Question

I did this:

git branch --track stats_page

to create a separate branch to try something out. But now everytime I push or pull I see:

$ git pull
From .
 * branch            master     -> FETCH_HEAD

and

$ git push
Total 0 (delta 0), reused 0 (delta 0)
To .
   ff0caf5..e360168  stats_page -> master

I’m far from a git master so I might be miss understanding what’s happening – but is my stats_page branch tracking to master?

I may have created the branch wrong – if so – can I fix it so that from now on it pushes and pulls to the remote stats_page branch? How?

Answer 1

Quick answer: The easiest way to fix this is in Make an existing Git branch track a remote branch?

What’s really going on here

I’ve always found descriptions of git’s “tracking branches” confusing myself. They sound, to me at least, like they do a lot more than they really do.

The trick, I think, is to realize that git doesn’t ever actually “track” a remote repository at all. (It does fetch from a remote repo.) Instead, a “tracking branch” tracks a branch in your own repo. To “track a remote branch”, you create a branch whose name says that it was originally copied from some other repo. (Under the covers, this means any branch stored under refs/remotes/, more or less.) The most common “other repo” is the one named origin, which is the one that an initial git clone sets up for you. As it says in the git-clone manual page:

This default configuration is achieved by creating references to the remote branch heads under refs/remotes/origin and by initializing remote.origin.url and remote.origin.fetch configuration variables.

When you use git branch --track to create a branch, it makes the new branch “track” the “start point” that you give it. (And, “track” mainly means “associate for various push/pull/fetch actions, and tell me more when I run git status.) So, assuming the repo you cloned-from already has a stats_page branch, you probably meant to do this:

git branch --track stats_page origin/stats_page

which tells git “make new branch named stats_page that tracks what I have in refs/remotes/origin/stats_page” (you normally leave off the refs/heads and refs/remotes parts, and git figures them out). You don’t even need the --track here, as the origin/ part on the right turns that on for you.

Once you do all that, in git’s terminology, you have a “local” branch (refs/heads/stats_page) that tracks a “remote” branch (refs/remotes/origin/stats_page). To actually update the “remote” branch you use git fetch, or more explicitly, git fetch origin, which goes out to the URL associated with origin and collects any new stuff and then adds it to your repo. Hence, despite calling this “remote”, it’s still remarkably local: it’s right there on your own local disk.

(I have taken to calling git “the Borg of SCMs”: “We are the git-Borg. We will add your commit distinctiveness to our own.” Most of what you do just adds new stuff to your own repo-Borg-collective.)

Once you’ve used git fetch to update your local copy of the “remote” branch (origin/stats_page), you can do a git merge to get it merged in (to stats_page, the name without the origin/ part). And, a git pull just runs those two in one easy step—which hides the fact that it’s the fetch that actually gets new stuff from the “remote”

Now, the problem you have is that when you ran:

git branch --track stats_page

you were on your master branch, so you created a new “local” branch named stats_page that tracks your master, instead of tracking origin/stats_page. Hence, git pull and git push just pull and push on your local branch named master. (Note: If you actually run git fetch, that still fetches into remotes/origin/*, so it still updates origin/stats_page, even though the one “local” branch is mistakenly tracking another “local” branch.)

One last note: usually when you make a “local” branch that tracks a remote branch of the same name, you also want to do a git checkout immediately. You can just git checkout -b stats_page origin/stats_page to get all three steps done at once. (If your version of git is old, it may not automatically track when you name origin/stats_page on the right; but if so you probably should just update your version of git.)

---

Footnote: Some people recommend against using git pull at all, and I sort of agree with them. git pull really mixes two very different operations. In particular git pull remote_A remote_B seems to surprise people when it does an octopus merge. Using git fetch never seems to surprise people; and once they get used to, and familiar with, an explicit git merge, the logic behind fast-forward merges makes a lot more sense. Still, git pull with no arguments is so convenient….