HomeworldSDL/Linux/BUILD.md

Compiling Homeworld SDL on Linux

Installing dependencies

Using Nix

If you have Nix installed, there is a flake.nix file listing the build depencies so you can just run the build in a nix develop environment without installing anything:

nix develop ./Linux

You can then go on with the Quick Start in that virtual environment.

Quick Start

More information can be found in the documentation files next to this one.

Meson

meson setup build
cd build
meson compile

You are free to replace build above with anything you like. It will be the name of the build directory

You can now run the compiled executable for the first time.

Autotools (Deprecated)

x86_64 (intel/amd 64-bit)

cd Linux
./bootstrap
../configure
make -j4

The -j4 flag passed to make is just an example. -j controls the number of "jobs" used by make to compile the sources. If your machine is equipped with, e.g., 12 cores, then replace -j4 by -j12 for a faster build.

The configuration step has a lot of flags, run ../configure --help to see them. (Notably the --disable-linux-fixme flag)

x86 (intel/amd 32-bit)

This is if you want to cross-compile the game to 32bit even if your machine is 64bit.

The process is the same as for x86_64, except for the ../configure step, as follows:

cd Linux
./bootstrap
CFLAGS='-m32' ../configure --disable-x86_64  # here
make

Note: this will output a binary without debug symbols. Building a 32b binary with debug symbols on a 64b machine is feasible (I have done it for debugging), but not supported by autoconf and therefore not very straightforward.

Hacking

So you want to dive into the code and start hacking, huh? Here are a few pointers to help you with that:

Clangd

Clangd is a language server that can work with many editors (including VSCode) via a plugin. It adds smart features to your editor: code completion, compile errors, go-to-definition and more.

To give proper hints, though, clangd needs to know the compile flags used (otherwise you'll get "header not found" errors). To that end, it uses a compile_commands.json file describing how each file was compiled.

With Meson

Meson automatically generates compile_commands.json, so if you named your build dir build as clangd expects, then you have nothing to do. Enjoy your modern development environment!

With Autotools (Deprecated)

You can use bear to auto-generate compile_commands.json. In the build steps outlined above, replace the make step with:

bear -- make -j4

⚠️ As of 2023-09-22, this breaks the build if you also used the --enable-sanitizers option in the configure step. You will have to run a first build without this option, then re-enable it, and re-build once the compile_commands.json has been generated.

Then link or copy the result to the root of the repo, so that clangd finds it automatically

ln -srv compile_commands.json ..

Sanitizers

LLVM's Sanitizers are a powerful suite of tools for memory debugging. They can detect and help debug many kinds of invalid or dangerous memory handling patterns (like buffer overflows, use after free, or leaks).

With Meson

The address and undefined sanitizers are enabled by default. You can disable them by passing the -Db_sanitize=none option to meson setup.

With Autotools (Deprecated)

You can build a debug version of the game that includes those sanitizers with

../configure --enable-sanitizers

Cross-compiling to wasm32-emscripten

You need to have emscripten installed and enable your installed emsdk tools in your current shell environment. (If you're using the nix shell, you can skip this step)

source emsdk_env.sh

Now you can setup meson for cross-compiling to wasm32 using emscripten:

meson setup --cross-file ../wasm/wasm32-emscripten.meson-cross-build-definition.txt -Db_sanitize=none -Dmovies=false build.emscripten ..

Now switch to the created build.emscripten folder and compile:

cd build.emscripten
meson compile

To automatically open the compiled wasm32 binaries in the browser using the provided index.html one can use the mini webserver provided with emscripten:

emrun .

Building the Docker image used in the CI

docker load < $(nix build --no-link --print-out-paths .#ci-docker-image)