Skip to content

Latest commit

 

History

History
186 lines (135 loc) · 9.82 KB

install-yarp.md

File metadata and controls

186 lines (135 loc) · 9.82 KB
Raw

Install YARP

We use YARP for communications. Official page: link

Install YARP (Ubuntu)

Installing YARP on Ubuntu is quite straightforward.

Note that you will be prompted for your password upon using sudo a couple of times.

As can be seen, here we are accounting for: YARP lib_math, the GUIs and mjpeg carrier.

sudo apt install build-essential git
sudo apt install libeigen3-dev # Needed for creating YARP lib_math used for kinematics, etc.
sudo apt install qtbase5-dev qtdeclarative5-dev qtmultimedia5-dev qtdeclarative5-qtquick2-plugin qtdeclarative5-window-plugin qtdeclarative5-qtmultimedia-plugin qtdeclarative5-controls-plugin qtdeclarative5-dialogs-plugin libqt5svg5
sudo apt install libjpeg8-dev # Needed for mjpeg carrier
sudo apt install libedit-dev # Enables keyboard arrow keys within an RPC communication channel via terminal
mkdir -p ~/repos; cd ~/repos # Create $HOME/repos if it doesn't exist; then, enter it
git clone https://github.com/robotology/yarp
cd yarp && mkdir build && cd build
cmake .. -DSKIP_ACE=ON -DCREATE_LIB_MATH=ON -DCREATE_GUIS=ON -DCREATE_OPTIONAL_CARRIERS=ON -DENABLE_yarpcar_mjpeg=ON # configure
make -j$(nproc) # Compile
sudo make install && sudo ldconfig && cd # Install and go home

For additional options use ccmake instead of cmake.

A nice feature is to enable yarp auto-completion:

source ~/repos/yarp/scripts/yarp_completion # Activate in current bash session
echo "source ~/repos/yarp/scripts/yarp_completion" >> ~/.bashrc # Activate in future bash sessions

Note for Linux Mint 17.3 Rosa

sudo apt install libqt5opengl5-dev # Avoid error on yarpmanager/builder GUI

Note for Linux Mint 18.3 Sylvia

(tested on 64-bit XFCE desktop)

sudo apt install libqt5opengl5-dev # Avoid error on yarpmanager/builder GUI
sudo apt-get install qml-module-qt-labs-settings qml-module-qt-labs-folderlistmodel # Fix QQmlApplicationEngine failed to load component

Install YARP (Windows)

Binary releases usually work well: http://www.yarp.it/download.html

Install Additional Plugins: Carriers

Install ROS support

Make sure you have installed previously YARP and that the ROS environment is not sourced, as it causes the build to fail. If you have the line source /opt/ros/indigo/setup.bash at the end of ~/.bashrc, comment it, save the file and open a new terminal.

cd ~/repos/yarp/build
cmake .. -DCREATE_OPTIONAL_CARRIERS=ON -DENABLE_yarpcar_tcpros=ON -DENABLE_yarpcar_xmlrpc=ON
make -j$(nproc) # Compile
sudo make install && sudo ldconfig && cd # Install and go home

Install Additional Plugins: Devices

Install additional YARP device: OpenNI2DeviceServer (Ubuntu)

Note: tested on YARP 2.3.68.x and YARP 2.3.72.1, not compatible with YARP 3.x.

Make sure you have previously installed YARP and:

cd ~/repos/yarp/build
cmake .. -DCREATE_DEVICE_LIBRARY_MODULES=ON -DENABLE_yarpmod_OpenNI2DeviceServer=ON -DENABLE_yarpmod_OpenNI2DeviceClient=ON -DOpenNI2_INCLUDE_DIR=/usr/local/include/OpenNI2/ -DOpenNI2_LIBRARY=/usr/local/lib/libOpenNI2.so
make -j$(nproc) # Compile
sudo make install && sudo ldconfig && cd # Install and go home

You should now be able to launch yarpdev --device OpenNI2DeviceServer. It is a complex device, see options with yarpdev --device OpenNI2DeviceServer --verbose (where there is an option to see modes) or example here.

Note (ref): installation of depthCamera is broken at YARP 2.3.72 (and probably at 2.3.70.x, too). Did work at 2.3.68.x, fixed upstream at 2.3.72.1. Workaround: do cmake . -DYARP_HAS_OpenNI2=ON and configure again.

Install additional YARP device: OpenNI2DeviceServer (Ubuntu) with NiTE (skeletons)

Note: not compatible with YARP 3.x.

NiTE only required for skeletons. In addition to above steps for OpenNI:

cd ~/repos/yarp/build
cmake .. -DNITE2_INCLUDE_LOCAL=/usr/local/include/NiTE-Linux-x64-2.2 -DNITE2_LIBRARY=/usr/local/lib/libNiTE2.so
make -j$(nproc) # Compile
sudo make install && sudo ldconfig && cd # Install and go home

You may need to launch yarpdev --device OpenNI2DeviceServer from /YOUR_PATH_TO/NiTE-Linux-x64-2.2/Redist if using NiTE.

Install Bindings

Swig is needed in order to build all language bindings. Refer to Install SWIG.

Install Python bindings

Make sure you have previously installed YARP and:

sudo apt update
sudo apt install libpython-dev # Python development package are not installed by default on clean distros
cd ~/repos/yarp/build
cmake .. -DYARP_COMPILE_BINDINGS=ON -DCREATE_PYTHON=ON
make -j$(nproc)  # compile
sudo make install && sudo ldconfig && cd # Install and go home

You should now be able to launch python and import yarp.

Install Python bindings (troubleshooting)

Also, extra care should be taken with multiple Python versions (e.g. 2.x vs 3.x). The following command has been tested on Ubuntu Xenial to force Python 3.5m (note distro version is 3.5m, where m is --with-pymalloc ref):

cmake -DYARP_COMPILE_BINDINGS:BOOL=ON -DCREATE_PYTHON:BOOL=ON -DYARP_USE_PYTHON_VERSION=3.5 -DPYTHON_INCLUDE_DIR=/usr/include/python3.5m -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.5m.so -DCMAKE_INSTALL_PYTHONDIR=lib/python3.5/dist-packages ..

In this specific case, CMAKE_INSTALL_PYTHONDIR is apparently ignored, so you should:

sudo ln -s /usr/local/lib/python3/dist-packages/_yarp.so /usr/local/lib/python3.5/dist-packages/
sudo ln -s /usr/local/lib/python3/dist-packages/yarp.py /usr/local/lib/python3.5/dist-packages/

Install Java bindings

First, install the Java JDK. There are several versions at hand, so make sure you pick the Java release you are later going to work with in your applications.

Make sure you have previously installed YARP and:

sudo apt update
sudo apt install openjdk-8-jdk  # on Trusty, use openjdk-7-jdk
cd ~/repos/yarp/build
cmake .. -DYARP_COMPILE_BINDINGS=ON -DCREATE_JAVA=ON
sudo make install; sudo ldconfig; cd  # install and go home

Install Java bindings (troubleshooting)

In case you run into trouble because of CMake not finding Java paths, we'll define here a variable that may hold different values depending on the system you are working on, and use it later. As a prerequisite, inspect the usual JVM installation path with ls /usr/lib/jvm.

On Ubuntu 16.04 64 bits, you should see a java-1.8.0-openjdk-amd64 directory symlink. Store its location in a new environment variable:

JAVA_JVM_PATH=/usr/lib/jvm/java-1.8.0-openjdk-amd64

On different installations, the path will look differently. Adjust the variable to your needs, or use the following script to handle this for you:

JAVA_JVM_PATH=$(find /usr/lib/jvm/ -name java-1.*.*-openjdk* -type l | head -n 1)

Now, run the CMake command again and proceed with the installation:

cd ~/repos/yarp/build
cmake .. -DJAVA_AWT_INCLUDE_PATH=$JAVA_JVM_PATH/include -DJAVA_AWT_LIBRARY=$JAVA_JVM_PATH/include/jawt.h -DJAVA_INCLUDE_PATH=$JAVA_JVM_PATH/include -DJAVA_INCLUDE_PATH2=$JAVA_JVM_PATH/include/linux -DJAVA_JVM_LIBRARY=$JAVA_JVM_PATH/include/jni.h
sudo make install && sudo ldconfig && cd  # Install and go home

You might need to set the CLASSPATH and LD_LIBRARY_PATH variables prior to calling YARP bindings in your application in order to find the generated *.jar and *.so files, respectively.

Install MATLAB bindings

Two options here:

  1. Classical way via Java bindings, which is similar to Python, then setting the classpath.txt and librarypath.txt files contained within MATLAB. Ref: http://wiki.icub.org/wiki/Calling_yarp_from_Matlab Working setups on Windows 10 (more complex than Ubuntu):
    • Good 64 bit: MATLAB R2017b (9.3, 64 bit) + yarp_2.3.70_v14_x86_amd64_1.exe + cmake 3.9.4 + VS 15 2017 + jdk-8u162-windows-x64.exe working nicely, only having to install things.i manually (see robotology/yarp#698) and doing javac -source 1.6 -target 1.6 *.java (which is the default in code, but must be done manually, and manual says 1.3).
    • Good 64 bit (portable): To avoid installing VS + JDK + CMake, here we've put together some pre-compiled YARP bindings (note vc_redist.x64.exe to compensate the lack of VS, and forced us to compile bindings in Release not Debug).
    • Ye-oldie 32 bit: MATLAB R2015b (8.6, 32 bit) + yarp_2.3.70_v14_x86_1.exe + cmake 3.9.4 + VS 15 2017 + jdk-8u152-windows-i586.exe + MATLAB R2015b (8.6 32 bit) working nicely, only having to install things.i manually (see robotology/yarp#698) and doing javac -source 1.3 -target 1.3 *.java with 1.5 instead.
  2. A different method: New repo directly against MATLAB: yarp-matlab-bindings. Some issues:   - Requires setting variables such as YARP_DIR and even YCM_DIR.
    • This seems to imply a separately built YCM, but see this issue.   - It links against MATLAB libs, so you must match compiler in addition to bits (Windows MATLAB 2017b only provides 64-bit mingw version).