Source Code Management with Git · 2009-12-29 15:22 by Black in Programming

Git is a distributed SCM designed by Linus Torvalds to manage the development of the Linux Kernel. Since it’s licensed under the GPL, it can be used freely by anyone.

Just like backups are a necessity for anyone who uses a Computer (or should be…), source code management is a necessity for serious developers. Not only does it track the past state of the project (which allows tracking the introduction of bugs), but it also allows the management of separate branches. That way, development can continue to add new experimental features while production uses only stable and tested code.

Git is a distributed SCM tool, unlike CVS and Subversion it does not require a central server and by design there is no central authoritative repository. Every repository contains the full history. Every file is hashed and added to a database. Every commit contains a tree of file hashes, a commit message and a pointer to the ancestor commits. All that is hashed and added to the database, so a commit’s hash can be used to cryptographically verify the integrity of the complete previous history. For a more technical perspective on git’s inner workings, read Git for Computer Scientists (It really is quite cool in it’s simplicity). Here’s a one sided comparison of git with some alternatives.

I have started to use git beginning of 2008 for my work on ExaminationRoom, and while the start was a bit hairy, having a history of my code development as well as my comments have helped me a lot, even as only developer. I worked on three computers, so keeping the code synchronized was critical. That too was easy thanks to the SCM, even without a reachable central server (One of the computers had no internet access, it was only used to drive two Projectors for the experiments.)

I still use git these days, and can’t recommend it more. Although most other projects are World of Warcraft addons… All my public code can be cloned from my repositories