The GMAT, GMAT Plugins, and CSALT build process is managed by CMake and custom scripts that automatically install required dependencies. (CSALT currently requires some manual configuration.) The build process can be broken down into several steps:
The first two three steps are generally "one-time" processes and result in a build system (e.g. Visual Studio solution or makefiles) that will intelligently rebuild GMAT components as needed when source or configuration files are changed. NOTE: These instructions refer to <GMAT>
as the top-level GMAT repository folder.
Figure 1. The <GMAT>
repository layout
GMAT uses to Git repositories for source code version control. The "Public" repository contains code that has been approved for general open source release. The internal repository contains proprietary and/or sensitive components that cannot be released open source. Public developers can obtain the source code from SourceForge (the Git command is: "git clone git://git.code.sf.net/p/gmat/git
gmat-git
"). Internal developers should contact the GMAT team for repository URLs.
When building with internal plugins, if the internal repository is placed next to the main |
GMAT is distributed with a Python script to configure most (but not all) dependencies. Table 1 describes all software dependencies for GMAT. The <GMAT>/depends
folder contains scripts to automatically download and configure the core GMAT dependencies. Note, to build without MATLAB or Python (or other non-Core dependencies), you can simply uncheck the plugins that require those dependencies later in the CMake configuration (see the "Used in GMAT" column in the table below). Core dependencies are required to build even the minimal GMAT distribution.
Table 1. GMAT Dependencies
Name | Version | Used in GMAT | Download |
---|---|---|---|
CSPICE | N0067 | Core Dependency | Configure Script |
wxWidgets | 3.0.2 | Core Dependency | Configure Script |
Xerces | 3.1.4 | Core Dependency | Configure Script |
MATLAB | R2015a+ | CInterface Plugin MatlabInterface Plugin | Self-download |
Python | 3.6, 3.7, 3.8, 3.9 | PythonInterface Plugin | Mac, Windows Linux (Package Manager) |
Additional CSALT Dependencies (See below for CSALT dependency configuration) | |||
boost | 1.69 | CSALT CSALTInterface Plugin | CMake configuration |
SNOPT | 7.5 | CSALT | Self-download; place in |
2017 Intel Fortran Redistributable Libraries | Release 4 (ww_ifort_redist_msi_2017.4.120.zip) 64-bit | CSALT CSALTInterface Plugin | Self-download; Windows users only |
Requirements:
Figure 2. <GMAT>/depends
folder layout after dependency configuration
configure.sh
(Mac/Linux) or configure.bat
(Windows) script to set up core GMAT dependencies.<GMAT>/depends
folder structure should look like Figure 2.configure.py
script to set up core GMAT dependencies.<GMAT>/depends
folder structure should look like Figure 2.→ Download optional GMAT dependencies.
As shown in Table 1, multiple GMAT components require a Python distribution be installed on the developer's system. At a minimum, developers are encouraged to install the same version of Python that is used in the GMAT test system. By default GMAT will search for Python versions listed in Table 1, and create versions of each component corresponding to those Python versions. Specifying which version of Python to use at runtime is described in Step 5 (Run GMAT) below.
Developers can override the Python versions that GMAT searches for using the GMAT_PYTHON3_VERSIONS
CMake variable. They can also specify a custom Python install location using the GMAT_PYTHON3X_ROOT_DIR
CMake variable. These variables are documented in Table 2. Modifying either of these variables is uncommon, and should be done with intention and care.
The GMAT Windows, macOS, and Linux build machines create distributable versions of GMAT, and therefore should install all versions of Python listed in Table 1. |
CSALT requires installation of the code and build files for two additional third-party components: Boost (code only) and SNOPT (code and libraries). The Boost code is downloaded automatically as part of the CMake configuration process. (See Step 3 of this guide.) The SNOPT code and libraries should be placed in a folder named snopt7. The dependency folder ("depends"), configured to include the CSALT dependencies, is shown in Figure 3. Note that some of the items in the depends folder do not appear until either (1) the configure.py script is run or (2) the CMake configuration process is run. (See Step 3 of this guide.)
Figure 3. <GMAT>/depends
folder layout, including the folders for CSALT builds
When the Boost and SNOPT folders are configured correctly, the code for the corresponding components should be directly inside of the Boost_1_71_0 and snopt7 folders, as shown in Figure 4.
Figure 4. A view showing contents of the boost and part of the snopt7 folders
CSALT is built to use SNOPT 7.5. Civil servant members of the GMAT development team can obtain a copy of the SNOPT code from the GMAT Project lead. Subcontractors should obtain a copy from their company GMAT point of contact. Other GMAT developers should contact Stanford Business Software, Inc. (SBS) for a licensed copy of the software package. Setup for the SNOPT package is different for developers that have access to the code from the GMAT project than for developers that are using code from SBS.
The SNOPT source code is contained in an archive that your team lead will provide to you. Unpack that archive into your depends folder, and verify the resulting snopt7 fonder contains the file structure shown in Figure 4.
Windows, Mac, and Linux developers with access to the GMAT internal repository can find prebuilt copies of the SNOPT libraries in platform specific subfolders of the <gmatinternal>/code/CSALTPlugin/snoptlibs folder. Precompiled libraries are available there for Windows 10, Linux (built on Ubuntu) and Mac OS X (built on High Sierra). Thise libraries will need to be copied into place in the bin and debug subdirectories of the GMAT output directory that you will configure later in the build configuration. In the GMAT CMake settings, that directory is labeled GMAT_BUILDOUTPUT_DIRECTORY, and defaults to the <GMAT>/application folder.
Developers working on other platforms will need to build SNOPT for their environments. The build procedure is described in the snopt7/INSTALL text file. Be sure to build both the SNOPT shared library and the C++ interface shared library for the target platform.
SBS delivers the SNOPT code in an archive file. Unpack the SNOPT 7.5 archive, snopt7.5-1.4 into the depends folder. On success, the depends folder should contain the snopt7 folder used to build CSALT. Follow the SNOPT build instructions, found in the INSTALL text file of the snopt7 folder, to create theSNOPT shared library and the SNOPT C++ interface library. Once built, the libraries snopt7 and snopt7_cpp (with platform specific file extensions) are located in the lib/.libs folder. Copy the shared libraries to the .lib folder, completing installation.
A note for Windows developers: SNOPT is coded in Fortran. You will need a Fortran compiler to build SNOPT for your workstation. The libraries built for the GMAT development team were compiled using the Intel Fortran compiler. Windows users of the CSALT libraries also install the Intel redistributable libraries as part of the GMAT/CSALT installation procedure. (See "Additional CSALT Dependencies" in Table 1)
Requirements (in addition to Step 1 requirements):
Launch Matlab, run the command: mex -setup
Tip: Select the "Grouped" option in the CMake GUI (Figure 5) to sort CMake variables and make them easier to find.
Tip: Select the "Advanced" option in the CMake GUI (Figure 5) to display variables, such as PYTHON_LIBRARY, that are normally hidden.
Tip: All CMake commands can also be performed on the command-line instead of using the GUI. See below for instructions.
Figure 5. Components of the CMake GUI
<GMAT>
cloned repository on your computer<GMAT>/build/<OS>-cmakebuild
for this value.Figure 6. Choosing a generator in CMake
4a
and 4b
) will be populated:Figure 7. Output of CMake Configure
Use the Configure results output box (Figure 3 section 4a
) to change variables in the CMake variable list (Figure 3 section 4b
) as follows:
In addition to errors, there are several CMake variables that allow you to control how the build system configures GMAT:
CMake Variable (Group) | Description | Associated CMake Error |
---|---|---|
CMAKE_BUILD_TYPE (CMAKE) | On makefile systems, this specifies the desired build type On VisualStudio/XCode systems, this specifies all possible build types Valid values: Release , Debug , RelWithDebInfo , MinSizeRel Note: On makefile systems, you should create a separate out-of-source build folder for each desired build type (Figure 3 box 2 ). (e.g. cmakebuild-release or cmakebuild-debug ) | N/A |
CMAKE_INSTALL_PREFIX (CMAKE) | Location to install GMAT when doing make install or building the VisualStudio-> INSTALL project Note: This is autofilled to <GMAT>/GMAT-<release>-<OS>-<BitType> for convenience. Note libraries are only placed here when building AND installing. Note this directory must be different than "Where to build the binaries", and the target of GMAT_BUILDOUTPUT_DIRECTORY. | N/A |
CSPICE_DIR (Ungrouped) | Path to CSPICE root directory (containing include/ , lib/ , etc.) | CSPICE NOT FOUND (make sure to run depends script from Step 1) |
F2C_DIR (Ungrouped) | Path to F2C root directory (containing f2c.h) Note: this should generally be CSPICE_DIR/include | F2C NOT FOUND (make sure to run depends script from Step 1) |
GMAT_ADDITIONAL_PLUGINS (GMAT) | Full path to a text file containing locations of additional (i.e. 3rd-party) GMAT plugins. The file contents should follow the following format: | N/A |
GMAT_BUILDOUTPUT_DIRECTORY (GMAT) | Location to place libraries and executables when building but NOT installing. Note this directory must be different than "Where to build the binaries", and the target of CMAKE_INSTALL_PREFIX. | |
GMAT_INCLUDE_API (GMAT) | Build the Python and/or Java APIs for GMAT. If the APIs for the Estimation and Station components are desired, then those plugins should also be enabled. | |
GMAT_PROPRIETARYPLUGINS_PATH (GMAT) | Full path to the top-level GMAT Proprietary Plugins (folder that contains CMakeLists.txt) This will be automatically found if you name it | N/A |
GMAT_PYTHON3_VERSIONS (GMAT) | Semicolon-separated list of Python versions to use with each Python-aware GMAT component (see Table 1). | |
GMAT_PYTHON3X_ROOT_DIR (GMAT) | Absolute path to the root directory of a specific Python 3.X installation. Leave blank to auto-find Python 3.X in standard install locations. Note: One of these variables will exist for each Python 3.X specified in GMAT_PYTHON3_VERSIONS | |
Matlab_ROOT_DIR (Matlab) | Path to MATLAB root directory (on Mac, this is the path to MATLAB_R20xxx.app) | Matlab NOT FOUND (make sure MATLAB is installed) |
PLUGIN_XXX (PLUGIN) | Whether to build a particular GMAT Plugin | N/A |
wxWidgets_ROOT_DIR (wxWidgets) | Mac/Linux: Path to wxWidgets wx-config utility(usually this is the wxWidgets bin/ folder)Windows: Path to wxWidgets (containing include/ and lib/ ) | wxWidgets NOT FOUND (make sure to run depends script from Step 1) |
PYTHON_LIBRARY (PYTHON) (Advanced Variable) | Use a specific installation of Python for GMAT's PythonInterface plugin. If this variable is blank, the latest version of Python that can be found in default locations will be used. Set this to the FULL PATH to pythonXX.lib (e.g. C:/path/to/python35.lib) if you installed Python to a custom location. | Python NOT FOUND (make sure Python is installed, and variable is set properly) |
XercesC_INCLUDE_DIR XercesC_LIBRARY (Xerces) | Xerces library and include folder locations. | Failed to find XercesC (make sure to run depends script from Step 1) |
When all CMake errors are handled and you have specified all desired GMAT options, click "Generate".
CMake will create the build system in the chosen out-of-source build folder (Figure 3 box 2).
CMake is fully scriptable and can be called from the command line instead of using the GUI. This is especially useful on operating systems (e.g. Red Hat Linux 7) where the GUI is unavailable.
cd <GMAT>/build; mkdir macosx-cmake; cd macosx-cmake
<GMAT>
folder and specify optionscmake [options] ../..
-DOPTION=VALUE
formatting. e.g. -DPLUGIN_CINTERFACE=OFF
. Multiple such options can be specified.Beginning with the beta leading to GMAT R2019a, builds for the Collocation Stand Alone Library and Toolkit (CSALT) are incorporated into the GMAT build system. The CSALT dependencies are included in the list of dependencies above. CSALT users set up the build system by following these additional steps:
Configuration
The CSALT build is controlled by checkboxes on the CMake GUI in the Ungrouped Entries section if CSALT has not been built before, or in the GMAT section if it has been built. The CSALT build toggle is labeled GMAT_INCLUDE_CSALT:
When you activate the CSALT build setting and configure the CMake build again, another CSALT setting, GMAT_INCLUDE_CSALT_TESTPROGRAM, will be added to the Ungrouped Entries group. This switch controls builds of the CSALT test program. The next time (and all subsequent times) that the CMake GUI configure button is toggled, the CSALT settings will be grouped together, as in the image above.
The GMAT CSALT interface is part of the GMAT internal repository. Users with access to that repository can build the GMAT interface plugins for CSALT (PLUGIN_PROPRIETARY_CSALT) and the EMTG spacecraft models (PLUGIN_PROPRIETARY_EMTG_MODELS).
Activating the CSALT build switches turns on the corresponding build settings, producing the CSALT. Installing the resulting build includes installation of the SNOPT and CSALT libraries and the CSALT test executable (if selected to build). The GMAT interfaces can be activated by adding the PLUGIN entries for the build CSALT libraries in the GMAT startup file.
GMAT.sln
Visual Studio solution. After loading, you should see the following projects:Figure 8. CMake-generated Visual Studio Solution
ALL_BUILD
: The default startup project. Ensures that all other projects are up-to-date, then builds them.ZERO_CHECK
: Performs the work to ensure all other projects are up-to-date. It is automatically built along with all other projects.INSTALL
: Creates a standalone GMAT folder containing all executables, plugins, data files, samples, and documentation.GmatBase
, GmatConsole
, GmatGUI
, Plugins: The various GMAT components. You can build these individually if desired.Release
, Debug
, etc.ALL_BUILD
project. Depending on your system speed and number of selected GMAT components, this may take a while!INSTALL
project if you want a fully standalone and relocatable version of GMAT.Makefiles are run through the command line, which on Mac and Linux can be accessed via the Terminal application.
In these instructions, <CMake_build_path>
is the path to the build system folder that you chose in Step 2 (Figure 3 box 2).
$
is your command prompt):$ cd <CMake_build_path>
$ make
make -jN
" to significantly speed up the compile timemake install
" if you want a fully standalone and distributable version of GMAT.See the following table in case there are build errors when compiling or installing GMAT.
Error | GMAT Component | Description | Fix |
---|---|---|---|
Unresolved External Symbol *_Py_* referenced in function ... | Python Interface | 32/64-bit Python found by CMake is different than architecture of compiler (VisualStudio, gcc, ...) | Make sure to install the correct 64-bit version of Python and specify it via the PYTHON_LIBRARY CMake variable. |
CInterface Matlab thunk files not produced during INSTALL step | C Interface | There is a known incompatibility between Matlab R2015a/b and XCode 7 that prevents the CInterface thunk files from being built. | Perform the Matlab initialization instructions. |
After building, you have several options for how to run GMAT:
CMAKE_INSTALL_PREFIX
folder that you chose in Step 2. The GMAT executables will be in the bin/
subfolder.<GMAT>/application
directory. You can run GMAT from the bin/
(or debug/
) subfolders without having to perform the optional INSTALL step. This allows for a more rapid edit-build-test development cycle. On Windows, you can also run the GmatConsole
and GmatGUI
projects directly from within VisualStudio. This allows for in-program debugging with breakpoints.Release
, Debug
, etc.)Right-click on the GmatConsole
or GmatGUI
project, and select "Set as Startup Project"
Select menu item Debug -> Start Debugging (or Start Without Debugging for Release configurations
As listed in Table 1, several GMAT components depend on Python. These components will have multiple built versions in the GMAT plugin/
subfolder, with names that clearly identify which version of Python they use.
When running GMAT, the gmat_startup_file.txt
file should contain a PLUGIN
line corresponding to the desired Python-supported version of each plugin.
wxWidgets v3.0.2 has a known bug (documented here) on Mac OSX 10.10+ that causes a build error. As of R2016a, the GMAT dependency configuration script (THIS IS NO LONGER AN ISSUE since switching to wxWidgets 3.0.4.configure.sh
) implements this fix internally, so GMAT users do not need to take any additional action for wxWidgets to build on Mac.
The GMAT project has incorporated the OpenFramesInterface (OFI) plugin as a 3D visualization replacement for OrbitView. Follow the OFI build instructions to include the OFI in your GMAT build.