Building ROS 2 on Windows

Table of Contents

This guide is about how to setup a development environment for ROS 2 on Windows.

System requirements 

Only Windows 10 is supported.

Language support 

Make sure you have a locale which supports UTF-8. For example, for a Chinese-language Windows 10 installation, you may need to install an English language pack.

Installing prerequisites 

Install Chocolatey 

Chocolatey is a package manager for Windows, install it by following their installation instructions:

https://chocolatey.org/

You’ll use Chocolatey to install some other developer tools.

Install Python 

Open a Command Prompt and type the following to install Python via Chocolatey:

choco install -y python --version 3.8.3

Install Visual C++ Redistributables 

Open a Command Prompt and type the following to install them via Chocolatey:

choco install -y vcredist2013 vcredist140

Install OpenSSL 

Download the Win64 OpenSSL v1.1.1L OpenSSL installer from this page. Scroll to the bottom of the page and download Win64 OpenSSL v1.1.1L. Don’t download the Win32 or Light versions.

Run the installer with default parameters, as the following commands assume you used the default installation directory.

This command sets an environment variable that persists over sessions:

setx -m OPENSSL_CONF "C:\Program Files\OpenSSL-Win64\bin\openssl.cfg"

You will need to append the OpenSSL-Win64 bin folder to your PATH. You can do this by clicking the Windows icon, typing “Environment Variables”, then clicking on “Edit the system environment variables”. In the resulting dialog, click “Environment Variables”, then click “Path” on the bottom pane, finally click “Edit” and add the path below.

C:\Program Files\OpenSSL-Win64\bin\

Install Visual Studio 

Install Visual Studio 2019.

If you already have a paid version of Visual Studio 2019 (Professional, Enterprise), skip this step.

Microsoft provides a free of charge version of Visual Studio 2019, named Community, which can be used to build applications that use ROS 2. You can download the installer directly through this link.

Make sure that the Visual C++ features are installed.

An easy way to make sure they’re installed is to select the Desktop development with C++ workflow during the install.

Make sure that no C++ CMake tools are installed by unselecting them in the list of components to be installed.

Install additional DDS implementations (optional)

If you would like to use another DDS or RTPS vendor besides the default, Eclipse Cyclone DDS, you can find instructions here.

Install OpenCV 

Some of the examples require OpenCV to be installed.

You can download a precompiled version of OpenCV 3.4.6 from https://github.com/ros2/ros2/releases/download/opencv-archives/opencv-3.4.6-vc16.VS2019.zip .

Assuming you unpacked it to C:\opencv, type the following on a Command Prompt (requires Admin privileges):

setx -m OpenCV_DIR C:\opencv

Since you are using a precompiled ROS version, we have to tell it where to find the OpenCV libraries. You have to extend the PATH variable to C:\opencv\x64\vc16\bin.

Install dependencies 

There are a few dependencies not available in the Chocolatey package database. In order to ease the manual installation process, we provide the necessary Chocolatey packages.

As some chocolatey packages rely on it, we start by installing CMake

choco install -y cmake

You will need to append the CMake bin folder C:\Program Files\CMake\bin to your PATH.

Please download these packages from this GitHub repository.

asio.1.12.1.nupkg
bullet.2.89.0.nupkg
cunit.2.1.3.nupkg
eigen-3.3.4.nupkg
tinyxml-usestl.2.6.2.nupkg
tinyxml2.6.0.0.nupkg
log4cxx.0.10.0.nupkg

Once these packages are downloaded, open an administrative shell and execute the following command:

choco install -y -s <PATH\TO\DOWNLOADS\> asio cunit eigen tinyxml-usestl tinyxml2 log4cxx bullet

Please replace <PATH\TO\DOWNLOADS> with the folder you downloaded the packages to.

You must also install some additional python dependencies:

python -m pip install -U catkin_pkg cryptography empy ifcfg importlib-metadata lark-parser lxml matplotlib netifaces numpy opencv-python PyQt5 pip pillow psutil pycairo pydot pyparsing pyyaml rosdistro setuptools

Install Qt5 

This section is only required if you are building rviz, but it comes with our default set of sources, so if you don’t know, then assume you are building it.

First get the installer from Qt’s website:

https://www.qt.io/download

Select the Open Source version and then the Qt Online Installer for Windows.

Run the installer and install Qt5.

We recommend you install it to the default location of C:\Qt, but if you choose somewhere else, make sure to update the paths below accordingly. When selecting components to install, the only thing you absolutely need is the appropriate MSVC 64-bit component under the Qt -> Qt 5.15.0 tree. We’re using 5.15.0 as of the writing of this document and that’s what we recommend since that’s all we test on Windows, but later Qt5 versions will probably work too. Be sure to select MSVC 2019 64-bit. After that, the default settings are fine.

Finally, set the Qt5_DIR environment variable in the cmd.exe where you intend to build so that CMake can find it:

set Qt5_DIR=C:\Qt\5.15.0\msvc2019_64
set QT_QPA_PLATFORM_PLUGIN_PATH=C:\Qt\5.15.0\msvc2019_64\plugins\platforms

You could set it permanently with setx -m Qt5_DIR C:\Qt\5.15.0\msvc2019_64 and setx -m QT_QPA_PLATFORM_PLUGIN_PATH C:\Qt\5.15.0\msvc2019_64\plugins\platforms instead, but that requires Administrator.

注解

This path might change based on which MSVC version you’re using or if you installed it to a different directory.

RQt dependencies 

To run rqt_graph you need to download and install Graphviz. The installer will ask if to add graphviz to PATH, choose to either add it to the current user or all users.

Additional prerequisites 

When building from source you’ll need a few additional prerequisites installed.

Install additional prerequisites from Chocolatey 

choco install -y cppcheck curl git winflexbison3

You will need to append the Git cmd folder C:\Program Files\Git\cmd to the PATH (you can do this by clicking the Windows icon, typing “Environment Variables”, then clicking on “Edit the system environment variables”. In the resulting dialog, click “Environment Variables”, the click “Path” on the bottom pane, then click “Edit” and add the path).

Install Python prerequisites 

Installing additional Python dependencies:

pip install -U colcon-common-extensions coverage flake8 flake8-blind-except flake8-builtins flake8-class-newline flake8-comprehensions flake8-deprecated flake8-docstrings flake8-import-order flake8-quotes mock mypy==0.761 pep8 pydocstyle pytest pytest-mock vcstool

Install miscellaneous prerequisites 

Next install xmllint:

Download the 64 bit binary archives of libxml2 (and its dependencies iconv and zlib) from https://www.zlatkovic.com/projects/libxml/
Unpack all archives into e.g. C:\xmllint
Add C:\xmllint\bin to the PATH.

Get the ROS 2 code 

Now that we have the development tools we can get the ROS 2 source code.

First setup a development folder, for example C:\dev\ros2_galactic:

md \dev\ros2_galactic\src
cd \dev\ros2_galactic

Get the ros2.repos file which defines the repositories to clone from:

# CMD
curl -sk https://raw.githubusercontent.com/ros2/ros2/galactic/ros2.repos -o ros2.repos

# PowerShell
curl https://raw.githubusercontent.com/ros2/ros2/galactic/ros2.repos -o ros2.repos

Next you can use vcs to import the repositories listed in the ros2.repos file:

# CMD
vcs import src < ros2.repos

# PowerShell
vcs import --input ros2.repos src

Install additional DDS implementations (optional)

If you would like to use another DDS or RTPS vendor besides the default, Eclipse Cyclone DDS, you can find instructions here.

Build the ROS 2 code 

To build ROS 2 you will need a Visual Studio Command Prompt (“x64 Native Tools Command Prompt for VS 2019”) running as Administrator.

Fast RTPS is bundled with the ROS 2 source and will always be built unless you put an AMENT_IGNORE file in the src\eProsima folder.

To build the \dev\ros2_galactic folder tree:

colcon build --merge-install

注解

We’re using --merge-install here to avoid a PATH variable that is too long at the end of the build. If you’re adapting these instructions to build a smaller workspace then you might be able to use the default behavior which is isolated install, i.e. where each package is installed to a different folder.

注解

If you are doing a debug build use python_d path\to\colcon_executable colcon. See Extra stuff for debug mode for more info on running Python code in debug builds on Windows.

Environment setup 

Start a command shell and source the ROS 2 setup file to set up the workspace:

call C:\dev\ros2_galactic\install\local_setup.bat

This will automatically set up the environment for any DDS vendors that support was built for.

It is normal that the previous command, if nothing else went wrong, outputs “The system cannot find the path specified.” exactly once.

Test and run 

Note that the first time you run any executable you will have to allow access to the network through a Windows Firewall popup.

You can run the tests using this command:

colcon test --merge-install

注解

--merge-install should only be used if it was also used in the build step.

Afterwards you can get a summary of the tests using this command:

colcon test-result

To run the examples, first open a clean new cmd.exe and set up the workspace by sourcing the local_setup.bat file. Then, run a C++ talker:

call install\local_setup.bat
ros2 run demo_nodes_cpp talker

In a separate shell you can do the same, but instead run a Python listener:

call install\local_setup.bat
ros2 run demo_nodes_py listener

You should see the talker saying that it’s Publishing messages and the listener saying I heard those messages. This verifies both the C++ and Python APIs are working properly. Hooray!

注解

It is not recommended to build in the same cmd prompt that you’ve sourced the local_setup.bat.

To be able to extract the Python source tarball, you can use PeaZip:

choco install -y peazip

You’ll also need SVN, since some of the Python source-build dependencies are checked out via SVN:

choco install -y svn hg

You’ll need to quit and restart the command prompt after installing the above.
Get and extract the Python 3.8.3 source from the tgz:
- https://www.python.org/ftp/python/3.8.3/Python-3.8.3.tgz
- To keep these instructions concise, please extract it to C:\dev\Python-3.8.3
Now, build the Python source in debug mode from a Visual Studio command prompt:

cd C:\dev\Python-3.8.3\PCbuild
get_externals.bat
build.bat -p x64 -d

Finally, copy the build products into the Python38 installation directories, next to the Release-mode Python executable and DLL’s:

cd C:\dev\Python-3.8.3\PCbuild\amd64
copy python_d.exe C:\Python38 /Y
copy python38_d.dll C:\Python38 /Y
copy python3_d.dll C:\Python38 /Y
copy python38_d.lib C:\Python38\libs /Y
copy python3_d.lib C:\Python38\libs /Y
copy sqlite3_d.dll C:\Python38\DLLs /Y
for %I in (*_d.pyd) do copy %I C:\Python38\DLLs /Y

Now, from a fresh command prompt, make sure that python_d works:

python_d -c "import _ctypes ; import coverage"

Once you have verified the operation of python_d, it is necessary to reinstall a few dependencies with the debug-enabled libraries:

python_d -m pip install --force-reinstall https://github.com/ros2/ros2/releases/download/numpy-archives/numpy-1.18.4-cp38-cp38d-win_amd64.whl
python_d -m pip install --force-reinstall https://github.com/ros2/ros2/releases/download/lxml-archives/lxml-4.5.1-cp38-cp38d-win_amd64.whl

To verify the installation of these dependencies:

python_d -c "from lxml import etree ; import numpy"

When you wish to return to building release binaries, it is necessary to uninstall the debug variants and use the release variants:

python -m pip uninstall numpy lxml
python -m pip install numpy lxml

To create executables python scripts(.exe), python_d should be used to invoke colcon

python_d path\to\colcon_executable build

Hooray, you’re done!

If you installed your workspace with colcon as instructed above, “uninstalling” could be just a matter of opening a new terminal and not sourcing the workspace’s setup file. This way, your environment will behave as though there is no Galactic install on your system.
If you’re also trying to free up space, you can delete the entire workspace directory with:
```
rmdir /s /q \ros2_galactic
```

Building ROS 2 on Windows

System requirements 

Language support 

Installing prerequisites 

Install Chocolatey 

Install Python 

Install Visual C++ Redistributables 

Install OpenSSL 

Install Visual Studio 

Install additional DDS implementations (optional)

Install OpenCV 

Install dependencies 

Install Qt5 

RQt dependencies 

Additional prerequisites 

Install additional prerequisites from Chocolatey 

Install Python prerequisites 

Install miscellaneous prerequisites 

Get the ROS 2 code 

Install additional DDS implementations (optional)

Build the ROS 2 code 

Environment setup 

Test and run 

Next steps after installing 

Using the ROS 1 bridge 

Additional RMW implementations (optional)

Extra stuff for Debug mode 

Stay up to date 

Troubleshooting 

Uninstall 