uncommitted-changes-sync

Use case

You are working on a git code repository that you want to run on a server. However, you would like to work on the files locally. Reasons for this could be that

• you don't have root access on the server and therefore can't install the helper tools that you need
• a mounted remote filesystem can be too slow

So you have the repository locally and remotely and need a way to quickly sync the two. Of course you could work locally, then commit, push and then pull on the server, but it is inconvenient. In particular, you don't want to create a commit yet, because you're still debugging. That's where uncommitted-changes-sync comes in! It gives you a way to sync your uncommited changes directly to a remote filesystem.

CAUTION

This will DELETE all untracked files and folders in your remote repository every time it syncs (unless they are in your .gitignore).

Installation

This only works on Linux. You need the program inotifywait installed locally. On debian/ubuntu, it can be found in the package inotify-tools:

sudo apt install inotify-tools

Then download this repository to your local computer.

How it works

The main functionality can be found in git_upd. Every time a local change to a file in one of the tracked repositories is detected, the following happens:

• Locally, create a git diff between the current state of the repository and the branch to compare with (typically origin/master)
• Send the diff via ssh to the server
• On the server, perform a git reset --hard to the comparison branch
• On the server, apply the diff

Usage

Create a config file and place it somewhere on your computer (on unix, the recommended path would be something like ~/.config/uncommitted_changes_sync, but anywhere is fine). It should have one line per repository that you want to sync. In each line, you put first the path to the local repository, then the comparison branch, then the ssh-name of the remote host, and then the remote path to the repository, all separated by spaces. Example:

/home/local_username/myrepo origin/main mycluster /home/remote_username/myrepo

Lines starting with "#" or empty lines are ignored.

If you haven't done so already, you should add your ssh-key to the ssh-daemon with

ssh-add

Then, run the sync command (adjust the paths):

~/repos/uncommitted_changes_sync/sync ~/.config/uncommitted_changes_sync

Gotchas

• You cannot use $HOME in your remote path
• Assuming your comparison branch is origin/master, then you need to make sure that origin/master points to the same commit locally and remotely. Note that you can make commits locally, even to master, as long as you don't push the commits to origin/master without pulling on the remote
• Changes in git submodules are not synced. However, you can sync each submodule individually by including them as extra repos in the config file
• If your diff is empty, the output from the server will be unrecognized input. Nothing bad happens, but it can be confusing.

Run in the background with systemd

systemd can run this script for you in the background. Once started, it will continue working until reboot. systemd also takes care of the logs for you, including time-stamps.

To run this in the background with systemd as a user unit (i.e. not requiring root permissions):

• Place the provided file systemd_unit.service into ~/.config/systemd/user/ and give it a more reasonable name like uncommitted-changes-sync.service. In the file, adapt the path as required.
• Run systemctl --user restart uncommitted-changes-sync.service

If you did not run ssh-add since booting, you have to run it first before then (re-)starting the systemd unit. This unfortunately also means that configuring the unit such that it will get started automatically on boot doesn't have the desired effect, as syncing will not work.