DFHack builds are available for all supported platforms; see Installing DFHack for installation instructions. If you are a DFHack end-user, modder, or plan on writing scripts [lua] (not plugins), it is generally recommended (and easier) to use these builds instead of compiling DFHack from source.
However, if you are looking to develop plugins, work on the DFHack core, make complex changes to DF-structures, or anything else that requires compiling DFHack from source, this document will walk you through the build process. Note that some steps may be unconventional compared to other projects, so be sure to pay close attention if this is your first time compiling DFHack.
DFHack uses Git for source control; instructions for installing Git can be found in the platform-specific sections below. The code is hosted on GitHub, and can be downloaded with:
git clone --recursive https://github.com/DFHack/dfhack cd dfhack
If your version of Git does not support the
--recursive flag, you will need
to omit it and run
git submodule update --init after entering the dfhack
This will check out the code on the default branch of the GitHub repo, currently
develop, which may be unstable. If you want code for the latest stable
release, you can check out the
master branch instead:
git checkout master git submodule update
In general, a single DFHack clone is suitable for development - most Git operations such as switching branches can be done on an existing clone. If you find yourself cloning DFHack frequently as part of your development process, or getting stuck on anything else Git-related, feel free to reach out to us for assistance.
If you plan to build DFHack on a machine without an internet connection (or with an unreliable connection), see Building DFHack Offline for additional instructions.
Working with submodules
DFHack uses submodules extensively to manage its subprojects (including the
scripts folder and DF-structures in
library/xml). Failing to keep
submodules in sync when switching between branches can result in build errors
or scripts that don’t work. In general, you should always update submodules
whenever you switch between branches in the main DFHack repo with
git submodule update. (If you are working on bleeding-edge DFHack and
have checked out the master branch of some submodules, running
in those submodules is also an option.)
Rarely, we add or remove submodules. If there are any changes to the existence
of submodules when you switch between branches, you should run
git submodule update --init instead (adding
--init to the above
Some common errors that can arise when failing to update submodules include:
fatal: <some path> does not existwhen performing Git operations
Build errors, particularly referring to structures in the
df::namespace or the
Not a known DF versionwhen starting DF
Run 'git submodule update --init'when running CMake
Submodules are a particularly confusing feature of Git. The Git Book has a thorough explanation of them (as well as of many other aspects of Git) and is a recommended resource if you run into any issues. Other DFHack developers are also able to help with any submodule-related (or Git-related) issues you may encounter.
Before you can compile the code you’ll need to configure your build with cmake. Some IDEs can do this for you, but it’s more common to do it from the command line. Windows developers can refer to the Windows section below for batch files that can be used to avoid opening a terminal/command-prompt.
You should seek cmake’s documentation online or via
cmake --help to see how the command works. See
the Build Options page for help finding the DFHack build options relevant to you.
Before compiling code, you’ll of course need code to compile. This will include the submodules, so be sure you’ve read the section about getting the code.
On Linux, DFHack acts as a library that shadows parts of the SDL API using LD_PRELOAD.
Building is fairly straightforward. Enter the
build folder (or create an
empty folder in the DFHack directory to use instead) and start the build like this:
cd build cmake .. -G Ninja -DCMAKE_BUILD_TYPE:string=Release -DCMAKE_INSTALL_PREFIX=<path to DF> ninja install # or ninja -jX install to specify the number of cores (X) to use
<path to DF> should be a path to a copy of Dwarf Fortress, of the appropriate version for the DFHack you are building. This will build the library along with the normal set of plugins and install them into your DF folder.
Alternatively, you can use ccmake instead of cmake:
cd build ccmake .. -G Ninja ninja install
This will show a curses-based interface that lets you set all of the extra options. You can also use a cmake-friendly IDE like KDevelop 4 or the cmake-gui program.
There are several different batch files in the
subfolders in the
build folder, along with a script that’s used for picking
the DF path. Use the subfolder corresponding to the architecture that you want
to build for.
set_df_path.vbs and point the dialog that pops up at
a suitable DF installation which is of the appropriate version for the DFHack
you are compiling. The result is the creation of the file
the build directory. It contains the full path to the destination directory.
You could therefore also create this file manually - or copy in a pre-prepared
version - if you prefer.
Next, run one of the scripts with
generate prefix. These create the MSVC
allwill create a solution with everything enabled (and the kitchen sink).
guiwill pop up the CMake GUI and let you choose what to build. This is probably what you want most of the time. Set the options you are interested in, then hit configure, then generate. More options can appear after the configure step.
minimalwill create a minimal solution with just the bare necessities - the main library and standard plugins.
releasewill create a solution with everything that should be included in release builds of DFHack. Note that this includes documentation, which requires Python.
Then you can either open the solution with MSVC or use one of the msbuild scripts.
After running the CMake generate script you will have a new folder called VC2022
or VC2022_32, depending on the architecture you specified. Open the file
dfhack.sln inside that folder. If you have multiple versions of Visual
Studio installed, make sure you open with Visual Studio 2022.
The first thing you must then do is change the build type. It defaults to Debug,
but this cannot be used on Windows. Debug is not binary-compatible with DF.
If you try to use a debug build with DF, you’ll only get crashes and for this
reason the Windows “debug” scripts actually do RelWithDebInfo builds.
After loading the Solution, change the Build Type to either
Then build the
INSTALL target listed under
In the build directory you will find several
buildprefix will only build DFHack.
installprefix will build DFHack and install it to the previously selected DF path.
packageprefix will build and create a .zip package of DFHack.
Compiling from the command line is generally the quickest and easiest option. Modern Windows terminal emulators such as Cmder or Windows Terminal provide a better experience by providing more scrollback and larger window sizes.
DFHack functions similarly on macOS and Linux, and the majority of the information above regarding the build process (CMake and Ninja) applies here as well.
DFHack can officially be built on macOS only with GCC 4.8 or 7. Anything newer than 7 will require you to perform extra steps to get DFHack to run (see Notes for GCC 8+ or OS X 10.10+ users), and your build will likely not be redistributable.
Get the DFHack source as per section How to get the code, above.
Set environment variables
Homebrew (if installed elsewhere, replace /usr/local with
export CC=/usr/local/bin/gcc-7 export CXX=/usr/local/bin/g++-7
export CC=/opt/local/bin/gcc-mp-7 export CXX=/opt/local/bin/g++-mp-7
Change the version numbers appropriately if you installed a different version of GCC.
If you are confident that you have GCC in your path, you can omit the absolute paths:
export CC=gcc-7 export CXX=g++-7
(adjust as needed for different GCC installations)
mkdir build-osx cd build-osx cmake .. -G Ninja -DCMAKE_BUILD_TYPE:string=Release -DCMAKE_INSTALL_PREFIX=<path to DF> ninja install # or ninja -jX install to specify the number of cores (X) to use
<path to DF> should be a path to a copy of Dwarf Fortress, of the appropriate version for the DFHack you are building.
If you have issues building on OS X 10.10 (Yosemite) or above, try defining the following environment variable:
If you build with a GCC version newer than 7, DFHack will probably crash
immediately on startup, or soon after. To fix this, you will need to replace
hack/libstdc++.6.dylib with a symlink to the
in your version of GCC:
cd <path to df>/hack && mv libstdc++.6.dylib libstdc++.6.dylib.orig && ln -s [PATH_TO_LIBSTDC++] .
For example, with GCC 6.3.0,
PATH_TO_LIBSTDC++ would be:
/usr/local/Cellar/gcc@6/6.3.0/lib/gcc/6/libstdc++.6.dylib # for 64-bit DFHack /usr/local/Cellar/gcc@6/6.3.0/lib/gcc/6/i386/libstdc++.6.dylib # for 32-bit DFHack
Note: If you build with a version of GCC that requires this, your DFHack
build will not be redistributable. (Even if you copy the
from your GCC version and distribute that too, it will fail on older OS X
versions.) For this reason, if you plan on distributing DFHack, it is highly
recommended to use GCC 4.8 or 7.
Alongside the above, you will need to follow these additional steps to get it running on Apple silicon.
Install an x86 copy of
homebrew alongside your existing one. This
stackoverflow answer describes the
Follow the normal macOS steps to install
gcc via your x86 copy of
homebrew. Note that this will install a GCC version newer than 7, so see
Notes for GCC 8+ or OS X 10.10+ users.
In your terminal, ensure you have your path set to the correct homebrew in
addition to the normal
CXX flags above:
You can use docker to build DFHack for Windows. These instructions were developed on a Linux host system.
On your Linux host, install and run the docker daemon and then run these commands:
xhost +local:root docker run -it --env="DISPLAY" --env="QT_X11_NO_MITSHM=1" --volume=/tmp/.X11-unix:/tmp/.X11-unix --user buildmaster --name dfhack-win ghcr.io/dfhack/build-env:msvc
xhost command and
--env parameters are there so you can eventually
run Dwarf Fortress from the container and have it display on your host.
docker run command above will give you a shell prompt (as the
buildmaster user) in the
container. Inside the container, run the following commands:
git clone https://github.com/DFHack/dfhack.git cd dfhack git submodule update --init cd build dfhack-configure windows 64 Release dfhack-make
dfhack-* scripts there are several commands that set up the wine
server. Each invocation of a Windows tool will cause wine to run in the container.
Preloading the wineserver and telling it not to exit will speed configuration and
compilation up considerably (approx. 10x). You can configure and build DFHack
ninja commands, but your build will go much slower.
First, create a directory in the container to house the Dwarf Fortress binary and assets:
If you can just download Dwarf Fortress directly into the container, then that’s fine. Otherwise, you can do something like this in your host Linux environment to copy an installed version to the container:
cd ~/.steam/steam/steamapps/common/Dwarf\ Fortress/ docker cp . dfhack-win:df/
Back in the container, run the following commands:
cd dfhack/build cmake .. -DCMAKE_INSTALL_PREFIX=/home/buildmaster/df ninja install cd ~/df wine64 "Dwarf Fortress.exe"
Closing your shell will kick you out of the container. Run this command on your Linux host when you want to reattach:
docker start -ai dfhack-win
If you edit code and need to rebuild, run
dfhack-make and then
That will handle all the wineserver management for you.
If you wish, you can use Docker to build just the Windows files to copy to your existing Steam installation on Linux.
dfhack into another directory, and run the build script:
git clone https://github.com/DFHack/dfhack.git cd dfhack git submodule update --init --recursive cd build ./build-win64-from-linux.sh
The script will mount your host’s
dfhack directory to docker, use it to
build the artifacts in
build/win64-cross, and put all the files needed to
If you need to run
sudo, run the script using
rather than directly:
As the script will tell you, you can then copy the files into your DF folder:
# Optional -- remove the old hack directory in case we leave files behind rm ~/.local/share/Steam/steamapps/common/"Dwarf Fortress"/hack cp -r win64-cross/output/* ~/.local/share/Steam/steamapps/common/"Dwarf Fortress"/
Afterward, just run DF as normal.
As of 0.43.05, DFHack downloads several files during the build process, depending on your target OS and architecture. If your build machine’s internet connection is unreliable, or nonexistent, you can download these files in advance.
First, you must locate the files you will need. These can be found in the dfhack-bin repo. Look for the most recent version number before or equal to the DF version which you are building for. For example, suppose “0.43.05” and “0.43.07” are listed. You should choose “0.43.05” if you are building for 0.43.05 or 0.43.06, and “0.43.07” if you are building for 0.43.07 or 0.43.08.
Then, download all of the files you need, and save them to
<path to DFHack
clone>/CMake/downloads/<any filename>. The destination filename you choose
does not matter, as long as the files end up in the
You need to download all of the files for the architecture(s) you are building
for. For example, if you are building for 32-bit Linux and 64-bit Windows,
download all files starting with
win64. GitHub should sort
files alphabetically, so all the files you need should be next to each other.
Any files containing “allegro” in their filename are only necessary for building stonesense. If you are not building Stonesense, you don’t have to download these, as they are larger than any other listed files.
It is recommended that you create a build folder and run CMake to verify that
you have downloaded everything at this point, assuming your download machine has
CMake installed. This involves running a “generate” batch script on Windows, or
a command starting with
cmake .. -G Ninja on Linux and macOS, following the
instructions in the sections above. CMake should automatically locate files that
you placed in
CMake/downloads, and use them instead of attempting to