In CI and our development environment we do not rely on the Go environment of
the host machine, and instead use Go installed in a container. To simplify
things we separate all of Boulder's build dependencies into its own
boulder-tools Docker image.
To build boulder-tools images, you'll need a Docker set up to do cross-platform builds (we build for both amd64 and arm64 so developers with Apple silicon can use boulder-tools in their dev environment). On Ubuntu the setup steps are:
docker buildx create --use --name=cross
sudo sudo apt-get install qemu binfmt-support qemu-user-static
After setup, the output of docker buildx ls should contain an entry like:
cross0 unix:///var/run/docker.sock running linux/amd64, linux/386, linux/arm64, linux/riscv64, linux/ppc64le, linux/s390x, linux/mips64le, linux/mips64, linux/arm/v7, linux/arm/v6
If you see an entry like:
cross0 unix:///var/run/docker.sock stopped
That's probably fine; the instance will be started when you run
tag_and_upload.sh (which runs docker buildx build).
Rather than install multiple versions of Go within the same boulder-tools
container we maintain separate images for each Go version we support.
When a new Go version is available we perform several steps to integrate it to our workflow:
- We add it to the
GO_VERSIONSarray intag_and_upload.sh. - We run the
tag_and_upload.shscript to build, tag, and upload aboulder-toolsimage for each of theGO_VERSIONS - We update
.github/workflows/boulder-ci.yml, adding the new docker image tag(s) to theBOULDER_TOOLS_TAGsection.
After some time when we have spot checked the new Go release and coordinated
a staging/prod environment upgrade with the operations team we can remove the
old GO_VERSIONS entries, delete their respective build matrix items, and update
docker-compose.yml.