roc/BUILDING_FROM_SOURCE.md

Building the Roc compiler from source

Installing LLVM, Python, Zig, valgrind, libunwind, and libc++-dev

To build the compiler, you need these installed:

• libunwind (macOS should already have this one installed)
• libc++-dev
• Python 2.7 (Windows only), python-is-python3 (Ubuntu)
• Zig, see below for version
• LLVM, see below for version

To run the test suite (via cargo test), you additionally need to install:

• valgrind (needs special treatment to install on macOS Alternatively, you can use cargo test --no-fail-fast or cargo test -p specific_tests to skip over the valgrind failures & tests.

For debugging LLVM IR, we use DebugIR. This dependency is only required to build with the --debug flag, and for normal developtment you should be fine without it.

libunwind & libc++-dev

MacOS systems should already have libunwind, but other systems will need to install it (On Ubuntu, this can be done with sudo apt-get install libunwind-dev). Some systems may already have libc++-dev on them, but if not, you may need to install it. (On Ubuntu, this can be done with sudo apt-get install libc++-dev.)

libcxb libraries

You may see an error like this during builds:

/usr/bin/ld: cannot find -lxcb-render
/usr/bin/ld: cannot find -lxcb-shape
/usr/bin/ld: cannot find -lxcb-xfixes

If so, you can fix it like so:

sudo apt-get install libxcb-render0-dev libxcb-shape0-dev libxcb-xfixes0-dev

Zig

version: 0.8.0

For any OS, you can use zigup to manage zig installations.

If you prefer a package manager, you can try the following:

• For MacOS, you can install with brew install zig
• For, Ubuntu, you can use Snap, you can install with snap install zig --classic --beta
• For other systems, checkout this page

If you want to install it manually, you can also download Zig directly here. Just make sure you download the right version, the bleeding edge master build is the first download link on this page.

LLVM

version: 12.0.x

For macOS, you can install LLVM 12 using brew install llvm@12 and then adding /usr/local/opt/llvm/bin to your PATH. You can confirm this worked by running llc --version - it should mention "LLVM version 12.0.0" at the top.

For Ubuntu and Debian, you can use the Automatic installation script at apt.llvm.org:

sudo bash -c "$(wget -O - https://apt.llvm.org/llvm.sh)"

If you use this script, you'll need to add clang and llvm-as to your PATH. By default, the script installs them as llvm-as-12 and clang-12, respectively. You can address this with symlinks like so:

sudo ln -s /usr/bin/clang-12 /usr/bin/clang

sudo ln -s /usr/bin/llvm-as-12 /usr/bin/llvm-as

There are also alternative installation options at http://releases.llvm.org/download.html

Troubleshooting

Using Nix

Install

Using nix is a quick way to get an environment bootstrapped with a single command.

Anyone having trouble installing the proper version of LLVM themselves might also prefer this method.

First, install nix:

curl -L https://nixos.org/nix/install | sh

If MacOS and using a version >= 10.15:

sh <(curl -L https://nixos.org/nix/install) --darwin-use-unencrypted-nix-store-volume

You may prefer to setup up the volume manually by following nix documentation.

You may need to restart your terminal

Usage

Now with nix installed you just need to run one command:

nix-shell

This may not output anything for a little while. This is normal, hang in there. Also make sure you are in the roc project root.

Also, if you're on NixOS you'll need to enable opengl at the system-wide level. You can do this in configuration.nix with hardware.opengl.enable = true;. If you don't do this, nix-shell will fail!

You should be in a shell with everything needed to build already installed. Next run:

cargo run repl

You should be in a repl now. Have fun!

Extra tips

If you plan on using nix-shell regularly, check out direnv and lorri. Whenever you cd into roc/, they will automatically load the Nix dependecies into your current shell, so you never have to run nix-shell directly!

Editor

When you want to run the editor from Ubuntu inside nix you need to install nixGL as well:

nix-shell
git clone https://github.com/guibou/nixGL
cd nixGL

If you have an Nvidia graphics card, run:

nix-env -f ./ -iA nixVulkanNvidia

If you have integrated Intel graphics, run:

nix-env -f ./ -iA nixVulkanIntel

Check the nixGL repo for other configurations.

Now you should be able to run the editor:

cd roc
nixVulkanNvidia cargo run edit `# replace Nvidia with the config you chose in the previous step`

Troubleshooting

Create an issue if you run into problems not listed here. That will help us improve this document for everyone who reads it in the future!

LLVM installation on Linux

On some Linux systems we've seen the error "failed to run custom build command for x11". On Ubuntu, running sudo apt install pkg-config cmake libx11-dev fixed this.

If you encounter cannot find -lz run sudo apt install zlib1g-dev.

If you encounter:

error: No suitable version of LLVM was found system-wide or pointed
       to by LLVM_SYS_120_PREFIX.

Add export LLVM_SYS_120_PREFIX=/usr/lib/llvm-12 to your ~/.bashrc or equivalent file for your shell.

LLVM installation on macOS

If installing LLVM fails, it might help to run sudo xcode-select -r before installing again.

It might also be useful to add these exports to your shell:

export LDFLAGS="-L/usr/local/opt/llvm/lib -Wl,-rpath,/usr/local/opt/llvm/lib"
export CPPFLAGS="-I/usr/local/opt/llvm/include"

LLVM installation on Windows

Installing LLVM's prebuilt binaries doesn't seem to be enough for the llvm-sys crate that Roc depends on, so I had to build LLVM from source on Windows. After lots of help from @IanMacKenzie (thank you, Ian!), here's what worked for me:

1. I downloaded and installed Build Tools for Visual Studio 2019 (a full Visual Studio install should work tool; the Build Tools are just the CLI tools, which is all I wanted)
2. In the installation configuration, under "additional components" I had to check both "C++ ATL for latest v142 build tools (x86 & x64)" and also "C++/CLI support for v142 build tools"
3. I launched the "x64 Native Tools Command Prompt for Visual Studio 2019" application (note: not the similarly-named "x86" one!)
4. Make sure Python 2.7 and CMake 3.17 are installed on your system.
5. I followed most of the steps under LLVM's building from source instructions up to the cmake -G ... command, which didn't work for me. Instead, at that point I did the following step.
6. I ran cmake -G "NMake Makefiles" -DCMAKE_BUILD_TYPE=Release ../llvm to generate a NMake makefile.
7. Once that completed, I ran nmake to build LLVM. (This took about 2 hours on my laptop.)
8. Finally, I set an environment variable LLVM_SYS_100_PREFIX to point to the build directory where I ran the cmake command.

Once all that was done, cargo ran successfully for Roc!

Use LLD for the linker

Using lld for Rust's linker makes build times a lot faster, and I highly recommend it.

Create ~/.cargo/config.toml if it does not exist and add this to it:

[build]
# Link with lld, per https://github.com/rust-lang/rust/issues/39915#issuecomment-538049306
# Use target-cpu=native, per https://deterministic.space/high-performance-rust.html
rustflags = ["-C", "link-arg=-fuse-ld=lld", "-C", "target-cpu=native"]

Then install lld version 9 (e.g. with $ sudo apt-get install lld-9) and add make sure there's a ld.lld executable on your PATH which is symlinked to lld-9.

That's it! Enjoy the faster builds.