忘记搬到另一个分支，而不是掌握Git的人

Question

I'm using Git with a rails project.

Today i make a lot of changes and i forget to move to another branch , so all these changes are in the master branch but not committed yet .

this is what i get when i do git status command :

# On branch master
# Changes not staged for commit:
#   (use "git add <file>..." to update what will be committed)
#   (use "git checkout -- <file>..." to discard changes in working directory)
#
#   modified:   app/controllers/sessions_controller.rb
#   modified:   app/controllers/users_controller.rb
#   modified:   app/helpers/sessions_helper.rb
#   modified:   app/models/user.rb
# .....

#
# Untracked files:
#   (use "git add <file>..." to include in what will be committed)
#
#   app/assets/javascripts/password_resets.js.coffee
#   app/assets/stylesheets/password_resets.css.scss
#   app/controllers/password_resets_controller.rb
# .....
# ......
no changes added to commit (use "git add" and/or "git commit -a")

how i can solve this error and move all these untracked files from master to another branch already exists ? thank you.

Answer 1

Just create a new branch and then commit, eg:

git checkout -b my_branch
git commit -am "My commit."

Answer 2

You simply need to create new branch.

Once you create a branch all the changes will be in the new branch.

and then clean your old stuff from master

// Checkout new branch
git checkout -b my_branch

// Prepare files for commit
git add .

// Commit changes
git commit -m "Your commit message"

// delete the dirty master
git branch -D master

// Now we are going to clean your changes from the master branch
// the -f is important in case you skipped the previous command to delete your master
git checkout -f master 

Answer 3

If there are no commits marked C1, C2, etc. below

You're in the simple case; just use Henrik N's answer .

If there is a tracking branch

Your updated question/comments suggests that you have both committed changes and uncommitted changes, and you did them all on some tracking branch, such as master that tracks origin/master or devel that tracks origin/devel , but that you have not pushed any of them yet. (I'll assume master but you can just change each master to the feature/development/whatever branch name below.)

As a diagram, your repo's commit tree and working directory/index now looks something like this:

M1 -...- M5 -- M6                    <-- origin/master
                 \
                  C1 -- C2           <-- master, HEAD=master
                          \
                           i, w      <-- "index" and working-tree files

Here i is the stuff in the index ( git add dir/file1 , git rm file2 , etc), which git status shows as changes to be committed , and w is the stuff in your working directory that git status shows as changes not staged for commit .

Here's what you want it to look like:

M1 -...- M5 -- M6                    <-- master, origin/master
                 \
                  C1 -- C2           <-- my_branch, HEAD=my_branch
                          \
                           i, w      <-- "index" and working-tree files

Remember that branch labels are like sticky-notes ("yellow stickies" or "Post-It® notes" or whatever): they have a name written on them and are pasted on to a commit. So what you have to do is add a new sticky-note, my_branch , pointing to commit C2, and also make HEAD refer to the new branch name:

git checkout -b my_branch

This uses the previous value of HEAD to name the commit that my_branch is pasted-on-to. Since that commit is C2 you now have both master and my_branch here. It then rewrites HEAD to point to the name my_branch . Now you have:

M1 -...- M5 -- M6                    <-- origin/master
                 \
                  C1 -- C2           <-- master, my_branch, HEAD=my_branch
                          \
                           i, w      <-- "index" and working-tree files

Now you need to move the sticky-note labeled master to point back where origin/master points. There are two ways to do this, with git reset , or with git branch . Using git reset is harder, so let's do it with git branch :

git branch -f master origin/master

The -f (force) flag tells git branch to change an existing sticky-note, rather than failing because it exists, and origin/master gives which commit master should point-to: so git peels the label off C2 and pastes it on M6 , and you have the setup you wanted (the 2nd diagram).

---

Summary: it's just two commands:

git checkout -b my_branch           # create new branch and change HEAD
git branch -f master origin/master  # restore master to origin/master

You're now ready to check in changes as usual, on branch my_branch . ( git add , git rm , git commit as needed.) That will add a new commit C3:

M1 -...- M5 -- M6                    <-- master, origin/master
                 \
                  C1 -- C2 -- C3     <-- my_branch, HEAD=my_branch

---

What if there is no tracking branch?

Well, that's OK, it's just a little harder. Go back to the "have now" and "want" diagrams. In these diagrams, the commit I labeled M6 is the one you want master to name. You now have to find M6. Commits have these long hexadecimal-numeric (SHA-1) "true names" that never change, like 5e013711f5d6eb3f643ef562d49a131852aa4aa1 for example. ( git rev-parse HEAD will show the "true name" for the current, ie, HEAD , commit.) You can find that "true name", or you can find some alternate or abbreviated name that also names the same thing.

There are a whole lot of options for names, but let's stick with numbers that come out of git log and git log --oneline . The latter gives you something like this:

97206f5 peerish: set socket options earlier
4881af5 add semtest.c
b3f8bea fdm: repair example

where the number is an abbreviated SHA-1 and the text is the first line of the commit message. Since the logs are shown in reverse order, the top line is the most recent commit ( C2 in the example above), and then each next line is the next one back (well, it's more complicated with merges, but close enough).

Let's say that from the above, you can immediately tell that the commit I labeled M6 is b3f8bea , ie, that you need to skip two commits. Simply supply that value as the last argument to git branch :

git branch -f master b3f8bea

If your repo is huge and old, or has a lot of merge commits, you might want your logs to include more "decoration". I got these aliases from somewhere long ago:

[alias]
    lol = git log --graph --decorate --oneline
    lola = git log --graph --decorate --oneline --all

so that I can run git lol and git lola .

Alternatively, if you know for sure there were exactly two commits, master~2 (and in this case, after adding my_branch , my_branch~2 ) will get you the same SHA-1. The git rev-parse command shows you exactly which SHA-1 any given name "means". In one repo I have here, HEAD~2 is the same as origin/featureX :

$ git rev-parse HEAD~2
0f5a13497dd3da8aff8e452c8f56630f83253e79
$ git rev-parse origin/featureX
0f5a13497dd3da8aff8e452c8f56630f83253e79

but in general git log (or git lol , etc.) will get you the raw SHA-1 and you can add labels from there, as needed.