Skip to main content

Testing scripts before publishing

load.sh normally downloads a script from shellscript.download and keeps a copy in $HOME/.shellscript/downloads. While you change a script, you want it to run your working copy instead, with the same helpers and environment the loader gives a published script. That is what --developer does.

Run a working copy with --developer​

--developer <path> makes load.sh run <path>/<script>.sh from a local folder. Nothing is downloaded, and the cached copy in $HOME/.shellscript/downloads is neither used nor changed.

# From a clone of the repository: run maven.sh from the working tree
load.sh --developer ./public/scripts maven

# Options of the script go after "--", as usual
load.sh --developer ./public/scripts maven -- --dry-run --version 3.9.6

# Check that the script is found and prepared, without running it
load.sh --developer ./public/scripts --dont-run maven

load.sh errors if <path>/<script>.sh does not exist. A script that shares a file from public/scripts/lib/ reads it from the local folder too: the post-load hook that downloads it is not called in this mode.

The same works from another project. A test suite can be pointed at your working copy before anything is published, for example:

load.sh --developer ~/src/shellscript-download/public/scripts qemu -- start --image ubuntu-24.04 --bridge virbr0

Test a change to load.sh itself​

The loader you have installed is $HOME/.shellscript/bin/load.sh, and it only changes when you reinstall it. To try a change to the loader, run the copy in the repository, with --developer so that the scripts come from the working tree as well:

bash ./public/scripts/load.sh --developer ./public/scripts qemu -- list

Capture a script's output​

The loader writes its own messages to stderr, and stdout carries only what the script prints, so a script's output can be captured:

ip=$(load.sh qemu -- address node1)

Loaders installed before this change print their banner on stdout. A caller that has to work with them keeps only the lines it expects.

Before opening a pull request​

bash -n public/scripts/<name>.sh     # syntax
npm run build                        # regenerates the script pages, docs/scripts and completion

Then test the script in a clean environment.

Test in a clean environment​

Test scripts in a clean, non-root Linux before submitting them. The simplest is a Docker container with the working tree mounted:

docker run -it --rm \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v "$(pwd)/public/scripts":/scripts:ro \
  -v "$(pwd)/public/install":/install:ro \
  ubuntu:24.04 bash

Inside the container, create an unprivileged user, since scripts refuse to run as root:

apt update && apt install -y curl sudo

useradd -m -s /bin/bash user
echo 'user ALL=(ALL) NOPASSWD:ALL' >> /etc/sudoers

su - user

Then install the loader from the mounted folder and run your script, with no network needed for either:

bash /install/loader --developer

load.sh --developer /scripts <name> -- --dry-run