The <src> is often the name of the branch you would want to push, but it can be any arbitrary "SHA-1 expression", such as master~4 or HEAD (see gitrevisions[7]).

The <dst> tells which ref on the remote side is updated with this push. Arbitrary expressions cannot be used here, an actual ref must be named. If git push [<repository>] without any <refspec> argument is set to update some ref at the destination with <src> with remote.<repository>.push configuration variable, :<dst> part can be omitted—​such a push will update a ref that <src> normally updates without any <refspec> on the command line. Otherwise, missing :<dst> means to update the same ref as the <src>.

If <dst> doesn’t start with refs/ (e.g. refs/heads/master) we will try to infer where in refs/* on the destination <repository> it belongs based on the type of <src> being pushed and whether <dst> is ambiguous.

The object referenced by <src> is used to update the <dst> reference on the remote side. Whether this is allowed depends on where in refs/* the <dst> reference lives as described in detail below, in those sections "update" means any modifications except deletes, which as noted after the next few sections are treated differently.

The refs/heads/* namespace will only accept commit objects, and updates only if they can be fast-forwarded.

The refs/tags/* namespace will accept any kind of object (as commits, trees and blobs can be tagged), and any updates to them will be rejected.

It’s possible to push any type of object to any namespace outside of refs/{tags,heads}/*. In the case of tags and commits, these will be treated as if they were the commits inside refs/heads/* for the purposes of whether the update is allowed.

I.e. a fast-forward of commits and tags outside refs/{tags,heads}/* is allowed, even in cases where what’s being fast-forwarded is not a commit, but a tag object which happens to point to a new commit which is a fast-forward of the commit the last tag (or commit) it’s replacing. Replacing a tag with an entirely different tag is also allowed, if it points to the same commit, as well as pushing a peeled tag, i.e. pushing the commit that existing tag object points to, or a new tag object which an existing commit points to.

Tree and blob objects outside of refs/{tags,heads}/* will be treated the same way as if they were inside refs/tags/*, any update of them will be rejected.

All of the rules described above about what’s not allowed as an update can be overridden by adding an the optional leading + to a refspec (or using --force command line option). The only exception to this is that no amount of forcing will make the refs/heads/* namespace accept a non-commit object. Hooks and configuration can also override or amend these rules, see e.g. receive.denyNonFastForwards in git-config[1] and pre-receive and update in githooks[5].

Pushing an empty <src> allows you to delete the <dst> ref from the remote repository. Deletions are always accepted without a leading + in the refspec (or --force), except when forbidden by configuration or hooks. See receive.denyDeletes in git-config[1] and pre-receive and update in githooks[5].

The special refspec : (or +: to allow non-fast-forward updates) directs Git to push "matching" branches: for every branch that exists on the local side, the remote side is updated if a branch of the same name already exists on the remote side.

tag <tag> means the same as refs/tags/<tag>:refs/tags/<tag>.

This option overrides this restriction if the current value of the remote ref is the expected value. "git push" fails otherwise.

Imagine that you have to rebase what you have already published. You will have to bypass the "must fast-forward" rule in order to replace the history you originally published with the rebased history. If somebody else built on top of your original history while you are rebasing, the tip of the branch at the remote may advance with her commit, and blindly pushing with --force will lose her work.

This option allows you to say that you expect the history you are updating is what you rebased and want to replace. If the remote ref still points at the commit you specified, you can be sure that no other people did anything to the ref. It is like taking a "lease" on the ref without explicitly locking it, and the remote ref is updated only if the "lease" is still valid.

--force-with-lease alone, without specifying the details, will protect all remote refs that are going to be updated by requiring their current value to be the same as the remote-tracking branch we have for them.

--force-with-lease=<refname>, without specifying the expected value, will protect the named ref (alone), if it is going to be updated, by requiring its current value to be the same as the remote-tracking branch we have for it.

--force-with-lease=<refname>:<expect> will protect the named ref (alone), if it is going to be updated, by requiring its current value to be the same as the specified value <expect> (which is allowed to be different from the remote-tracking branch we have for the refname, or we do not even have to have such a remote-tracking branch when this form is used). If <expect> is the empty string, then the named ref must not already exist.

Note that all forms other than --force-with-lease=<refname>:<expect> that specifies the expected current value of the ref explicitly are still experimental and their semantics may change as we gain experience with this feature.

"--no-force-with-lease" will cancel all the previous --force-with-lease on the command line.

A general note on safety: supplying this option without an expected value, i.e. as --force-with-lease or --force-with-lease=<refname> interacts very badly with anything that implicitly runs git fetch on the remote to be pushed to in the background, e.g. git fetch origin on your repository in a cronjob.

The protection it offers over --force is ensuring that subsequent changes your work wasn’t based on aren’t clobbered, but this is trivially defeated if some background process is updating refs in the background. We don’t have anything except the remote tracking info to go by as a heuristic for refs you’re expected to have seen & are willing to clobber.

If your editor or some other system is running git fetch in the background for you a way to mitigate this is to simply set up another remote:

Now when the background process runs git fetch origin the references on origin-push won’t be updated, and thus commands like:

Will fail unless you manually run git fetch origin-push. This method is of course entirely defeated by something that runs git fetch --all, in that case you’d need to either disable it or do something more tedious like:

I.e. create a base tag for versions of the upstream code that you’ve seen and are willing to overwrite, then rewrite history, and finally force push changes to master if the remote version is still at base, regardless of what your local remotes/origin/master has been updated to in the background.

This flag disables these checks, and can cause the remote repository to lose commits; use it with care.

Note that --force applies to all the refs that are pushed, hence using it with push.default set to matching or with multiple push destinations configured with remote.*.push may overwrite refs other than the current branch (including local refs that are strictly behind their remote counterpart). To force a push to only one branch, use a + in front of the refspec to push (e.g git push origin +master to force a push to the master branch). See the <refspec>... section above for details.

For a failed update, more details are given:

The default behavior of this command when no <refspec> is given can be configured by setting the push option of the remote, or the push.default configuration variable.

For example, to default to pushing only the current branch to origin use git config remote.origin.push HEAD. Any valid <refspec> (like the ones in the examples below) can be configured as the default for git push origin.

See the section describing <refspec>... above for a discussion of the matching semantics.

This is to emulate git fetch run on the mothership using git push that is run in the opposite direction in order to integrate the work done on satellite, and is often necessary when you can only make connection in one way (i.e. satellite can ssh into mothership but mothership cannot initiate connection to satellite because the latter is behind a firewall or does not run sshd).

After running this git push on the satellite machine, you would ssh into the mothership and run git merge there to complete the emulation of git pull that were run on mothership to pull changes made on satellite.

The above command would change the origin repository to

Commits A and B would no longer belong to a branch with a symbolic name, and so would be unreachable. As such, these commits would be removed by a git gc command on the origin repository.