Today I faced one of those "this is Witchcraft" moments.

CI was failing for a specific commit. I checked out that exact commit locally… everything clean. Built fine. No errors.

Witchcraft 🧙! "I'm exactly on the same commit, and in my local machine it does build with no problems".

This was a JS project. And in JS (TS doesn’t matter here), my first suspect is always dependencies.
So, I removed node_modules and rebuilt from scratch.

Witchcraft x2 🧙🧙! still couldn’t reproduce the error.

So I went to the workflow file and started following step by step what CI was doing.

Witchcraft x3 🧙🧙🧙! it can't b... oh wait.

Of course the CI runs everything in a clean, pure node:20-alpine image.

So I fired up the same container locally to see what was going on.

Now, this is just he context to what i want to share:
I use this tiny script like...a lot.

If you use oh-my-zsh, you can add it to ~/.oh-my-zsh/custom/docker-run.zsh (adapt to your setup):

docker-run.zsh

function docker-run() {
    if [[ -z "$1" ]]; then
        echo "Usage: docker-run <image>"
        return 1
    fi

    local container_image="$1"
    local workdir="/app"

    docker run --rm -it \
        -v "$PWD:$workdir" \
        -w "$workdir" \
        "$container_image" \
        /bin/sh
}

Super simple. Just go to your project folder:

$» docker-run node:20-alpine
/app #

The devil’s in the details:

• --rm so the container is gone when you exit
• $PWD mounts the current project inside the container
• it so you’re dropped into the terminal
• sh cause these minimal images usually don’t have bash

Back to the mysterious "Witchcraft!🧙 error" —>
Turned out there was a node version mismatch: v20.18.1 locally vs. v20.19.2 in the container.

No magic for today.

--
thanks for reading.