.hpp file CMake
BOOST_INCLUDEDIR Idem Environment
Boost_LIBRARY_DIRS Directory containing the compiled Boost libraries CMake
BOOST_LIBRARYDIR Idem Environment
Boost_(xyz)_LIBRARY_RELEASE Full pathname to a release build of the compiled 'xyz' Boost library CMake
Boost_(xyz)_LIBRARY_DEBUG Full pathname to a debug build of the compiled 'xyz' Boost library CMake

GMP and MPFR Libraries

Under Windows, auto-linking is used, so only the directory containing the libraries is needed and you would specify GMP|MPFR_LIBRARY_DIR rather than GMP|MPFR_LIBRARIES. On the other hand, under Linux the actual library filename is needed. Thus you would specify GMP|MPFR_LIBRARIES. In no case you need to specify both.

CGAL uses both Gmp and Mpfr so both need to be supported. If either of them is unavailable the usage of Gmp and of Mpfr will be disabled.

Variable Description Type
CGAL_DISABLE_GMP Indicates whether to search and use Gmp/Mpfr or not CMake
GMP_DIR Directory of Gmp default installation Environment
GMP_INCLUDE_DIR Directory containing the gmp.h file CMake
GMP_INC_DIR Idem Environment
GMP_LIBRARIES_DIR Directory containing the compiled Gmp library CMake
GMP_LIB_DIR Idem Environment
GMP_LIBRARIES Full pathname of the compiled Gmp library CMake
MPFR_INCLUDE_DIR Directory containing the mpfr.h file CMake
MPFR_INC_DIR Idem Environment
MPFR_LIBRARIES_DIR Directory containing the compiled Mpfr library CMake
MPFR_LIB_DIR Idem Environment
MPFR_LIBRARIES Full pathname of the compiled Mpfr library CMake

Under Linux, the Gmpxx is also searched for, and you may specify the following variables:

Variable Description Type
GMPXX_DIR Directory of gmpxx default installation Environment
GMPXX_INCLUDE_DIR Directory containing the gmpxx.h file CMake
GMPXX_LIBRARIES Full pathname of the compiled Gmpxx library CMake

Qt5 Library

You must set the cmake or environment variable Qt5_DIR to point to the path to the directory containing the file Qt5Config.cmake created by your Qt5 installation. If you are using the open source edition it should be <path>/qt-everywhere-opensource-src-<version>/qtbase/lib/cmake/Qt5.

LEDA Library

When the LEDA libraries are not automatically found, yet they are installed on the system with base names 'leda' and 'ledaD' (for the release and debug versions resp.), it might be sufficient to just indicate the library directory via the LEDA_LIBRARY_DIRS variable. If that doesn't work because, for example, the names are different, you can provide the full pathnames of each variant via LEDA_LIBRARY_RELEASE and LEDA_LIBRARY_DEBUG.

The variables specifying definitions and flags can be left undefined if they are not needed by LEDA.

Variable Description Type
WITH_LEDA Indicates whether to search and use LEDA or not CMake
LEDA_DIR Directory of LEDA default installation Environment
LEDA_INCLUDE_DIR Directory containing the file LEDA/system/basic.h CMake
LEDA_LIBRARIES Directory containing the compiled LEDA libraries CMake
LEDA_INC_DIR Directory containing the file LEDA/system/basic.h Environment
LEDA_LIB_DIR Directory containing the compiled LEDA libraries Environment
LEDA_LIBRARY_RELEASE Full pathname to a release build of the LEDA library CMake
LEDA_LIBRARY_DEBUG Full pathname to a debug build of the LEDA library CMake
LEDA_DEFINITIONS Preprocessor definitions CMake
LEDA_CXX_FLAGS Compiler flags CMake
LEDA_LINKER_FLAGS Linker flags CMake

MPFI Library

CGAL provides a number type based on this library, but the CGAL library itself does not depend on Mpfi. This means that this library must be configured when compiling an application that uses the above number type.

When Mpfi files are not on the standard path, the locations of the headers and library files must be specified by using environment variables.

Variable Description Type
MPFI_DIR Directory of MPFI default installation Environment
MPFI_INCLUDE_DIR Directory containing the mpfi.h file CMake
MPFI_INC_DIR Idem Environment
MPFI_LIBRARIES_DIR Directory containing the compiled Mpfi library CMake
MPFI_LIB_DIR Idem Environment
MPFI_LIBRARIES Full pathname of the compiled Mpfi library CMake

RS and RS3 Library

As said before, only the CGAL univariate algebraic kernel depends on the library Rs. As the algebraic kernel is not compiled as a part of the CGAL library, this library is not detected nor configured at installation time.

CMake will try to find Rs in the standard header and library directories. When it is not automatically detected, the locations of the headers and library files must be specified using environment variables.

Rs needs Gmp 4.2 or later and Mpfi 1.3.4 or later. The variables related to the latter library may also need to be defined.

Variable Description Type
RS_DIR Directory of Rs default installation Environment
RS_INCLUDE_DIR Directory containing the rs_exports.h file CMake
RS_INC_DIR Idem Environment
RS_LIBRARIES_DIR Directory containing the compiled Rs library CMake
RS_LIB_DIR Idem Environment
RS_LIBRARIES Full pathname of the compiled Rs library CMake

Similar variables exist for Rs3.

Variable Description Type
RS3_DIR Directory of Rs3 default installation Environment
RS3_INCLUDE_DIR Directory containing the file rs3_fncts.h file CMake
RS3_INC_DIR Idem Environment
RS3_LIBRARIES_DIR Directory containing the compiled Rs3 library CMake
RS3_LIB_DIR Idem Environment
RS3_LIBRARIES Full pathname of the compiled Rs3 library CMake

NTL Library

Some polynomial computations in CGAL's algebraic kernel are speed up when Ntl is available. As the algebraic kernel is not compiled as a part of the CGAL library, this library is not detected nor configured at installation time.

CMake will try to find Ntl in the standard header and library directories. When it is not automatically detected, the locations of the headers and library files must be specified using environment variables.

Variable Description Type
NTL_DIR Directory of NTL default installation Environment
NTL_INCLUDE_DIR Directory containing the NTL/ZZX.h file CMake
NTL_INC_DIR Idem Environment
NTL_LIBRARIES_DIR Directory containing the compiled Ntl library CMake
NTL_LIB_DIR Idem Environment
NTL_LIBRARIES Full pathname of the compiled Ntl library CMake

Eigen Library

Eigen is a header-only template library. Only the directory containing the header files of Eigen 3.1 (or greater) is needed.

Variable Description Type
EIGEN3_INCLUDE_DIR Directory containing the file signature_of_eigen3_matrix_library CMake
EIGEN3_INC_DIR Idem Environment

ESBTL Library

One skin surface example requires the Esbtl library in order to read Pdb files.

If Esbtl is not automatically found, setting the ESBTL_INC_DIR environment variable is sufficient.

Variable Description Type
ESBTL_DIR Directory of ESBTL default installation Environment
ESBTL_INC_DIR Directory containing the ESBTL/default.h file Environment
ESBTL_INCLUDE_DIR Directory containing the ESBTL/default.h file CMake

TBB Library

If Tbb is not automatically found, the user must set the TBBROOT environment variable. The environment variable TBB_ARCH_PLATFORM=<arch>/<compiler> must be set. <arch> is ia32 or intel64. <compiler> describes the Linux kernel, gcc version or Visual Studio version used. It should be set to what is used in $TBBROOT/lib/<arch>.

For windows users, the folder TBBROOT/bin/<arch>/<compiler> should be added to the PATH variable.

Note that the variables in the table below are being used.

Variable Description Type
TBBROOT Directory of Tbb default installation Environment
TBB_INCLUDE_DIRS Directory containing the tbb/tbb.h file CMake
TBB_LIBRARY_DIRS Directory(ies) containing the compiled TBB libraries CMake
TBB_LIBRARIES Full pathnames of the compiled TBB libraries (both release and debug versions, using "optimized" and "debug" CMake keywords). Note that if the debug versions are not found, the release versions will be used instead for the debug mode. CMake
TBB_RELEASE_LIBRARY Full pathname of the compiled TBB release library CMake
TBB_MALLOC_RELEASE_LIBRARY Full pathname of the compiled TBB release malloc library CMake
TBB_DEBUG_LIBRARY Full pathname of the compiled TBB debug library CMake
TBB_MALLOC_DEBUG_LIBRARY Full pathname of the compiled TBB debug malloc library CMake
TBB_MALLOCPROXY_DEBUG_LIBRARY Full pathname of the compiled TBB debug malloc_proxy library (optional) CMake
TBB_MALLOCPROXY_RELEASE_LIBRARY Full pathname of the compiled TBB release malloc_proxy library (optional) CMake

Compiler Workarounds

A number of boolean flags are used to workaround compiler bugs and limitations. They all start with the prefix CGAL_CFG. These flags are used to work around compiler bugs and limitations. For example, the flag CGAL_CFG_NO_CPP0X_LONG_LONG denotes that the compiler does not know the type long long.

For each installation a file <CGAL/compiler_config.h> is defined, with the correct settings of all flags. This file is generated automatically by CMake, and it is located in the include directory of where you run CMake. For an in-source configuration this means CGAL-4.14.2/include.

The test programs used to generate the compiler_config.h file can be found in config/testfiles. Both compiler_config.h and the test programs contain a short description of the problem. In case of trouble with one of the CGAL_CFG flags, it is a good idea to take a look at it.

The file CGAL/compiler_config.h is included from <CGAL/config.h>. which is included by all CGAL header files.

Compiler Optimizations

By default CMake generates makefiles for Release mode, with optimization flags switched on, and vcproj files for Release and Debug modes.