How to GitHub: Fork, Branch, Track, Squash and Pull Request

by Rich Jones
Mar 27, 2012 This guide will teach you how to properly contribute to open source projects on GitHub. It assumes that you already know about how to use Git for version control and that you already have a GitHub account. Psstt.. if you already have a GitHub account and you want to earn more money , sign up for Gun.io with your GitHub profile and we'll pair you with people looking for developers based on your portfolio!

Getting Started

GitHub displays these instructions when you start a new project.

GitHub is pretty great about giving advice to users starting new repositories, but it isn't very helpful when it comes to contributing changes back to other projects. Hopefully, this guide will help. Before you get started, find the page of the project you're looking to improve. Poke around in the code a little bit, familiarize yourself with their development styles, check the commit log to see who is contributing and check out the profile of the core maintainer.

Check the Network

The network graph. Notice that somebody is already working on a 'mobile' branch, so you probably wouldn't want to duplicate their effort.

The first thing to do is check the Network tab on the project to see all the other forks that other people have made. Spend a few minutes digging around in them, as it's quite possible that somebody is already working on the problem that you'd like to see solved. It'll also give you an idea of the activity of the project, and how likely it is that your changes will be merged in.

Opening an Issue

You've got issues, man.

Next, head over to the Issues tab. Have a look around, see how many issues there are and if anybody has opened up the issue that you're interested in working on. This is an important step that many people forget about, and they just submit major pull requests to maintainers without considering that the maintainers might not have the same intentions with the software as they do. This is especially true if a new feature requires user interface/design changes, as often, that's the aspect of programs that people are the most protective of. If your issue doesn't exist already, open up a new issue. Standard human interaction rules apply here; be friendly, be polite, say thanks for making the project, describe the bug or feature you'd like to work on and then offer to help. Hopefully, they'll reply shortly with some input on how to solve the problem.

Making Your Fork

Hardcore Forking Action

Here's the fun part! Hit 'Fork'. Now you've got your own version! Go to the page, get the ssh: url from the box at the top and then
git clone **your ssh/git url**

to your local machine. Hooray! You have the code on your local machine now.

Make Your Fork Track the Original Upstream Repo

It's a fork. Hur hur hur.

This step isn't absolutely necessary, but I find it very useful if you plan on working on this project for anything more than a very quick fix. Use the following commands to add the 'upsteam' (original project location) as a remote branch so that you can get their updates into your branch. Replace the 'upstreamname' and 'projectname' values with that actual user/project name that you're trying to track. 1
git remote add --track master upstream git://github.com/upstreamname/projectname.git

This Gist brought to you by GitHub.

gistfile1.txt view raw

This will add the original project as a remote named 'upstream'. To get the code, type:
git fetch upstream

Then, to merge it into your own project, type:

git merge upstream/master

Tada! Now you'll have an up-to-date version of the upstream code in your current branch.

Setting Up A Development Branch

Guys, remember the old internet? Oh man.

Now you're getting ready to start hacking, you'll want to switch off of the 'master' branch and onto a different branch for your

new feature. It's important to do this because you can only have onePull Requestper branch, so if you want to submit more than one fix, you'll need to have multiple branches. Make a new branch like this:
git branch newfeature

Then switch to it like this:

git checkout newfeature

Now you're on your new branch. You can confirm this by simply typing 'git branch'.

Hack!
This part's up to you. Hack along as you normally would until the code is in the state where you're happy with it, it performs the task as described and it passes all the tests you've written for it. Yayyyy!

Squashing Your Commits

Squash. Hur hur hur.

If you're a heavy committer like me, you've probably got lots of poorly messaged commits ('works!', 'broken', 'fuck', 'woo', etc.). This is a bad habit, but I haven't been able to break it yet and I know I'm not the only one! Before you submit your pull request back upstream, you'll want to squash these commits into a small handful of well-labeled commits. To do this, we're going to use the git rebase command. First, take a look at the commits we've made with git log and figure out the commits that we want to squash. If we wanted to squash the last 3 commits into one, we'd open up an an interactive rebase like this:
git rebase -i HEAD~3

This will bring you into your editor with some text that will look something like this:
pick df94881 Allow install to SD pick a7323e5 README Junkyism pick 3ead26f rm classpath from git

To squash those commits into one, change to something like this:

pick df94881 Allow install to SD squash a7323e5 README Junkyism squash 3ead26f rm classpath from git

Then, save/quit, and you'll be brought to into another editor session. Describe the changes as well as you can and save/quit again. Hooray! You've squashed your ugly commits into one nice one. Now you're ready to submit a pull request.

Submitting a Pull Request

Once you've commited and squashed your changes, push them to your remote like this:
git push origin newfeature

Then go to that page on GitHub and change branches to the one for your new feature.

Submit a Pull Request!

Then, click on the little button that says 'Pull Request'. This will bring you to a page asking you to describe your change. Describe it thoroughly.

Describe your Pull Request.

Then press 'Submit Pull Request'. Hooray! You did it. Now, you're not quite done yet. If the maintainer finds some problems with your code, they won't want to pull your changes until you fix them. Fortunately, whenever you commit and push more things to that branch of your code, they will be included in that pull request until it is closed.

Accepting And Merging a Pull Request

Bonus! If you're on the receiving end of a pull request, how do you merge the changes? Easy - press the button! GitHub now has an auto-merge button which does everything for you in one click. However, it doesn't always work, in which case you'll have to do the merge on your own machine, like so:
git git git git git checkout master remote add contributor git://github.com/contributor/project fetch contributor merge contributor/newfeature push origin master

And then their changes will be properly merged into your main master branch.

So, Your Pull Request Was Rejected. Now What?

Some forks are unavoidable.

Sometimes, for technical or organizational reasons, your pull request will be rejected. This can feel really frustrating, and there are a few different ways you can proceed. Just remember to act rationally. The simplest thing is to simply accept their judgement. It's their project, and a good maintainer knows when to say "no." If their argument is logically sound, you should accept it. If you don't think it's a particularly important feature, hopefully whoever is looking at the project will check the Network and Issues tabs of the upstream project and will notice your changes. However, I think this is a pretty poor solution in cases when the upstream maintainer is wrong or unresponsive. A better thing to do is write about it. Write about it on your blog, start a discussion on a mailing list, and solicit opinions from the community about what the best way to proceed is. It'll also give some Google-juice to your project/issue, which will help other people who ran into the same problem you faced. The last option is to sever ties with the upstream and declare yourself the new maintainer of the project . This should only be as a last resort and should only really be done when the upstream project is dead or has gone completely off the rails. That being said, this kind of software deadheading can actually breathe a lot of new life into a project - just look at how LibreOffice has managed to revive the OpenOffice project by severing ties with Oracle. If you do this, you should rename your project to differentiate it from the upstream, explicitly state your reasons for the schism in your README, and be sure to give proper copyright credit according the the open source license they originally chose. Maintaining an open source project carries quite a lot of responsibility, so make sure you're prepared to care for the project once you create such a schism.

Wrap Up
Hopefully this little guide was useful for getting you started with collaborative software development on GitHub! If you're a developer looking for more jobs, we at Gun.io would like to help! You can sign up with GitHub and we'll automatically pair you up with freelance and full-time job offers based on your existing code portfolio!

Need a Hacker?
gun.io is where top developers find gigs. Bad developers are screened out, so you'll only hire great hackers at gun.io.

Pssst! Hackers, sign up here!

Comments

About gun.io
We match top developers with high-quality freelance and full-time jobs.

Infrastructure Programmer
Cryptic Studios

Los Gatos

Artisteer eZ Publish Integration

$350

Front-End Developer - js/jQuery/HTML/CSS, Shot in the dark here, but VisualForce and APEX (?)
$1025

Sign up with GitHub and we'll match you with great gigs based on your open-source portfolio!

25 comments 2 reactions
Leave a message...
Discussion Community

22

Share

Steve Klabnik 11 months ago

You should really rebase the commits from upstream rather than merge, so you don't make a crazy graph with tons of merge commits.
4

Reply

Share

g b > Steve Klabnik 11 months ago

Something like this? http://blog.bleeds.info/sofa/_...

man, this is all too much work. I need a Makefile just for the git stuff.
0

Reply

Share

Tim Smart > g b 2 months ago

`git rebase master` on whatever branch you are on (change 'master' with the upstream branch). Just replacing 'merge' with 'rebase' gets the job done. The main problem comes when pushing your commits to the remote server. 'rebase' creates non-fast-forward commits which any sane git server setup will reject. This means to update your branch you will have to delete the remote branch and re-push it. `git push origin :my-branch && git push origin my-branch` and you are done.
0

Reply

Share

Josh Kaufman 11 months ago

This is a great post. I'm new to Github, and didn't know how to track the original repo of a fork or squash commits. Thanks for putting this together!
1

Reply

Share

Eric Pardee 3 days ago

I found this helpful, thank you.

0

Reply

Share

Charles Edward Smith 24 days ago

Okay, so how does one do this with TortoiseGIT?

0

Reply

Share

Steve Powell a month ago

I tried to follow the `Make Your Fork Track the Original Upstream Repo` section but found that I needed to use the `git @ github:username/projectname` format for it to work properly (spaces added to avoid re-formatting). Otherwise `git fetch` simply times out unable to make contact. Thanks for a useful page.
0

Reply

Share

Sonny Darvishzadeh a month ago

I'm new to GitHub and this is a good start, thanks for the article
0

Reply

Share

Shoop a month ago

Great post! But I would be interested in: "If you're on the [sending] end of a pull request, how do you merge the changes [if the pull request has been accepted?]" :)
0

Reply

Share

Dineshkummarc 2 months ago

thanks Rich Jones, what if i forked using github UI and want to get the updated commits for the particular fork. please update me.
0

Reply

Share

Levi Notik 3 months ago

Awesome post. I always come back to this.

0

Reply

Share

Greg Gauthier 4 months ago

How do you do a pull request without having to fork? Work wants me to use pull requests, but I got my knuckles slapped for forking the repo in order to do that according to githubs instructions.
0

Reply

Share

Hari K T 7 months ago

Great Post :)
0

Reply

Share

Can I translate this post in my language?

0

Reply

Share

Rich > lethee 11 months ago

Yes!, Please be my guest. It'd be nice if you linked back to here though..
0

Reply

Share

lethee > Rich 11 months ago

The URL will be http://dogfeet.github.com/arti... and be published in some days.

1

Reply

Share

g b 11 months ago

before the squashing step, it's missing a step on how to send your stuff to your github fork, before you do any pull request! what if you are working with another contributor on that feature? what if you are working at home and at work?
0

Reply

Share

Chris 11 months ago

Seems "git remote add upstream ..." is missing from the section "Make Your Fork Track the Original Upstream Repo".
0

Reply

Share

isomorphisms 11 months ago

Where's the "undo" button? I've struggled with git-rewind every time I've tried to use it.
0

Reply

Share

DiabloD3 11 months ago

There is a rather large lack of git flow usage going on here.

0

Reply

Share

Dude > DiabloD3 11 months ago

Erm, git flow? Git is a distributed version control system that supports TONS of different workflows. You should choose one based on merits and applicability to the project. Plus, git flow sucks. Get over it.
1

Reply

Share

Decklin 11 months ago

You can also merge pull requests from your own working copy without adding the contributor as a remote: git fetch refs/pull/123/head:pull-123 git merge pull-123 There is also a 'refs/pull/123/remote/merge' which gives you the commit that the merge button would produce, but unfortunately the commit message is not very useful.
0

Reply

Share

Guest 11 months ago

Naming the parent repository 'upstream' is not a good idea since all branches have upstreams that can be set with --set-upstream switch.
0

Reply

Share

Chris Ryland 11 months ago

Note that the step "git remote add --track master upstream ..." isn't necessary--the "git clone" will already automatically track the remote master for you.
0

Reply

Share

Chris Ballinger > Chris Ryland 11 months ago

It won't automatically track the upstream's remote master though. It will only track your own fork's remote master.
2

Reply

Share

ALSO ON GUNIO BLOG

What's this?

1 comment 23 days ago 2 comments 24 days ago

r Comment feed

m Subscribe via email

Pssst.. do you want to write for Gun.io? Or, do you have a blog you'd like to see syndicated here? Get in touch!

About
gun.io is where hackers find jobs. Every month, tens of thousands of the best independent software developers from the open source community use gun.io to find high-quality freelance and full-time jobs.

Links
Philosophy Community Premium Services Team Blog @GUNdotIO

Contact
Need assistance hiring? Let us help! Your email address Or just mail the boss, [email protected], directly! Send it!