- Building libgit2
- Compiling a project that uses libgit2
You must have CMake installed to compile libgit2. If it is not installed, visit the CMake download page and download the appropriate binary for your version of Windows.
Important: All terminal commands in this guide must be run from within a Visual Studio command prompt with administrator privileges. To launch this, search the start menu for “Developer Command Prompt” and choose the appropriate option. (The exact wording may vary based on your version of Visual Studio.)
You must have cmake installed to compile libgit2. You must have OpenSSL and libSSH2 installed for libgit2 to build correctly. The easiest way to install all of these is through Homebrew:
brew install cmake brew install openssl brew install libssh2
When you have downloaded the libgit2 source, unzip it and
cd into the folder. In that folder, run one of the following, depending on your operating system:
mkdir build cd build cmake .. -DBUILD_SHARED_LIBS=OFF -DBUILD_CLAR=OFF -DTHREADSAFE=ON cmake --build .
mkdir build cd build cmake .. -DBUILD_SHARED_LIBS=OFF -DBUILD_CLAR=OFF -DTHREADSAFE=ON -DCMAKE_OSX_ARCHITECTURES="i386;x86_64" cmake --build .
BUILD_SHARED_LIBS=OFFmakes libgit2 build as a static library (.lib for Windows, .a for Mac) instead of a dynamic library (.dll for Windows, .dylib for Mac). This is recommended by the libgit2 authors, as it simplifies project distribution.
BUILD_CLAR=OFFprevents cmake from also building the libgit2 unit test suite (which is only useful for libgit2 development.)
THREADSAFE=ONensures that libgit2 will be safe for use in a multithreaded application.
CMAKE_OSX_ARCHITECTURES="i386;x86_64"is a necessary flag when building for OS X or macOS. (See the libgit2 readme.)
Once libgit2 is successfully built, run the following to install the library and C/C++ headers into a directory on your system:
To install libgit2 globally:
cmake --build . --target install
To install libgit2 into a specific project:
cmake .. -DCMAKE_INSTALL_PREFIX=\path\to\your\project cmake --build . --target install
cmake .. -DCMAKE_INSTALL_PREFIX=<install directory> cmake --build . --target install
<install directory> can be replaced with whatever path you like. You might do
/usr/local for a system-wide installation, or
/path/to/your/project for a project-specific installation.
Compiling a project that uses libgit2
This is the hard part. libgit2 has secret dependencies that tend to cause lots of linker errors.
libgit2 depends on the following libraries: winhttp, Rpcrt4, Crypt32.
First follow the official instructions for linking libgit2 into a Visual Studio project.
When that is done, add the following as additional dependencies: winhttp.lib, Rpcrt4.lib, and Crypt32.lib. When finished, your Additional Dependencies line should look like the following:
libgit2 depends on the following:
- Frameworks: Security and CoreFoundation
- Libraries: curl, z, iconv, ssh2
g++ -I/path/to/git2/include \ -L/path/to/git2/lib \ -framework CoreFoundation -framework Security \ -lcurl -lz -liconv -lssh2 -lgit2 \ my_project.cpp
If you installed libgit2 in a system-standard location, you may omit the
-L lines, which tell g++ extra include and library paths.
Resolving linker errors
If you still get linker errors even when your compiler is configured correctly, you may not have all the required libraries installed. libssh2 is the most likely culprit, since it is not installed by default on Mac. You can install it with Homebrew like so:
brew install libssh2
curl, z, and iconv should be installed by default on your Mac, but if for some reason they are not, they can be installed through Homebrew as well:
brew install curl brew install zlib brew install libiconv
Homebrew will probably present warnings and extra messages when installing these typically standard libraries. Be careful.