How I set up a Cantera development environment on Windows 7

Setting up a development environment for Cantera on Windows 7 can be a little trickier than on Linux. This post contains instructions for how I set up my development environment on Windows 7.

The first step is to download the SDK 7.1 compiler and its SP1 update. Installation should be a “Next->Next” affair, but if the install fails, The Mathworks recommends the following steps:

  1. Uninstall any Visual C++ 2010 Redistributable packages
  2. Install the SDK again, but uncheck the options for the compiler and the Visual C++ Redistributable
  3. Install the compiler update, which will install the compilers
  4. Reinstall the Visual C++ Redistributable from this link

If you would like to have Visual Studio installed in addition to the SDK, the instructions here give the proper order to install everything.

If you only have the SDK and its update installed, Microsoft has not-so-helpfully removed a file that is necessary to build Cantera, called ammintrin.h. There is a post at The Mathworks forum with a link to the file, and I’ll also post it here for reference. Add this file to the C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\include directory, and it should be all set. If you have Visual Studio installed, this will not be a problem.

Next, install SUNDIALS (I’ve helpfully posted some instructions here). Then, download and unzip the Boost libraries: http://www.boost.org/ I recommend unzipping into the root of the C:\ drive, so there will be a folder called boost_1_56_0 (for the most recent version as of this writing) with all the required files inside. Note that if you’re not careful, you’ll get a folder C:\boost_1_56_0\boost_1_56_0, which is not what we want, it should be just C:\boost_1_56_0\<files go here>.

Now we need other components of the development environment, particularly Python and all its packages. The most recent versions of Python are 2.7.8 and 3.4.1, which can be downloaded respectively here and here. Make sure to get the proper version for your operating system (32-bit or 64-bit). Installation into the default directories is preferable, and make sure to have the python.exe added to your path.

If you forget to add python.exe to your path in the installer, open the System control panel (Win+Pause), choose “Advanced System Settings” in the left panel, then choose the “Advanced” tab, and at the bottom click “Environment Variables”. In the “System variables” box, find “Path” and add C:\Python34;C:\Python34\Scripts;C:\Python27;C:\Python27\Scripts; to the front of the text box.

While we’re here, we need to add one environment variable. Click “New”, and in the first text box add VS90COMNTOOLS and in the second box add %VS100COMNTOOLS%. This sets the variable called VS90COMNTOOLS to be equal to the value of the variable called VS100COMNTOOLS. VS100COMNTOOLS is set by the SDK when it is installed, but Python 2 depends on VS90COMNTOOLS being installed. This allows us to install certain packages with pip in Python 2. Note that not all packages will work with this trick (in particular, matplotlib can’t be installed with pip this way).

Python 3.4 comes with pip installed, but 2.7 does not, plus we need some other packages anyways. Head to Christoph Gohlke’s page and download the following packages. Make sure to get the correct bitness for your Python versions.

2.7:

3.4:

Install these after you install Python. Now, open the “Windows SDK 7.1 Command Prompt” (searching for 7.1 in the Start Menu search box usually brings it up) and install some additional packages for Python with the following command:

pip2 install sphinxcontrib-matlabdomain sphinxcontrib-doxylink 3to2

This will fetch all the Python packages required to build the documentation and the package (3to2) to convert the examples from Python 3 syntax to Python 2 syntax.

You should also install Doxygen at this point, so that the C++ documentation can be built. The Windows installer for the most recent version of Doxygen (at the time of this writing, 1.8.8) is available on SourceForge. Finally, install the WiX Toolset so that MSI installers for Cantera can be built.

That should (finally) be all the software needed to get up and running in developing Cantera. The last thing to do is actually get the Cantera source code. There are several ways to get the code; my preferred way is to use GitHub. Ray Speth has a clone of the main Cantera repository on GitHub: https://github.com/Cantera/cantera. What you should do is create a fork of this project into your own personal GitHub account. This way, you can add branches, do all your work, push them to your forked repository, and create a pull request so that Ray can see the code and help you improve it. Then, either use the GitHub for Windows client, or clone the repository yourself.

Once in the main Cantera directory, we need to tell SCons how to build Cantera. Create a file called cantera.conf and put as its contents:

msvc_version = '10.0'
python_cmd = 'C:/Python27/python.exe'
python3_cmd = 'C:/Python34/python.exe'
matlab_path = 'C:/Program Files/MATLAB/R2014a'
sundials_include = 'C:/sundials-2.5.0-install/include'
sundials_libdir = 'C:/sundials-2.5.0-install/lib'
env_vars = 'all'
boost_inc_dir = 'C:/boost_1_56_0/'

Now, scons build and scons msi should work. If you get errors, read on.

The first warning you might see is that 3to2 is not installed properly, even though we installed it earlier. This is because the SConstruct file calls 3to2 as an executable, which doesn’t work on Windows. The fix for this is to edit two lines, one in the root SConstruct file, one in the interfaces/cython/SConscript file. First, in SConstruct change:

ret = getCommandOutput('3to2','-l')

to

ret = getCommandOutput(env['python_cmd'], 'C:/Python27/Scripts/3to2','-l')

Then, in interfaces/cython/SConscript, change:

subprocess.call(['3to2', '--no-diff', '-n', '-w','-x', 'str',

to

subprocess.call([env['python_cmd'], 'C:/Python27/Scripts/3to2', '--no-diff', '-n', '-w','-x', 'str',

Note that these are basically hacks until the problem can be fixed in the source.

As of eeab08, this problem has been fixed.

Second, you may get actual errors when trying to build the MATLAB interface (if you aren’t building this interface, you might not see this error). In case you get an error in the MATLAB interface about redefining _char16t, open src/matlab/ctmatutils.h and comment out lines 12-14. These lines were introduced to fix some problems in MATLAB R2010a, and aren’t required for at least R2013a and R2014a.