Awesome

Binpot

The statically compiled, cross architecture, Docker based, binaries pot.

Blog post

Usage

The usage is focused to build other Docker images.

For example:

FROM alpine:3.14
COPY --from=qmcgaw/binpot:helm /bin /usr/local/bin/helm

🔍 Search for all image tags 💡

• The Docker image tags for qmcgaw/binpot follow the following formatting:
  • :name for the latest stable version of the program name
  • :name-v0.0.0 for the semver version of the program name
• Each Docker image is based on scratch and only contain the binary program at the path /bin.
• Each Docker image and binary program is built for each of the possible CPU architecture supported by Docker, unless indicated otherwise.
• Each binary has permissions 500 (read and execute for the user owner)
• You can change the ownership in the COPY command using --chown=1000 for example, without duplicating the binary in your Docker image layers.
• Each Docker image has an entrypoint [ "/bin" ] so you can run it as well

Need help? ▶️ Create a discussion

Thinking of copying the binary for your host?

Programs available

ℹ️ all architectures means: linux/amd64, linux/386, linux/arm64, linux/arm/v7, linux/arm/v6, linux/ppc64le, linux/s390x, linux/riscv64

Want more!? ▶️ Create an issue!

How it works

1. For each program, a Dockerfile describes how to build it. The final binary is placed on a final scratch based Docker image. Example: helm has ./dockerfiles/helm/Dockerfile
2. For each program, a Github Action workflow is triggered when its Dockerfile or the workflow itself is changed. This workflow takes care of:
   1. Cross build the program for all CPU architectures
      • If one architecture is not supported such as for dlv, build the unavailable program
   2. Pushing the images containing the program to Docker Hub

Note on dlv

💁 Concerning dlv: all images are built for all architectures even if the program does not support all of them. A substitute Go program printing dlv v1.7.0 is unavailable on <platform name> and exiting with exit code 1 is used for unsupported platforms. This is like so so you can still cross build with all the architectures, especially if the program is an optional dependency. This is often the case for VSCode development containers for instance. In this case, if you try to build for arm/v7 and need dlv as an optional dependency, your COPY --from=qmcgaw/binpot:dlv will not fail.

Copy the binary on your host

If you want to use the binary directly on your host, you can do it with Docker. This has the advantage that it will automatically get the right binary for your host platform.

In the following we want to install helm on our host.

For example:

export PROGRAM="helm" && \
  docker pull "qmcgaw/binpot:$PROGRAM" && \
  containerid="$(docker create qmcgaw/binpot:$PROGRAM)" && \
  docker cp "$containerid:/bin" "/usr/local/bin/$PROGRAM" && \
  docker rm "$containerid"

Test Helm works with helm

TODOs

• Change version of go-outline once a release tag is made: Github issue