Nix is the easiest way to build from source, it should be on top.
9.5 KiB
Building the Roc compiler from source
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.
If you are running ArchLinux or a derivative like Manjaro, you'll need to run sudo sysctl -w kernel.unprivileged_userns_clone=1
before installing nix.
Install nix:
curl -L https://nixos.org/nix/install | sh
You will need to start a fresh terminal session to use nix.
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.
Use cargo run help
to see all subcommands.
To use the repl
subcommand, execute cargo run repl
.
Use cargo build
to build the whole project.
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 dependencies into your current shell, so you never have to run nix-shell directly!
Editor
The editor is a WIP and not ready yet to replace your favorite editor, although if you want to try it out on nix, read on.
cargo run edit
should work from NixOS, if you use a nix-shell from inside another OS, follow the instructions below.
Nvidia GPU
Outside of a nix shell, execute the following:
nix-channel --add https://github.com/guibou/nixGL/archive/main.tar.gz nixgl && nix-channel --update
nix-env -iA nixgl.auto.nixVulkanNvidia
Running the editor does not work with nix-shell --pure
.
nix-shell
460.91.03 may be different for you, type nixVulkanNvidia and press tab to autocomplete for your version.
nixVulkanNvidia-460.91.03 cargo run edit
Integrated Intel Graphics
❗ ** Our Nix setup currently cannot run the editor with integrated intel graphics, see #1856 ** ❗
Outside of a nix shell, run:
git clone https://github.com/guibou/nixGL
cd nixGL
nix-env -f ./ -iA nixVulkanIntel
cd to the roc repo, and run (without --pure):
nix-shell
nixVulkanIntel cargo run edit
Other configs
Check the nixGL repo for other graphics configurations.
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!
Manual Install
To build the compiler, you need these installed:
- Zig, see below for version
libxkbcommon
- macOS seems to have it already; on Ubuntu or Debian you can get it withapt-get install libxkbcommon-dev
- On Debian/Ubuntu
sudo apt-get install pkg-config
- 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 usecargo test --no-fail-fast
orcargo 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.
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
$(brew --prefix llvm@12)/bin
to your PATH
. You can confirm this worked by
running llc --version
- it should mention "LLVM version 12.0.0" at the top.
You may also need to manually specify a prefix env var like so:
export LLVM_SYS_120_PREFIX=/usr/local/opt/llvm@12
For Ubuntu and Debian:
sudo apt -y install lsb-release software-properties-common gnupg
wget https://apt.llvm.org/llvm.sh
chmod +x llvm.sh
./llvm.sh 12
If you use this script, you'll need to add clang
and llvm-as
to your PATH
.
By default, the script installs them as clang-12
and llvm-as-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
Building
Use cargo build
to build the whole project.
Use cargo run help
to see all subcommands.
To use the repl
subcommand, execute cargo run repl
.
LLVM installation on Linux
For a current list of all dependency versions and their names in apt, see the Earthfile.
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:
- 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)
- 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" [note: as of September 2021 this should no longer be necessary - the next time anyone tries this, please try it without this step and make a PR to delete this step if it's no longer needed!]
- I launched the "x64 Native Tools Command Prompt for Visual Studio 2019" application (note: not the similarly-named "x86" one!)
- Make sure Python 2.7 and CMake 3.17 are installed on your system.
- 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. - I ran
cmake -G "NMake Makefiles" -DCMAKE_BUILD_TYPE=Release ../llvm
to generate a NMake makefile. - Once that completed, I ran
nmake
to build LLVM. (This took about 2 hours on my laptop.) - Finally, I set an environment variable
LLVM_SYS_100_PREFIX
to point to thebuild
directory where I ran thecmake
command.
Once all that was done, cargo
ran successfully for Roc!
Build speed on WSL/WSL2
If your Roc project folder is in the Windows filesystem but you're compiling from Linux, rebuilds may be as much as 20x slower than they should be! Disk access during linking seems to be the bottleneck. It's recommended to move your folder to the Linux filesystem.
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 12 (e.g. with $ sudo apt-get install lld-12
)
and add make sure there's a ld.lld
executable on your PATH
which
is symlinked to lld-12
.
That's it! Enjoy the faster builds.