Difference between revisions of "Git"

From ArchWiki
Jump to navigation Jump to search
m (Corrected git clone git@github.com:torvalds/linux.git to git clone git://github.com/torvalds/linux.git)
(→‎Git repositories rights: Report dead link)
Line 307: Line 307:
=== Git repositories rights ===
=== Git repositories rights ===
To restrict read/write access, you can simply use Unix rights, see http://sitaramc.github.com/gitolite/doc/overkill.html
To restrict read/write access, you can simply use Unix rights, see http://sitaramc.github.com/gitolite/doc/overkill.html{{linkrot|2013|11|06}}
For a fine-grained rights access, see [[gitolite]] and [[gitosis]]
For a fine-grained rights access, see [[gitolite]] and [[gitosis]]

Revision as of 06:17, 7 November 2013

zh-CN:Git Template:Article summary start Template:Article summary text Template:Article summary heading Template:Article summary wiki: Generally about contributing to pacman, although it still serves as a practical Git tutorial Template:Article summary wiki Template:Article summary wiki Template:Article summary wiki Template:Article summary wiki Template:Article summary link Template:Article summary end

Git is the version control system (VCS) coded by Linus Torvalds (the creator of the Linux kernel) after being criticized for using the proprietary BitKeeper with the Linux kernel. Git is now used to maintain sources for the Linux kernel as well as thousands of other projects, including Pacman, Arch's package manager.

There is extensive documentation, including guides and tutorials, available from the official web site.


git can be installed with pacman from the official repositories. If you care about using Git with other VCS software, mail servers, or using Git's GUI, pay close attention to the optional dependencies.

Bash completion (e.g. hitting Tab to complete commands you are typing) should work if you add this line to your ~/.bashrc file:

source /usr/share/git/completion/git-completion.bash

Alternatively, you can install the bash-completion package to load the completions automatically for new shells.

If you want to use Git's built-in GUI (e.g. gitk or git gui) you should install the tk package, or you will get a rather cryptic message:

/usr/bin/gitk: line 3: exec: wish: not found.


Git reads its configuration from a few INI type configuration files. In each git repository .git/config is used for configuration options specific to that repository. Per-user ("global") configuration in $HOME/.gitconfig is used as a fall-back from the repository configuration. You can edit the files directly but the preferred method is to use the git-config utility. For example,

$ git config --global core.editor "nano -w"

adds editor = nano -w to the [core] section of your ~/.gitconfig file.

The man page for the git-config utility has a fairly long list of variables which can be set.

The two settings you should set before using Git are your name and email. These are used to sign commits you make.

$ git config --global user.name "Firstname Lastname"
$ git config --global user.email "your_email@youremail.com"

Colors in Git Prompt

color.ui is also a very useful option to set - it colorizes all Git output.

$ git config --global color.ui true

Basic Usage

Cloning a repository

git clone <repo location> <dir>

will clone a Git repository in a new directory inside your current directory. Leaving out <dir> will cause it to name the folder after the Git repository. For example,

git clone git://github.com/torvalds/linux.git

clones Github's mirror of the Linux kernel into a directory named "linux".

Committing files

Git's commit process involves two steps:

  1. Add new files, add changes for existing files (both with git add <files>), and/or remove files (with git rm). These changes are put in a staging area called the index.
  2. Call git commit to commit the changes.

Git commit will open up a text editor to provide a commit message. You can set this editor to whatever you want by changing the core.editor option with git config.

Alternatively, you can use git commit -m <message> to supply the commit message without opening the text editor.

Other useful commit tricks:

git commit -a lets you commit changes you have made to files already under Git control without having to take the step of adding the changes to the index. You still have to add new files with git add.

git add -p lets you commit specific parts of files you have changed. This is useful if you have made a bunch of changes that you think would be best split into several commits.

Pushing your changes

To push your changes up to a server (such as Github), use

git push <server name> <branch>

Adding -u will make this server the default one to push to for this branch. If you have cloned the repository as described above, the server will default to the location you cloned the repository from (nicknamed "origin") and the branch will default to the master branch. In other words, if you have followed this guide's instructions in cloning, git push will suffice. You can set up Git to push to multiple servers if you want, but that is a more advanced topic. Branches will be discussed later in this guide.

Pulling from the server

If you are working on multiple machines and want to update your local repository to what the server has, change directory into your local repository and use

git pull <server name> <branch>

Similarly to push, the server name and branch should have sane defaults, so git pull should suffice. Git pull is actually shorthand for doing two things:

  1. Calling git fetch, which updates the local copy of what the server has. Such branches are called "remote" branches because they are mirroring remote servers.
  2. Calling git merge, which merges what the remote branch has with what you have. If your commit history is the same as the server's commit history, you will be automatically fast-forwarded to the latest commit on the server. If your history does not match (maybe someone else has pushed commits since you last synced), the two histories will be merged.

It is not a bad idea to get into the practice of using these two commands instead of git pull. This way you can check to make sure that the server contains what you would expect before merging.

Examining history

The command git log shows the history of your current branch. Note that each commit is identified by a SHA-1 hash. The author, commit date, and commit message follow. A more useful command is

git log --graph --oneline --decorate

which provides a display similar to TortoiseGit's log window. It shows the following:

  • The first 7 digits of each commit's SHA-1 hash (enough to be unique)
  • The --graph option shows how any branches (if there are others) fork off from the current branch.
  • The --oneline option shows only the first line of each commit message
  • The --decorate option shows all commit labels (branches and tags)

It may be convenient to alias this command as git graph by doing the following:

git config --global alias.graph 'log --graph --oneline --decorate'

Now typing git graph will run git log --graph --oneline --decorate. git graph and git log may be given the --all flag in order to view all branches instead of just the current one. Adding --stat to one of these commands is also useful - it shows which files each commit changed and how many lines were changed in each file.

Dealing With Merges

Merges happen when you pull, as a result of a rebase operation, and when you merge one branch into another. Like other version control tools, when Git cannot automatically merge a commit, it turns to you. See this section of the Git Book for an explanation on how to resolve merge conflicts. If you screw up and would like to back out of the merge, you can usually abort the merge using the --abort flag with whatever command started the merge (e.g. git merge --abort, git pull --abort, git rebase --abort).

Taking Advantage of DVCS

The above commands only provide the basics. The real power and convenience in Git (and other distributed version control systems) come from leveraging its local commits and fast branching. A typical Git workflow looks like this:

  1. Create and check out a branch to add a feature.
  2. Make as many commits as you would like on that branch while developing that feature.
  3. Squash, rearrange, and edit your commits until you are satisfied with the commits enough to push them to the central server and make them public.
  4. Merge your branch back into the main branch.
  5. Delete your branch, if you desire.
  6. Push your changes to the central server.

Creating a branch

git branch <branch name>

can be used to create a branch that will branch off the current commit. After it has been created, you should switch to it using

git checkout <branch name>

A simpler method is to do both in one step with

git checkout -b <branch name>

To see a list of branches, and which branch is currently checked out, use

git branch

A word on commits

Many of the following commands take commits as arguments. A commit can be identified by any of the following:

  • Its 40-digit SHA-1 hash (the first 7 digits are usually sufficient to identify it uniquely)
  • Any commit label such as a branch or tag name
  • The label HEAD always refers to the currently checked-out commit (usually the head of the branch, unless you used git checkout to jump back in history to an old commit)
  • Any of the above plus ~ to refer to previous commits. For example, HEAD~ refers to one commit before HEAD and HEAD~5 refers to five commits before HEAD.

Commits as checkpoints

In Subversion and other older, centralized version control systems, commits are permanent - once you make them, they are there on the server for everyone to see. In Git, your commits are local and you can combine, rearrange, and edit them before pushing them to the server. This gives you more flexibility and lets you use commits as checkpoints. Commit early and commit often.

Editing the previous commit

git commit --amend

allows you to modify the previous commit. The contents of the index will be applied to it, allowing you to add more files or changes you forgot to put in. You can also use it to edit the commit message, if you would like.

Squashing, rearranging, and changing history

git rebase -i <commit>

will bring up a list of all commits between <commit> and the present, including HEAD but excluding <commit>. This command allows you rewrite history. To the left of each commit, a command is specified. Your options are as follows:

  • The "pick" command (the default) uses that commit in the rewritten history.
  • The "reword" command lets you change a commit message without changing the commit's contents.
  • The "edit" command will cause Git to pause during the history rewrite at this commit. You can then modify it with git commit --amend or insert new commits.
  • The "squash" command will cause a commit to be folded into the previous one. You will be prompted to enter a message for the combined commit.
  • The "fixup" command works like squash, but discards the message of the commit being squashed instead of prompting for a new message.
  • Commits can be erased from history by deleting them from the list of commits
  • Commits can be re-ordered by re-ordering them in the list. When you are done modifying the list, Git will prompt you to resolve any resulting merge problems (after doing so, continue rebasing with git rebase --continue)

When you are done modifying the list, Git will perform the desired actions. If Git stops at a commit (due to merge conflicts caused by re-ordering the commits or due to the "edit" command), use git rebase --continue to resume. You can always back out of the rebase operation with git rebase --abort.

Warning: Only use git rebase -i on local commits that have not yet been pushed to anybody else. Modifying commits that are on the central server will cause merge problems for obvious reasons.
Note: Vim makes these rebase operations very simple since lines can be cut and pasted with few keystrokes.

Git Prompt

The Git package comes with a prompt script. To enable the prompt addition you will need to source the git-prompt.sh script and add $(__git_ps1 " (%s)") to your PS1 variable.

  • Add the following line to your ~/.bashrc/~/.zshrc:
source /usr/share/git/completion/git-prompt.sh
  • For Bash:
PS1='[\u@\h \W$(__git_ps1 " (%s)")]\$ '
Note: For information about coloring your Bash prompt, see Color_Bash_Prompt
Note: For information about coloring your Git prompt, see #Colors in Git Prompt
  • For zsh:
PS1='[%n@%m %c$(__git_ps1 " (%s)")]\$ '

The %s is replaced by the current branch name. Git information is displayed only if you are navigating in a Git repository. You can enable extra information by setting and exporting certain variables to a non-empty value as shown in the following table:

Variable Information
GIT_PS1_SHOWDIRTYSTATE * for unstaged and + for staged changes
GIT_PS1_SHOWSTASHSTATE $ if something is stashed
GIT_PS1_SHOWUNTRACKEDFILES % if there are untracked files

In addition you can set the GIT_PS1_SHOWUPSTREAM variable to "auto" in order to see < if you are behind upstream, > if you are ahead and <> if you have diverged.

Note: If you experience that $(__git_ps1) returns ((unknown)), then there's a .git folder in your current directory which doesn't contain any repository, and therefore Git does not recognize it. This can for example happen if you for some reason mistake Git's config-file to be ~/.git/config instead of ~/.gitconfig.

Transfer Protocols

Smart HTTP

Since version 1.6.6 git is able to use the HTTP(S) protocol as efficiently as SSH or Git by utilizing the git-http-backend. Furthermore it is not only possible to clone or pull from repositories, but also to push into repositories over HTTP(S).

The setup for this is rather simple as all you need to have installed is the Apache web server (with mod_cgi, mod_alias, and mod_env enabled) and of course, git:

# pacman -S apache git

Once you have your basic setup up and running, add the following to your Apache's config usually located at /etc/httpd/conf/httpd.conf:

<Directory "/usr/lib/git-core*">
    Order allow,deny
    Allow from all

SetEnv GIT_PROJECT_ROOT /srv/git
ScriptAlias /git/ /usr/lib/git-core/git-http-backend/

The above example config assumes that your git repositories are located at /srv/git and that you want to access them via something like http(s)://your_address.tld/git/your_repo.git. Feel free to customize this to your needs.

Note: Of course you have to make sure that your Apache can read and write (if you want to enable push access) on your git repositories.

For more detailed documentation, visit the following links:


You first need to have a public SSH key. For that follow the guide at Using SSH Keys. To set up SSH itself, you need to follow the SSH guide. This assumes you have a public SSH key now and that your SSH is working. Open your SSH key in your favorite editor (default public key name is ~/.ssh/id_rsa.pub), and copy its content (Ctrl+c). Now go to your user where you have made your Git repository, since we now need to allow that SSH key to log in on that user to access the Git repository. Open ~/.ssh/authorized_keys in your favorite editor, and paste the contents of id_rsa.pub in it. Be sure it is all on one line! That is important! It should look somewhat like this:

Warning: Do not copy the line below! It is an example! It will not work if you use that line!
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAAAgQCboOH6AotCh4OcwJgsB4AtXzDo9Gzhl+BAHuEvnDRHNSYIURqGN4CrP+b5Bx/iLrRFOBv58TcZz1jyJ2PaGwT74kvVOe9JCCdgw4nSMBV44cy+6cTJiv6f1tw8pHRS2H6nHC9SCSAWkMX4rpiSQ0wkhjug+GtBWOXDaotIzrFwLw== username@hostname

Now you can checkout your Git repository this way (change where needed. Here it is using the git username and localhost):

git clone git@localhost:my_repository.git

You should now get an SSH yes/no question. Type yes followed by Enter. Then you should have your repository checked out. Because this is with SSH, you also do have commit rights now. For that look at Git and Super Quick Git Guide.

Specifying a non-standard port

Connecting on a port other than 22 can be configured on a per-host basis in /etc/ssh/ssh_config or ~/.ssh/config. To set up ports for a repository, specify the path in .git/config using the port number N and the absolute path /PATH/TO/REPO:


Typically the repository resides in the home directory of the user which allows you to use tilde-expansion. Thus to connect on port N=443,

url = git@example.org:repo.git


url = ssh://git@example.org:443/~git/repo.git

Git daemon

Note: The git daemon only allows read access. For write access see #Git SSH.

This will allow URLs like "git clone git://localhost/my_repository.git".

Start git-daemon with root privileges.

# systemctl start git-daemon.socket

To run the git-daemon every time at boot, enable the service:

# systemctl enable git-daemon.socket

The daemon is started with the following parameters which are placed in the systemd service file.

ExecStart=-/usr/lib/git-core/git-daemon --inetd --export-all --base-path=/srv/git

So you have to place your repositories in /srv/git/ to be able to clone them with git daemon.

Clients can now simply use:

git clone git://localhost/my_repository.git

Git repositories rights

To restrict read/write access, you can simply use Unix rights, see http://sitaramc.github.com/gitolite/doc/overkill.htmlTemplate:Linkrot

For a fine-grained rights access, see gitolite and gitosis

See also