For the past few weeks, I have been focusing most of my time at work on a project, which we have coined the . In its essence, it is a data gateway built on top of a Raspberry Pi, which bridges machine and energy metrics from existing infrastructure to our data pipeline. spark As our main focus was set on high frequency read-outs at up to 100Hz, it was quickly sorted out that python, usually my go-to language, wasn’t the way to go. After some research, I quickly settled on Rust due to its cutting edge performance and ‘ mantra. if i compiles, it’s safe’ But as always, everything comes with a price: While Rust is blazingly fast and enables us to develop and ship new features and fixes with confidence, it needs to be compiled for a particular architecture — in our case armv7. Working on a MacbookPro, I initially tackled this problem by pulling the entire repo and building the application on the Raspberry Pi itself, which resulted in up to 30 minutes of impatience. In search of a more streamlined build process, I ended up implementing a cross-compilation pipeline built on top of , which I would like to share with you. Docker What’s In It For You? This post is designed to be a step-by-step manual for you to build your own automated cross-compilation pipeline: First, we will set up a small Rust program, which we will later cross-compile for armv7. Second, we will create a custom Docker image, which will handle the heavy lifting of cross-compiling system dependencies as well as our application. Finally, we will write a shell script to execute the cross-compilation process. Now that we’ve set our scope, let’s get ready for take off! Preparing our Rust Application As this post mainly focuses on automating the cross-compilation of Rust applications, we will set up a bare scaffold and add a system dependency, which I found particularly tricky: OpenSSL. We start by creating a new directory named and ing into it. Then, we initialize a new cargo project by running in our terminal of choice. Now, we should have a project directory called with the following folder structure: hello-rpi cd cargo new --bin hello-rpi hello-rpi hello-rpi||--- hello-rpi* |--- Cargo.toml||--- src* |--- main.rs* With our minimal application generated, add as a dependency in our file and declare as an external crate in . After our small adjustments, our and should something like this: openssl = “0.10.5” Cargo.toml openssl main.rs Cargo.toml main.rs Cargo.toml [package]name = "hello-rpi"version = "0.1.0"authors = ["your handle < om>"] your@mail.c [dependencies]openssl = "0.10.5" main.rs extern crate openssl; fn main() {println!("Hello, world!");} Finally, go ahead and add the following lines to your crate’s cargo config file found in : hello-rpi/hello-rpi/.cargo/config [target.armv7-unknown-linux-gnueabihf]linker = "arm-linux-gnueabihf-gcc" This will tell cargo which linker to use, when we specify as our target, which is our Pi’s architecture. With these few lines of code, we’re done on the Rust end. armv7-unkown-linux-gnueabih Creating our Custom Docker Image With our minimal application set and ready to go, we now focus on automating the cross compilation process of our application. To this end, we will create a Docker image, which will handle cross-compiling our application as well as our dependency, i.e. OpenSSL for us. Fortunately, we don’t have to start from scratch — Stephen Thirlwall’s Docker image , which may be found , provides an awesome boilerplate to build up on. When run without any arguments, a container based on this image returns a helper script, which in turn let’s you interact with the container and start the actual cross-compilation. raspberry-pi-cross-compiler here So let’s dive right in! Create a new folder in our project’s (not crate!) root named , create a and paste the following into it: rpxc Dockerfile Wow, there’s a lot going on here — let’s break it down a bit: First, we point our base image to sdthirwall’s and install some packages via — nothing out of the ordinary. raspberry-pi-cross-compiler apt-get We then use the expression to set environment variables pointing to our and installations as well as our OpenSSL and directories. ENV Cargo Rustup lib include The subsequent block is borrowed from the image and deals with installing the Rust toolchain and adding as a compilation target. As indicated by its name, this is the compiler for our Raspberry’s architecture. RUN rust:1.26.0 armv7-unknown-linux-gnueabihf The second command then downloads, un-tars and cross-compiles OpenSSL for our target architecture. Note that our environment variables pointing to OpenSSL’s and directory match up with our install’s path. RUN lib include Not that complicated, right? What’s more, this setup facilitates including any additional system dependency in no time by simply extending our — it doesn’t get easier than that. Dockerfile Automating The Build Process With our set up, we’re meremly one step short of our build pipeline. More precisely, we want to Dockerfile Build our Docker image we just defined Start a container to obtain the cross-compilation script I mentioned above Use the script to start the cross-compilation process Transforming this procedure into a shell script and adding some clean up yields six actual lines of code, which need to be placed in the directory to work correctly. rpxc And that’s it! We have defined custom Docker image, which handles cross-compiling system dependencies at build time and written a shell script to actually build and cross-compile the entire application. Now, you can simply into , run the script above, sit back and watch the magic happening. cd rpxc Summary In this post, we built a build pipeline, which let’s us cross-compile any Rust application for the armv7 architecture by simply executing a single shell script. To this end, we built a custom , added application specific system dependencies and wrote a small shell script, which kicks off the cross-compilation process. Dockerfile This setup is favourable for multiple reasons: First and foremost, you can build your application on any machine and platform, as long as Docker is installed. Thus, you can easily integrate the entire build process as the final step of your continuous integration pipeline. What’s more, you can add any additional system dependency by extending the , which makes this solution scalable towards more complex applications, which may depend on a large number of system dependencies. Dockerfile That’s it from my side! I hope you’ve enjoyed my first post and that I was able to get you kick started with cross-compiling Rust for Pis. If you have any questions or critique regarding my post, let me know in the comments below! Raspberry I am one of the co-founders of Enlyze. We are a startup based in Aachen, Germany, which treats power consumption as a universal indicator for any appliance’s state of health. Our team of six consists of passionate AI researchers and electrial as well as software engineers. If you’d like to get in touch, shoot us a mail to hello@enlyze.com. Full disclosure:
