CODEWITHSUNDEEP
git
DJJ
DJJ

Reputation: 199

Why would you used '--track' options in git branches?

This is my understanding about tracking git branches, please correct me if I am missing something here. To track a remote branch, you would first run git fetch origin, which brings a copy image of the remote git repository. Let's say you list the branches you want to create a tacking branch to mybranch, you'd simply run git branch mybranch and then git will automatically create a remote branch and make it tracking the corresponding (name correlation) remote branch. Very simple! The question is, why do most of the git folks talk about the option --track if git is capable to identify the branch you just create (based on name correlation) is available in origin or remote repository?

Upvotes: 2

Views: 3055

Answers (2)

torek
torek

Reputation: 487805

There are some other details that are not quite right but the key is here:

you'd simply run git branch mybranch and then git will ... make it tracking the corresponding (name correlation) remote branch

When you use git branch mybranch to create a branch, it doesn't do that. It does do that, at least by default, if you run git branch mybranch origin/mybranch.

If you use git branch mybranch, though, Git creates mybranch based on the current commit hash, as if you had done:

git branch mybranch $(git rev-parse HEAD)

and the default behavior when naming a remote-tracking branch is controlled by the configuration setting in branch.autoSetupMerge.

The git checkout command will also sometimes, but not always, create a new branch using an implied --track and remote-tracking branch. Specifically, given git checkout name, there are numerous optional paths, depending on whether name can be resolved:

  • name is the name of an existing branch: git checkout tries to switch to that existing branch.
  • name is the name of an existing tag, or otherwise resolves to a commit through the multi-step procedure described in the gitrevisions documentation: git checkout tries to check out that particular commit as a detached HEAD.
  • name is not any of the above, but there exists exactly one remote-tracking branch which, if stripped of its remote, matches name (e.g., you enter git checkout develop when there is no develop but there is an origin/develop): in this case, Git will create a new branch develop that "tracks" (has as its upstream) origin/develop, just as git branch would with the appropriate arguments and appropriate branch.autoSetupMerge and/or branch.autoSetupRebase settings.

If all of the above fail, git checkout tries interpreting name as a path name. You can force git checkout to skip the above steps (and go straight for the path name interpretation) using the -- option-terminator: git checkout -- master means check out a file named master and not a branch.

Upvotes: 5

Mark Adelsberger
Mark Adelsberger

Reputation: 45659

The automatic tracking you mention is a convenience "shorthand", but it doesn't always work.

Sometimes there are multiple remotes with the same branch name; in that case git doesn't know which one to track and won't automatically set up tracking info.

Also, sometimes you want a local branch to track a remote branch that doesn't have the same name.

Some developers might, I suppose, never need to explicitly set tracking info; but some do, so you can.

Upvotes: 2

Related Questions