Dockerfile that fetches existing built releases for multi-platform build

[jadehopepunk]

This is mainly a draft for discussion at this stage. This dockerfile works, and fast, even if it's an unholy abomination. Particular limitations are:

• it only supports two docker platforms (linux/amd64 and linux/arm/v7) although adding more is fairly easy
• It requires the release number to be passed in, or else it defaults to v0.5.0

To use it, you need to use docker buildx. On my linux box, that's done by:

• using Docker Desktop (not Docker Engine, that doesn't work)
• setting up a build engine using the instructions at https://www.docker.com/blog/multi-arch-images/
• running the command below:

docker buildx build --platform linux/amd64,linux/arm/v7 -t redbugcollective/aquadoggo:latest --push .

I've tried pulling and running that image on linux x86, and on a physical Raspberry Pi, and it works.

So, to build this with github actions, we're going to need to set it up using the steps at https://docs.docker.com/build/ci/github-actions/multi-platform/

How are you feeling about this approach?

📋 Checklist

• [ ] Add tests that cover your changes
• [ ] Add this PR to the Unreleased section in CHANGELOG.md
• [ ] Link this PR to any issues it closes
• [ ] New files contain a SPDX license header

[sandreae]

Hi @jadehopepunk, thanks for this PR :+1:

Can you remind me how you ended up with this approach? As I remember, it was because compile times for some targets were getting too long.

I don't hate this approach, positives are that it re-uses existing compiled binaries, and it's quick, but of course it is an unusual beast ;-p

It's been a while since I did any serious docker work, but another approach I can see is to have a two step docker build process:

• compile binary for desired platform on a standard linux base using cargo build targets
• copy the binary into a final docker image with destination platform base image

I would guess this gets around any slow build times you were seeing, our ci build times for different platforms are quite quick: https://github.com/p2panda/aquadoggo/actions/runs/6039734188/usage