• Windows Build
Windows Build

This document describes how to build SUMO under Windows using only freely available (this does not mean "open source") tools. Instructions on how to build SUMO on Windows using an Open Source toolchain are included in our building on Linux page. Please note that you may also download pre-build Windows binaries.

Prerequisites#

CMake settings can easily be modified, loaded and saved using Project -> CMake Settings in Visual Studio.


Editing the CMake settings using Visual Studio

  • (after 30 days) Setup a Microsoft account (if you do not already have one) and register your Visual Studio

Further notes#

If you need a different python version or want to test with multiple Pythons you can either install them directly from Visual Studio or Download Python for Windows and install it. Most SUMO tools should work with Python 2 and 3. Please make sure that you install the recommended python modules as explained above and that the PATH environment variable contains the directory where the desired Python interpreter is.

If you want to build SUMO in Debug mode, then you will need the Python Debug binaries. The Python interepreter that comes with Visual Studio doesn't install these libraries by default (at least in the case of Visual Studio Community 2022); you need to install the Data Science stack in addition to the Python and C++ development stacks mentionned above. More precisely, the component you need is called Python native development tools, make sure you check the correct box.

If you want to clone / checkout a special SUMO version, you can of course do it from the command line (if you have installed the command line tools) using git clone --recursive https://github.com/eclipse/sumo or download and extract a source package, see Downloads.

The command for the Libraries is: git clone --recursive https://github.com/DLR-TS/SUMOLibraries. If you do not place the libraries in the same folder as SUMO, you will need to set the SUMO_LIBRARIES environment variable to the directory.

If Visual Studio fails at first try, it is probably because the SUMOLibraries are not cloned yet or are in an unusual location. The errors are due to Xerces not being found. Try:

  • Select Project->"Generate Cache" to try again (it is CMake->Generate in earlier Visual Studio versions) or see below.

Adapting settings#

If you need to modify settings, you can edit the CMakeCache.txt by opening it in your favorite text editor or via the Project->CMake-Cache menu. The following things might be useful:

  • If the libraries are not found, set SUMO_LIBRARIES:PATH to something like C:/Users/testus/source/repos/SUMOLibraries) and retry
  • If a wrong Python interpreter or library is found, edit the PYTHON_* variables
  • If you want to disable building the GUI (Fox) or usage of Proj, set the according library entries to an empty string

To make a debug build, you should not change CMAKE_BUILD_TYPE in the cache file. You should choose a different configuration in the GUI instead.

Libraries#

We provide a central location for getting all dependent libraries at https://github.com/DLR-TS/SUMOLibraries. The easiest way is to clone this repository and define an environment variable SUMO_LIBRARIES pointing to the resulting directory. They are build using Visual Studio 2019, but may be used with earlier and later versions as well. You may need to install the Visual C++ 2019 Runtime Distributable for running SUMO (tested with Visual Studio 2019).

For details on building your own and also on how to use different versions and additional libraries see Installing/Windows_Libraries. Make sure that the subdirectories fox-1.6.54/lib, proj_gdal-1911/bin and xerces-c-3.2.0/bin are in PATH. Note: for 32 bits compilations folders are 32bits/fox-1.6.54/lib, 32bits/proj_gdal-1911/bin and 32bits/xerces-c-3.2.0/bin. You can add both to the path but always add the 64 bit version first.

Install python packages#

Compiling netedit requires a list of python packages to generate templates. Install it using pip:

pip install google lxml rtree pandas matplotlib pulp pyproj ezdxf scipy fmpy ortools texttest pyautogui pyperclip

Manual CMake configuration#

Method 1: Via CMake GUI#

  • Start the CMake gui
  • Select the source path (e.g. D:\projects\sumo)
  • Select the build path (e.g. D:\projects\sumo\cmake-build)
    • It can be identical to the source path, but we recommend to separate them
    • Build directory will be created if necessary
  • Start configuration by pressing the "Configure" button
    • Select compiler (the default should do)
  • Create the Visual Studio solution by clicking on "Generate"
  • Open the generated solution (.sln file) (e.g. D:\projects\sumo\cmake-build\SUMO.sln) using the File Explorer or by clicking on "Open Project"
  • Build the configurations you want
    • Binaries will appear in D:\projects\sumo\bin

Visual guide:


Selecting Solution (Visual Studio, Eclipse, etc.) in the CMake gui


Libraries successfully configured


Generated solution

Method 2: Via CMake command line#

  • Go to the source path (e.g. cd D:\projects\sumo)
  • Create a build directory (e.g. mkdir cmake-build && cd cmake-build)
  • Run CMake with the correct Visual Studio version (e.g. cmake .. -G "Visual Studio 16 2019 Win64")
    • For a list of all possible generators run cmake --help
  • Open the solution D:\projects\sumo\cmake-build\SUMO.sln
    • Or build from the command line using cmake --build . --config Release

Tests#

If you plan to extend SUMO yourself, or just want to know whether everything compiled OK, it is strongly recommended to have a look at Developer/Tests. This tool makes it easier to check whether some existing functionality was broken by the extensions.

CLang#

If you want to compile using CLang in Windows, just add "ClangCL" in CMake's "Optional Toolset" (-T).

Troubleshooting#

  • Note on installation paths: MSVC seems to have difficulties with include and library paths containing spaces (for instance C:\Program Files). Thus try to avoid installing SUMO or any of the libraries in such paths.
  • Python cannot be found
    • Make sure that there are not two different Python versions installed
    • Python libraries can be specified manually (e.g. <PythonFolder>\libs\python<version>.lib)


Python library fields in the CMake Cache


Two different Python versions at the same time

You probably have installed a library to a path containing white spaces in its name. In such a case, the according environment variable should be embedded in quotes (").

Example: set FOX="D:\my libs\fox-1.6.36"

Failure on pre-build event (missing version.h or *typemap.h)#

If Visual Studio reports a failed pre-build event you can safely ignore this, unless you are building from the source code repository. In this case you should probably install Python. Even if Python is installed the file associations may be broken which causes the generation of src/version.h via the tools/build/version.py script to fail. Either repair your file associations or undefine HAVE_VERSION_H in src/windows_config.h.

If you did install Python correctly, double check that it passes command line arguments. For a quick fix, you can also execute the following commands manually:

tools\build\version.py
tools\build\typemap.py

Execution cannot proceed because MSVCR120.dll/MSVCR140.dll was not found#

Install Microsoft Visual C++ Redistributable Packages for Visual Studio 2012 (for MSVCR120.dll) or Microsoft Visual C++ Redistributable Packages for Visual Studio 2015 (for MSVCR140.dll). You can check if all dependencies are correct using Dependencies.

In debug mode, execution cannot proceed because MSVCR120D.dll/MSVCR140D.dll was not found#

Your version of Visual Studio doesn't support Debugging, you can only compile in release mode.

How to build JuPedSim and then build SUMO with JuPedSim#

Build JuPedSim#

1. You need to clone the repository https://github.com/JuPedSim/jpscore and checkout the SUMO-Integration branch.

2. You need to install vcpkg (documentation available here). We need vcpkg to install JuPedSim's dependencies.

3. You need to change the default platform used when installing packages. To do so, set the dedicated environment variable as follows:

export VCPKG_DEFAULT_TRIPLET=x64-windows

Note that the scope of this variable is terminal-bound: if you close your terminal you will have to set the variable again.

4. Then we use vcpkg in a certain way to install a specific version of the fmt library (later versions give compilation issues with JuPedSim). To do so, we use the following manifest file:

{
  "name": "jupedsim",
  "dependencies": [
    {
      "name": "fmt",
      "version>=": "8.0.1"
    },
  "builtin-baseline": "35443ee2753f46c52ac342fa6c9c48e9f5eb9105"
}

The last line is compulsory. Store it as vcpkg.json and put it at the root directory of vcpkg, on the same level as vcpkg.exe. Then simply run vcpkg.exe install, vcpkg will detect your manifest file and download the packages inside accordingly. At the moment this is the only way to download a package with a specific version.

5. Now delete the manifest file. Using vcpkg too, download and compile the other third-party dependencies with the command generic command vcpkg.exe install package-name:triplet (a default triplet has been set above):

vcpkg.exe install boost zlib spdlog poly2tri glm cgal pybind11

Note that the triplet x64-windows is used for both the Release and the Debug compilation modes.

6. Copy the files for the fmt library that are located in /c/Users/[username]/[path-to-vcpkg]/vcpkg_installed/x64-windows into /c/Users/[username]/[path-to-vcpkg]/installed/x64-windows so that all the third-party libraries share a common tree. In case you want to build the Debug binaries, you also need to copy /c/Users/[username]/[path-to-vcpkg]/installed/x64-windows/debug/lib/poly2tri.lib to /c/Users/[username]/[path-to-vcpkg]/installed/x64-windows/lib/poly2tri.lib (replacement).

7. In the 'jpscore' directory, create a 'build' directory (or better: a 'build_release' directory, in case you also want Debug binaries). From that directory, run CMake as follows:

cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=/c/Users/[username]/[path-to-vcpkg]/installed/x64-windows/ -DBUILD_JPSVIS=OFF -G"Visual Studio 15 2017 Win64" ..

The flag BUILD_JPSVIS is turned off to disable the build of JuPedSim's visualization tool. Use -DCMAKE_BUILD_TYPE=Debug for Debug binaries.

8. Then open the Visual Studio solution file that has been created in that directory, select either the Debug or Release mode at the bottom of Visual Studio and then build.

9. You can copy to some place the freshly built binaries by launching the following command from your build directory:

cmake --install . --prefix /c/Users/[username]/[path-to-jpscore]/install

You will need this installation path later when compiling SUMO with JuPedSim. Add --config Debug to the previous command line for Debug binaries.

Build SUMO with JuPedSim#

1. Build SUMO as usual, Visual Studio will say that it can't find JuPedSim. Then in the CMake cache file, search for the JUPEDSIM_DIR variable and set it with the installation path mentioned above:

JUPEDSIM_DIR:PATH=C:\Users\[username]\[path-to-jpscore]\install

Notice how the path is written (Windows-style convention).

2. Then rebuild SUMO; JuPedSim is found.

3. The JuPedSIM and third-party binaries need to be copied to the SUMO 'bin' directory for execution (jupedsim.dll, mpfr-6.dll, gmp-10.dll, spdlog.dll and fmt.dll). These binaries are located in the 'build/bin' subdirectory of jpscore or in the subdirectory 'installed' of vcpkg (or 'installed/debug' for Debug binaries; in that case some binaries have an additional 'd' at the end).

4. Also note that in order to use JuPedSim, you need to put <pedestrian.model value="jupedsim"/> in your SUMO config file.