Compiling from Source

Attention

This section is now critically out of date!

This section contains instructions for compiling the OSCAR-5 distribution from source. Currently only Linux-based systems are supported.

System dependencies

The following sections lists the major system dependencies required to compile the source distribution:

Complete C++/C build environment

A valid C++ compiler and associated std library must be available. Compilation has been tested on the following compilers:

GNU g++ v4.8.5

Apple LLVM v8.10

GNU g++ v7.x.x

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`build-essential`	`sudo apt-get install build-essential`
CentOS	`'Development Tools'`	`sudo yum groupinstall 'Development Tools'`

Python and python development package

Since rapyds makes use of C++ python bindings, the python development library is required.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`python2.7`	`sudo apt-get install python2.7`
	`python2.7-dev`	`sudo apt-get install python2.7-dev`
CentOS	`python27-python-devel.x86_64`	`sudo yum install python27-python-devel.x86_64`

Boost

All extensions require the Boost <https://boost.org> C++ library. Compilation has only been tested using version 1.58, but any version larger than this should work. This version is distributed with the source under the third_party directory.

Attention

The boost.serialization package is not guaranteed to be backwards compatible. Thus, if you want to share your archived data between different users, make sure that the same version of boost is use.

CMake

The system uses the cross platform utility cmake as its main build configuration tool. Any version larger than 2.8.11 can be used.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`cmake`	`sudo apt-get install cmake`
CentOS	`cmake`	`sudo yum install cmake`

High precision libraries

The Computational Geometry Algorithms Library (CGAL), which is used extensively for geometry processing, requires the high precision libraries GMP and MPFR.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`libgmp-dev`, `libmpfr-dev`	`sudo apt-get install libgmp-dev libmpfr-dev`
CentOS	`gmp-devel`, `mpfr-devel`	`sudo yum install gmp-devel mpfr-devel`

Miscellaneous

The python package used to communicate with remote servers requires the ffi development libraries.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`libffi-dev`	`sudo apt-get install libffi-dev`
CentOS	`libffi-devel`	`sudo yum install -y libffi libffi-devel`

Additional python dependencies

The following sections list python specific dependencies (extensions) that is better to install system wide.

Matplotlib

The extensive python plotting framework matplotlib is used to create static and interactive plots. Although it can be installed via a python distribution tool (like pip), you can prevent a lot of compilation headaches by using the package available in your distribution.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`python-matplotlib`	`sudo apt-get install python-matplotlib`
CentOS	`python-matplotlib`	`sudo yum install python-matplotlib`

VTK python bindings

Required for three dimensional visualization using the VTK pipeline.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`python-vtk`	`sudo apt-get install python-vtk`
CentOS	`vtk-python`	`sudo yum install vtk-python`

QT4 python bindings

Required for various user interface elements.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`python-qt4`	`sudo apt-get install python-qt4`
CentOS	`PyQt4`	`sudo yum install PyQt4`

Building external dependencies

The third_party directory contains a number of third party packages that you should compile if they are not already available on the system.

Boost

In a terminal:

cd OSDEV_ROOT/third_party/
tar -xvf boost_1_58_0.tar.bz2
cd boost_1_58_0
./bootstrap.sh
./b2

If you are able (and willing) to install this boost version system wide:

sudo ./b2 install

otherwise, you can point cmake to this local boost when compiling the extensions.

CGAL

In a terminal:

cd osdev/third_party/
tar -xvf CGAL-4.6.3.tar.bz2
cd CGAL-4.6.3
mkdir build
cd build
cmake ..
make
sudo make install

Note

The bundled CGAL version includes a modification that allows thread safe reference counting of objects, which enables multi-threaded builds of polyhedral structures. If you want to use your own source distribution, you can apply the patch located in the patches directory. See the README file for instructions.

Building and installing the extensions

Finally, the extension libraries can be compiled and installed:

cd OSDEV_ROOT
mkdir build
cd build
cmake ..
make
make install

Pointing to local boost

If you want to use a local boost installation, pass its installation root to the cmake command using the DBOOST_ROOT flag. For instance,

cd OSDEV_ROOT
mkdir build
cd build
cmake .. -DBoost_NO_SYSTEM_PATHS=TRUE  -DBOOST_ROOT:PATHNAME=OSDEV_ROOT/third_party/boost_1_58_0
make
make install

Setup on local host

Setup python package management

If not already installed, you will need the python package management utility pip.

Linux Distribution	Package Name(s)	Command
Debian (Ubuntu)	`python-pip`	`sudo apt-get install python-pip`
CentOS	`python-pip`	`yum -y install python-pip`

Configuring a virtual environment

When using the pre-processing system rapyds locally, it is recommended to use a python virtual environment. This avoids creating conflicts with your existing system packages.

Note

A python virtual environment is a fully function python system, including an interpreter and installed packages, which is isolated from the system wide python environment. Thus, packages installed in a virtual environment will only be available in that environment, and won’t interfere with the default system.

Get python virtual environment utilities

Activate python virtual environment using pip:

$ sudo pip install virtualenv

The virtualenvwrapper greatly simplifies the creation and use of virtual environments. Install it with pip:

$ sudo pip install virtualenvwrapper

Creating a virtual environment

Create virtual environment called rapyds:

$ mkvirtualenv rapyds --system-site-packages

Complete system configuration

To complete the installation run setup.py located in the installation root directory. If using a virtual environment, activate it first:

$ workon rapyds
$(rapyds) python setup.py --first-time

otherwise:

$ python setup.py --system-wide