6.3 GitHub - Maintaining a Project

Someone comes along and makes a change to your code and sends you a Pull Request. You should get an email notifying you about the new Pull Request and it should look something like Email notification of a new Pull Request.

There are a few things to notice about this email. It will give you a small diffstat — a list of files that have changed in the Pull Request and by how much. It gives you a link to the Pull Request on GitHub. It also gives you a few URLs that you can use from the command line.

If you notice the line that says git pull <url> patch-1, this is a simple way to merge in a remote branch without having to add a remote. We went over this quickly in Checking Out Remote Branches. If you wish, you can create and switch to a topic branch and then run this command to merge in the Pull Request changes.

The other interesting URLs are the .diff and .patch URLs, which as you may guess, provide unified diff and patch versions of the Pull Request. You could technically merge in the Pull Request work with something like this:

As we covered in The GitHub Flow, you can now have a conversation with the person who opened the Pull Request. You can comment on specific lines of code, comment on whole commits or comment on the entire Pull Request itself, using GitHub Flavored Markdown everywhere.

Every time someone else comments on the Pull Request you will continue to get email notifications so you know there is activity happening. They will each have a link to the Pull Request where the activity is happening and you can also directly respond to the email to comment on the Pull Request thread.

Once the code is in a place you like and want to merge it in, you can either pull the code down and merge it locally, either with the git pull <url> <branch> syntax we saw earlier, or by adding the fork as a remote and fetching and merging.

If the merge is trivial, you can also just hit the “Merge” button on the GitHub site. This will do a “non-fast-forward” merge, creating a merge commit even if a fast-forward merge was possible. This means that no matter what, every time you hit the merge button, a merge commit is created. As you can see in Merge button and instructions for merging a Pull Request manually, GitHub gives you all of this information if you click the hint link.

If you decide you don’t want to merge it, you can also just close the Pull Request and the person who opened it will be notified.

If you’re dealing with a lot of Pull Requests and don’t want to add a bunch of remotes or do one time pulls every time, there is a neat trick that GitHub allows you to do. This is a bit of an advanced trick and we’ll go over the details of this a bit more in The Refspec, but it can be pretty useful.

GitHub actually advertises the Pull Request branches for a repository as sort of pseudo-branches on the server. By default you don’t get them when you clone, but they are there in an obscured way and you can access them pretty easily.

To demonstrate this, we’re going to use a low-level command (often referred to as a “plumbing” command, which we’ll read about more in Plumbing and Porcelain) called ls-remote. This command is generally not used in day-to-day Git operations but it’s useful to show us what references are present on the server.

If we run this command against the “blink” repository we were using earlier, we will get a list of all the branches and tags and other references in the repository.

Of course, if you’re in your repository and you run git ls-remote origin or whatever remote you want to check, it will show you something similar to this.

If the repository is on GitHub and you have any Pull Requests that have been opened, you’ll get these references that are prefixed with refs/pull/. These are basically branches, but since they’re not under refs/heads/ you don’t get them normally when you clone or fetch from the server — the process of fetching ignores them normally.

There are two references per Pull Request - the one that ends in /head points to exactly the same commit as the last commit in the Pull Request branch. So if someone opens a Pull Request in our repository and their branch is named bug-fix and it points to commit a5a775, then in our repository we will not have a bug-fix branch (since that’s in their fork), but we will have pull/<pr#>/head that points to a5a775. This means that we can pretty easily pull down every Pull Request branch in one go without having to add a bunch of remotes.

Now, you could do something like fetching the reference directly.

This tells Git, “Connect to the origin remote, and download the ref named refs/pull/958/head.” Git happily obeys, and downloads everything you need to construct that ref, and puts a pointer to the commit you want under .git/FETCH_HEAD. You can follow that up with git merge FETCH_HEAD into a branch you want to test it in, but that merge commit message looks a bit weird. Also, if you’re reviewing a lot of pull requests, this gets tedious.

There’s also a way to fetch all of the pull requests, and keep them up to date whenever you connect to the remote. Open up .git/config in your favorite editor, and look for the origin remote. It should look a bit like this:

That line that begins with fetch = is a “refspec.” It’s a way of mapping names on the remote with names in your local .git directory. This particular one tells Git, "the things on the remote that are under refs/heads should go in my local repository under refs/remotes/origin." You can modify this section to add another refspec:

That last line tells Git, “All the refs that look like refs/pull/123/head should be stored locally like refs/remotes/origin/pr/123.” Now, if you save that file, and do a git fetch:

Now all of the remote pull requests are represented locally with refs that act much like tracking branches; they’re read-only, and they update when you do a fetch. This makes it super easy to try the code from a pull request locally:

The eagle-eyed among you would note the head on the end of the remote portion of the refspec. There’s also a refs/pull/#/merge ref on the GitHub side, which represents the commit that would result if you push the “merge” button on the site. This can allow you to test the merge before even hitting the button.

Not only can you open Pull Requests that target the main or master branch, you can actually open a Pull Request targeting any branch in the network. In fact, you can even target another Pull Request.

If you see a Pull Request that is moving in the right direction and you have an idea for a change that depends on it or you’re not sure is a good idea, or you just don’t have push access to the target branch, you can open a Pull Request directly to it.

When you go to open a Pull Request, there is a box at the top of the page that specifies which branch you’re requesting to pull to and which you’re requesting to pull from. If you hit the “Edit” button at the right of that box you can change not only the branches but also which fork.

Here you can fairly easily specify to merge your new branch into another Pull Request or another fork of the project.

When we mention “notifications” here with respect to GitHub, we mean a specific way that GitHub tries to get in touch with you when events happen and there are a few different ways you can configure them. If you go to the “Notification center” tab from the settings page, you can see some of the options you have.

The two choices are to get notifications over “Email” and over “Web” and you can choose either, neither or both for when you actively participate in things and for activity on repositories you are watching.

If you are using a branch other than “master” as your default branch that you want people to open Pull Requests on or see by default, you can change that in your repository’s settings page under the “Options” tab.

Simply change the default branch in the dropdown and that will be the default for all major operations from then on, including which branch is checked out by default when someone clones the repository.

If you would like to transfer a project to another user or an organization in GitHub, there is a “Transfer ownership” option at the bottom of the same “Options” tab of your repository settings page that allows you to do this.

This is helpful if you are abandoning a project and someone wants to take it over, or if your project is getting bigger and want to move it into an organization.

Not only does this move the repository along with all its watchers and stars to another place, it also sets up a redirect from your URL to the new place. It will also redirect clones and fetches from Git, not just web requests.