Git is a distributed version control system developed by Linus Torvalds. It is used for managing Linux kernel development as well as for many other projects.
There is a variety of good information about Git available online, including:
Most commands have built-in documentation you can access with the --help option:
$ git init --help
You can generally access the same documentation as Unix man pages, e.g:
git init to create a git repository in your current directory:
$ mkdir myproject $ cd myproject $ git init
git init creates a git repository in your current working directory. You will add files to this repository using git add. This gives you a repository (the
.git directory) and a working copy (everything else).
If you are going to start tracking an existing project with git, you will generally start like this:
$ cd myproject $ git init
This initializes the repository. Next, add the existing files to the repository:
$ git add . $ git commit -m 'initial import'
git add command schedules files to be committed to the repository.
$ git add file1.c file2.c
Unlike Subversion, if you modify a file you (generally) need to git add that file in order to make the changes part of the next commit.
git reset command to "undo" an add operation:
$ git reset HEAD
This resets the index but leaves your working directory untouched. You can also use
git reset to revert to a previous commit; read the documentation for more information.
git commit to commit files to your local repository:
$ git commit
git commit by itself will commit any changes scheduled using
git add. If you would like to commit all locally modified files, use the
$ git commit -a
You may also commit a subset of modified files by specifying paths on the
$ git commit path/to/modified/file
git mv to rename files in the repository:
$ git mv oldname newname
git tracks files by cryptographic checksum, rather than by name, the
git mv command is not strictly necessary. If you manually rename a file and then do a
git rm file followed by a
git add file, git will correctly recognize that you have simply renamed it (because the checksum is still the same).
git rm to remove files from the repository:
$ git rm file1.c
Use git status to see a list of modified files:
$ git status
The output of
git status will look something like this:
$ git status # On branch master # Changed but not updated: # (use "git add <file>..." to update what will be committed) # (use "git checkout -- <file>..." to discard changes in working directory) # # modified: version-control.rst # # Untracked files: # (use "git add <file>..." to include in what will be committed) # # examples/ no changes added to commit (use "git add" and/or "git commit -a")
The files listed as "changed but not updated" are files that you have modified but not yet added to the repository. "Untracked files" are files that have not previously been added to the repository.
Use git diff to see pending changes in your working copy:
$ git diff
The output of git diff is standard diff output, e.g.:
$ git diff diff --git a/version-control.rst b/version-control.rst index e518192..b1c519a 100644 --- a/version-control.rst +++ b/version-control.rst @@ -243,6 +243,34 @@ commit`` to commit them to the (local) repository:: Using git: What's changed? ========================== +Use ``git status`` to see a list of modified files:: + + git status + +.. container:: handout + + The output will look something like this:: +
You can also use git diff to see the changes between arbitrary revisions of your project:
git diff <commit>
-Changes between two previous commits:
git diff <commit1> <commit2>
git clone command to check out a working copy of a remote repository:
$ git clone REPOSITORY [DIRECTORY]
git clone will clone the remote repository to a new directory in your current directory named after the repository, unless you explicitly provide a name with the DIRECTORY argument.
This is analogous to Subversion's checkout operation.
You can only clone the top-level repository; unlike Subversion, git does not allow you to clone individual subtrees.
git pull to update your local repository from the remote repository and merge changes into your working copy:
$ git pull [REPOSITORY [REFSPEC]]
git pull by itself will pull changes from the remote repository defined by the branch.master.remote config option (which will typically be the repository from which you originally cloned your working copy). If there are multiple remote repositories associated with your working copy, you can specify a repository (and branch) on the command line, e.g, to pull changes from the branch master at a remote named origin:
$ git pull origin master
git push to send your committed changes to a remote repository:
$ git push [REPOSITORY [REFSPEC]]
git push by itself will push your changes to the remote repository defined by the
branch.master.remote config option (which will typically be the repository from which you originally cloned your working copy). If there are multiple remote repositories associated with your working copy, you can specify a repository (and branch) on the command line, e.g, to push your changes to branch master at a remote named origin:
$ git push origin master
If you attempt to push to a repository that is newer than your working copy you will see an error similar to the following:
$ git push To dottiness.seas.harvard.edu:repos/myproject ! [rejected] master -> master (non-fast forward) error: failed to push some refs to 'dottiness.seas.harvard.edu:repos/myproject' To fix this, run git pull and deal with any conflicts.
To fix this, you need to pull from that repo, merge changes, and then push.
If you will be sharing a repository with others (or with yourself on multiple computers), you will need to create a "bare" repository – that is, a repository without a working copy. You do this with the
-b flag to
$ git init -b
You can then clone this repository, pull from it, and push to it as described in the previous section.
A conflict occurrs when two people make overlapping changes.
They are detected when you attempt to update your working copy via git pull.
You may discard your changes, discard the repository changes, or attempt to correct things manually. If you attempt to pull in changes that conflict with your working tree, you will see an error similar to the following:
$ git pull remote: Counting objects: 5, done. remote: Compressing objects: 100% (3/3), done. remote: Total 3 (delta 2), reused 0 (delta 0) Unpacking objects: 100% (3/3), done. From /Users/lars/projects/version-control-workshop/work/repo2 4245cb6..84f1112 master -> origin/master Auto-merging README CONFLICT (content): Merge conflict in README Automatic merge failed; fix conflicts and then commit the result.
To resolve the conflict manually:
To discard your changes (and accept the remote repository version):
git checkout --theirs README
To override the repository with your changes:
git checkout --ours README
When you complete the above tasks:
The git log command shows you the history of your repository:
$ git log [PATH]
git log with no arguments shows you the commit messages for each revision in your repository:
$ git log commit 7c8c3e71893d7481fdd9c13ec8f53cb9c61fac50 Author: Lars Kellogg-Stedman <email@example.com> Date: Thu Mar 18 12:46:46 2010 -0400 changed GNU to Microsoft commit 257f2f3ff44c2165c1182d3673a825fcadf121aa Author: Lars Kellogg-Stedman <firstname.lastname@example.org> Date: Thu Mar 18 12:46:46 2010 -0400 made a change commit 99c4fb8f37e48284d79c7396aaf755b514d6a249 Author: Lars Kellogg-Stedman <email@example.com> Date: Thu Mar 18 12:46:45 2010 -0400 made some changes commit 20cc63576f7c88541f5b9471e20f4d1c5f8afcb9 Author: Lars Kellogg-Stedman <firstname.lastname@example.org> Date: Thu Mar 18 12:46:45 2010 -0400 initial import
Git has explicit support for tagging and branching.
git tagmanipulates tags
git checkoutmanipulate branches
Create a tag:
$ git tag [-a] TAGNAME
Creates a lightweight tag (an alias for a commit object). Add
-a to create an annotated tag (i.e., with an associated message). It is also possible to create cryptographically signed tags.
To list available tags:
$ git tag
$ git branch
Create a new branch:
$ git checkout -b BRANCHNAME
To switch to a branch:
$ git checkout BRANCHNAME
For example, you want to enhance your code with some awesome experimental code. You create a new seas-workshop-dev branch and switch to it:
$ git checkout -b seas-workshop-dev
You make some changes, and when things are working you commit your branch:
$ git commit -m 'made some awesome changes' -a
And then merge it into the master branch:
$ git checkout master $ git merge seas-workshop-dev Updating 1288ed3..33e4a4c Fast-forward version-control.rst | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
Git is not really just like Subversion (or most other version control solutions). That's mainly because of the git "index".
git addadds files to the index
git commitcommits files from the index to the repository.
The diff commands uses the index:
git diffis the difference between your working copy and the index.
git diff HEADis the difference between your working copy and the local repository.
git diff --cachedis the difference between the index and the local repository.
Refer back to this illustration if you get confused:
(This image used with permission.)
Git can integrate with other version control systems.
You can use git as your Subversion client. This gives you many of the benefits of a DVCS while still interacting with a Subversion repository.
Cloning a remote repository:
git svn clone [ -s ] REPO_URL
-s flag informs git that your Subversion repository uses the recommended repository layout (i.e., that the top level of your repository contains trunk/, tags/, and branches/ directories). The HEAD of your working copy will track the trunk.
This instructs git to clone the entire repository, including the complete revision history. This may take a while for repositories with a long history. You can use the -r option to request a partial history. From the man page:
-r <ARG>, --revision <ARG> Used with the fetch command. This allows revision ranges for partial/cauterized history to be supported. $NUMBER, $NUMBER1:$NUMBER2 (numeric ranges), $NUMBER:HEAD, and BASE:$NUMBER are all supported. This can allow you to make partial mirrors when running fetch; but is generally not recommended because history will be skipped and lost.
Committing your changes back to the Subversion repository:
git svn dcommit
Before you push your changes to the Subversion repository you need to first commit any pending modifications to your local repository. Otherwise, git will complain:
$ git svn dcommit Cannot dcommit with a dirty index. Commit your changes first, or stash them with `git stash'. at /usr/libexec/git-core/git-svn line 491
To fix this, commit your changes:
$ git commit -m 'a meaningful commit message' -a
And then send your changes to the Subversion repository:
$ git svn dcommit Committing to https://source.seas.harvard.edu/svn/version-control-workshop/trunk ... M seealso.rst Committed r38 M seealso.rst r38 = 03254f2c0b3d5e068a87566caef84454558b85b0 (refs/remotes/trunk) No changes between current HEAD and refs/remotes/trunk Resetting to the latest refs/remotes/trunk Unstaged changes after reset: M git.rst M git.rst Committed r39 M git.rst r39 = d1f884a3f945f6083541e28ab7a09ca8efc6343b (refs/remotes/trunk) No changes between current HEAD and refs/remotes/trunk Resetting to the latest refs/remotes/trunk
Updating your working copy from the Subversion repository:
git svn rebase
As with git svn dcommit, you must have a clean working copy before running the rebase command.
You can import a CVS repository into git (this is a one-time, one-way operation). The CVS import feature requires
cvsps, a tool for collating CVS changes into changesets.
This may take a while:
$ export CVSHOME=:pserver:email@example.com $ cvs login $ git cvsimport -o cvs_head -C my-project