How to stop yourself from force-pushing to `master`

posted on August 30, 12019 HE

If you’re like me, you probably use git, and if you’re like me, you probably use git badly. Years of working on projects by myself have ingrained in me a few habits that don’t transfer well to team environments – perhaps most notably, pushing (even force-pushing) to the master branch.

This post documents the steps I’ve taken to catch myself before this happens.

Git hooks, briefly

A git hook is a program that can be run by git at various points in your git workflow. Typical examples include pre-commit (run before making a commit) and post-checkout (run after switching branches).

I’m not bothered by making bad commits – in fact, I often do this on purpose to rebase later. What I’m trying to do is prevent pushing these bad commits, so I make a pre-push hook.

Complete program text1

# don't allow --force-pushing to master branch

hook_name="hooks/$(basename $0)"
cur_branch=$(git name-rev --name-only --no-undefined --always HEAD)
push_cmd=$(ps --pid $PPID --format "command=")


# putting regexes in quotes makes them fail, because bash ¯\_(ツ)_/¯
if [[ "$cur_branch" =~ $protected_branches ]]; then
    if [[ "$push_cmd" =~ $forceful_flags ]]; then
        echo -e "${hook_name}: don't force-push to $cur_branch"
        exit 1
        echo -ne "${hook_name}: are you aware that you are on branch ${cur_branch}? "
        read confirmation < /dev/tty
        if [[ ! "$confirmation" =~ $affirmative ]]; then
            exit 2
exit 0

You can find the latest version of this file in my dotfiles repository.


The code is pretty short and straightforward, but there are a few things worth explaining:

Getting the current branch name

git name-rev exists to make getting the symbolic names of branches easy.

Reading from stdin

Git hooks are not intended to run interactively. This is a problem if you are trying to write a confirmation (“are you sure?”) program.

To circumvent this, read directly from /dev/tty.

Checking the command-line arguments

pre-push will be forked from the git command that you run. With this in mind, we can pass the parent process ID to ps and it will output the command that was run.2

Exit status

If a hook exits with a non-zero exit status, git won’t follow through with the operation. I exploit this by exiting with 1 when we don’t want to push, and 0 when we do.

To set up

To apply this hook globally, to all current and future repos:

> cd ~
> mkdir -p .config/global_git_hooks/
> git config --global core.hooksPath .config/global_git_hooks

You can name your hooks directory whatever you want; here, I chose the name .config/global_git_hooks/. Save the file above as pre-push in the appropriate directory.

Ensure the file is executable:

> chmod +x ~/.config/global_git_hooks/pre-push

That’s all it takes – the program will be run every time invoke git push.

Restoring the default behaviour

Sometimes, you might want the default behaviour. No problem! cd into the repository and edit the local config with git config core.hooksPath $GIT_DIR/hooks. This will override your global and allow custom settings on a repo-per-repo basis.

What if I know what I’m doing and I really want to anyway?

You can skip the execution of hooks with git push --no-verify. Best of luck with that.

  1. I don’t always use bash, but the regex match operator (=~) was too convenient to refuse.↩︎

  2. My first instinct was to read from /proc, but this turned out to be more difficult than running ps.↩︎

More from around the web:

rustc_codegen_gcc: Progress Report #4
September 29

What is rustc_codegen_gcc? rustc_codegen_gcc is a GCC codegen for rustc, meaning that it can be loaded by the existing rustc frontend, but benefits from GCC by having more architectures supported and having access to GCC’s optimizations.

via Antoyo's Blog
Case Study: Using CRDTs to improve a distributed system
September 11

At a previous job, I was lucky enough to be faced with the challenge of improving a distributed service’s robustness to network faults. In this post, I share how I attacked the problem using Conflict-free/Convergent Replicated Data Types (CRDTs). A Chronic…

via Richard Kallos: Richard Kallos
Making Reasonable Use of Computer Resources: Part 2
August 28

In my last post I made the case that programmers should try and write their software in a way that makes reasonable use of the resources of the computer. By reasonable use, I didn’t mean to try and perform every crazy micro-optimization possible to extrac…

via Occasionally sane

generated by openring