G.2. Git Tutorial

gittutorial - A tutorial introduction to Git

This tutorial explains how to import a new project into Git, make changes to it, and share changes with other developers.

If you are instead primarily interested in using Git to fetch a project, for example, to test the latest version, you may prefer to start with the first two chapters of The Git User's Manual.

First, note that you can get documentation for a command such as git log --graph with:

$ man git-log

or:

$ git help log

With the latter, you can use the manual viewer of your choice; see Section G.3.58, “git-help(1)” for more information.

It is a good idea to introduce yourself to Git with your name and public email address before doing any operation. The easiest way to do so is:

$ git config --global user.name "Your Name Comes Here"
$ git config --global user.email [email protected]

Assume you have a tarball project.tar.gz with your initial work. You can place it under Git revision control as follows.

$ tar xzf project.tar.gz
$ cd project
$ git init

Git will reply

Initialized empty Git repository in .git/

You've now initialized the working directory--you may notice a new directory created, named ".git".

Next, tell Git to take a snapshot of the contents of all files under the current directory (note the .), with git add:

$ git add .

This snapshot is now stored in a temporary staging area which Git calls the "index". You can permanently store the contents of the index in the repository with git commit:

$ git commit

This will prompt you for a commit message. You've now stored the first version of your project in Git.

Modify some files, then add their updated contents to the index:

$ git add file1 file2 file3

You are now ready to commit. You can see what is about to be committed using git diff with the --cached option:

$ git diff --cached

(Without --cached, git diff will show you any changes that you've made but not yet added to the index.) You can also get a brief summary of the situation with git status:

$ git status
On branch master
Changes to be committed:
Your branch is up-to-date with 'origin/master'.
  (use "git reset HEAD <file>..." to unstage)

        modified:   file1
        modified:   file2
        modified:   file3

If you need to make any further adjustments, do so now, and then add any newly modified content to the index. Finally, commit your changes with:

$ git commit

This will again prompt you for a message describing the change, and then record a new version of the project.

Alternatively, instead of running git add beforehand, you can use

$ git commit -a

which will automatically notice any modified (but not new) files, add them to the index, and commit, all in one step.

A note on commit messages: Though not required, it's a good idea to begin the commit message with a single short (less than 50 character) line summarizing the change, followed by a blank line and then a more thorough description. The text up to the first blank line in a commit message is treated as the commit title, and that title is used throughout Git. For example, Section G.3.50, “git-format-patch(1)” turns a commit into email, and it uses the title on the Subject line and the rest of the commit in the body.

Many revision control systems provide an add command that tells the system to start tracking changes to a new file. Git's add command does something simpler and more powerful: git add is used both for new and newly modified files, and in both cases it takes a snapshot of the given files and stages that content in the index, ready for inclusion in the next commit.

At any point you can view the history of your changes using

$ git log

If you also want to see complete diffs at each step, use

$ git log -p

Often the overview of the change is useful to get a feel of each step

$ git log --stat --summary

A single Git repository can maintain multiple branches of development. To create a new branch named "experimental", use

$ git branch experimental

If you now run

$ git branch

you'll get a list of all existing branches:

  experimental
* master

The "experimental" branch is the one you just created, and the "master" branch is a default branch that was created for you automatically. The asterisk marks the branch you are currently on; type

$ git checkout experimental

to switch to the experimental branch. Now edit a file, commit the change, and switch back to the master branch:

(edit file)
$ git commit -a
$ git checkout master

Check that the change you made is no longer visible, since it was made on the experimental branch and you're back on the master branch.

You can make a different change on the master branch:

(edit file)
$ git commit -a

at this point the two branches have diverged, with different changes made in each. To merge the changes made in experimental into master, run

$ git merge experimental

If the changes don't conflict, you're done. If there are conflicts, markers will be left in the problematic files showing the conflict;

$ git diff

will show this. Once you've edited the files to resolve the conflicts,

$ git commit -a

will commit the result of the merge. Finally,

$ gitk

will show a nice graphical representation of the resulting history.

At this point you could delete the experimental branch with

$ git branch -d experimental

This command ensures that the changes in the experimental branch are already in the current branch.

If you develop on a branch crazy-idea, then regret it, you can always delete the branch with

$ git branch -D crazy-idea

Branches are cheap and easy, so this is a good way to try something out.

Suppose that Alice has started a new project with a Git repository in /home/alice/project, and that Bob, who has a home directory on the same machine, wants to contribute.

Bob begins with:

bob$ git clone /home/alice/project myrepo

This creates a new directory "myrepo" containing a clone of Alice's repository. The clone is on an equal footing with the original project, possessing its own copy of the original project's history.

Bob then makes some changes and commits them:

(edit files)
bob$ git commit -a
(repeat as necessary)

When he's ready, he tells Alice to pull changes from the repository at /home/bob/myrepo. She does this with:

alice$ cd /home/alice/project
alice$ git pull /home/bob/myrepo master

This merges the changes from Bob's "master" branch into Alice's current branch. If Alice has made her own changes in the meantime, then she may need to manually fix any conflicts.

The "pull" command thus performs two operations: it fetches changes from a remote branch, then merges them into the current branch.

Note that in general, Alice would want her local changes committed before initiating this "pull". If Bob's work conflicts with what Alice did since their histories forked, Alice will use her working tree and the index to resolve conflicts, and existing local changes will interfere with the conflict resolution process (Git will still perform the fetch but will refuse to merge --- Alice will have to get rid of her local changes in some way and pull again when this happens).

Alice can peek at what Bob did without merging first, using the "fetch" command; this allows Alice to inspect what Bob did, using a special symbol "FETCH_HEAD", in order to determine if he has anything worth pulling, like this:

alice$ git fetch /home/bob/myrepo master
alice$ git log -p HEAD..FETCH_HEAD

This operation is safe even if Alice has uncommitted local changes. The range notation "HEAD..FETCH_HEAD" means "show everything that is reachable from the FETCH_HEAD but exclude anything that is reachable from HEAD". Alice already knows everything that leads to her current state (HEAD), and reviews what Bob has in his state (FETCH_HEAD) that she has not seen with this command.

If Alice wants to visualize what Bob did since their histories forked she can issue the following command:

$ gitk HEAD..FETCH_HEAD

This uses the same two-dot range notation we saw earlier with git log.

Alice may want to view what both of them did since they forked. She can use three-dot form instead of the two-dot form:

$ gitk HEAD...FETCH_HEAD

This means "show everything that is reachable from either one, but exclude anything that is reachable from both of them".

Please note that these range notation can be used with both gitk and "git log".

After inspecting what Bob did, if there is nothing urgent, Alice may decide to continue working without pulling from Bob. If Bob's history does have something Alice would immediately need, Alice may choose to stash her work-in-progress first, do a "pull", and then finally unstash her work-in-progress on top of the resulting history.

When you are working in a small closely knit group, it is not unusual to interact with the same repository over and over again. By defining remote repository shorthand, you can make it easier:

alice$ git remote add bob /home/bob/myrepo

With this, Alice can perform the first part of the "pull" operation alone using the git fetch command without merging them with her own branch, using:

alice$ git fetch bob

Unlike the longhand form, when Alice fetches from Bob using a remote repository shorthand set up with git remote, what was fetched is stored in a remote-tracking branch, in this case bob/master. So after this:

alice$ git log -p master..bob/master

shows a list of all the changes that Bob made since he branched from Alice's master branch.

After examining those changes, Alice could merge the changes into her master branch:

alice$ git merge bob/master

This merge can also be done by pulling from her own remote-tracking branch, like this:

alice$ git pull . remotes/bob/master

Note that git pull always merges into the current branch, regardless of what else is given on the command line.

Later, Bob can update his repo with Alice's latest changes using

bob$ git pull

Note that he doesn't need to give the path to Alice's repository; when Bob cloned Alice's repository, Git stored the location of her repository in the repository configuration, and that location is used for pulls:

bob$ git config --get remote.origin.url
/home/alice/project

(The complete configuration created by git clone is visible using git config -l, and the Section G.3.27, “git-config(1)” man page explains the meaning of each option.)

Git also keeps a pristine copy of Alice's master branch under the name "origin/master":

bob$ git branch -r
  origin/master

If Bob later decides to work from a different host, he can still perform clones and pulls using the ssh protocol:

bob$ git clone alice.org:/home/alice/project myrepo

Alternatively, Git has a native protocol, or can use http; see Section G.3.95, “git-pull(1)” for details.

Git can also be used in a CVS-like mode, with a central repository that various users push changes to; see Section G.3.96, “git-push(1)” and Section G.2.4, “gitcvs-migration(7)”.

Git history is represented as a series of interrelated commits. We have already seen that the git log command can list those commits. Note that first line of each git log entry also gives a name for the commit:

$ git log
commit c82a22c39cbc32576f64f5c6b3f24b99ea8149c7
Author: Junio C Hamano <[email protected]>
Date:   Tue May 16 17:18:22 2006 -0700

    merge-base: Clarify the comments on post processing.

We can give this name to git show to see the details about this commit.

$ git show c82a22c39cbc32576f64f5c6b3f24b99ea8149c7

But there are other ways to refer to commits. You can use any initial part of the name that is long enough to uniquely identify the commit:

$ git show c82a22c39c   # the first few characters of the name are
                        # usually enough
$ git show HEAD         # the tip of the current branch
$ git show experimental # the tip of the "experimental" branch

Every commit usually has one "parent" commit which points to the previous state of the project:

$ git show HEAD^  # to see the parent of HEAD
$ git show HEAD^^ # to see the grandparent of HEAD
$ git show HEAD~4 # to see the great-great grandparent of HEAD

Note that merge commits may have more than one parent:

$ git show HEAD^1 # show the first parent of HEAD (same as HEAD^)
$ git show HEAD^2 # show the second parent of HEAD

You can also give commits names of your own; after running

$ git tag v2.5 1b2e1d63ff

you can refer to 1b2e1d63ff by the name "v2.5". If you intend to share this name with other people (for example, to identify a release version), you should create a "tag" object, and perhaps sign it; see Section G.3.134, “git-tag(1)” for details.

Any Git command that needs to know a commit can take any of these names. For example:

$ git diff v2.5 HEAD     # compare the current HEAD to v2.5
$ git branch stable v2.5 # start a new branch named "stable" based
                         # at v2.5
$ git reset --hard HEAD^ # reset your current branch and working
                         # directory to its state at HEAD^

Be careful with that last command: in addition to losing any changes in the working directory, it will also remove all later commits from this branch. If this branch is the only branch containing those commits, they will be lost. Also, don't use git reset on a publicly-visible branch that other developers pull from, as it will force needless merges on other developers to clean up the history. If you need to undo changes that you have pushed, use git revert instead.

The git grep command can search for strings in any version of your project, so

$ git grep "hello" v2.5

searches for all occurrences of "hello" in v2.5.

If you leave out the commit name, git grep will search any of the files it manages in your current directory. So

$ git grep "hello"

is a quick way to search just the files that are tracked by Git.

Many Git commands also take sets of commits, which can be specified in a number of ways. Here are some examples with git log:

$ git log v2.5..v2.6            # commits between v2.5 and v2.6
$ git log v2.5..                # commits since v2.5
$ git log --since="2 weeks ago" # commits from the last 2 weeks
$ git log v2.5.. Makefile       # commits since v2.5 which modify
                                # Makefile

You can also give git log a "range" of commits where the first is not necessarily an ancestor of the second; for example, if the tips of the branches "stable" and "master" diverged from a common commit some time ago, then

$ git log stable..master

will list commits made in the master branch but not in the stable branch, while

$ git log master..stable

will show the list of commits made on the stable branch but not the master branch.

The git log command has a weakness: it must present commits in a list. When the history has lines of development that diverged and then merged back together, the order in which git log presents those commits is meaningless.

Most projects with multiple contributors (such as the Linux kernel, or Git itself) have frequent merges, and gitk does a better job of visualizing their history. For example,

$ gitk --since="2 weeks ago" drivers/

allows you to browse any commits from the last 2 weeks of commits that modified files under the "drivers" directory. (Note: you can adjust gitk's fonts by holding down the control key while pressing "-" or "+".)

Finally, most commands that take filenames will optionally allow you to precede any filename by a commit, to specify a particular version of the file:

$ git diff v2.5:Makefile HEAD:Makefile.in

You can also use git show to see any such file:

$ git show v2.5:Makefile

This tutorial should be enough to perform basic distributed revision control for your projects. However, to fully understand the depth and power of Git you need to understand two simple ideas on which it is based:

Part two of this tutorial explains the object database, the index file, and a few other odds and ends that you'll need to make the most of Git. You can find it at Section G.2.2, “gittutorial-2(7)”.

If you don't want to continue with that right away, a few other digressions that may be interesting at this point are:

Section G.2.2, “gittutorial-2(7)”, Section G.2.4, “gitcvs-migration(7)”, Section G.2.3, “gitcore-tutorial(7)”, Section G.4.16, “gitglossary(7)”, Section G.3.58, “git-help(1)”, Section G.4.15, “gitworkflows(7)”, Section G.2.5, “giteveryday(7)”, The Git User's Manual

Part of the Section G.3.1, “git(1)” suite.

gittutorial-2 - A tutorial introduction to Git: part two

You should work through Section G.2.1, “gittutorial(7)” before reading this tutorial.

The goal of this tutorial is to introduce two fundamental pieces of Git's architecture--the object database and the index file--and to provide the reader with everything necessary to understand the rest of the Git documentation.

Let's start a new project and create a small amount of history:

$ mkdir test-project
$ cd test-project
$ git init
Initialized empty Git repository in .git/
$ echo 'hello world' > file.txt
$ git add .
$ git commit -a -m "initial commit"
[master (root-commit) 54196cc] initial commit
 1 file changed, 1 insertion(+)
 create mode 100644 file.txt
$ echo 'hello world!' >file.txt
$ git commit -a -m "add emphasis"
[master c4d59f3] add emphasis
 1 file changed, 1 insertion(+), 1 deletion(-)

What are the 7 digits of hex that Git responded to the commit with?

We saw in part one of the tutorial that commits have names like this. It turns out that every object in the Git history is stored under a 40-digit hex name. That name is the SHA-1 hash of the object's contents; among other things, this ensures that Git will never store the same data twice (since identical data is given an identical SHA-1 name), and that the contents of a Git object will never change (since that would change the object's name as well). The 7 char hex strings here are simply the abbreviation of such 40 character long strings. Abbreviations can be used everywhere where the 40 character strings can be used, so long as they are unambiguous.

It is expected that the content of the commit object you created while following the example above generates a different SHA-1 hash than the one shown above because the commit object records the time when it was created and the name of the person performing the commit.

We can ask Git about this particular object with the cat-file command. Don't copy the 40 hex digits from this example but use those from your own version. Note that you can shorten it to only a few characters to save yourself typing all 40 hex digits:

$ git cat-file -t 54196cc2
commit
$ git cat-file commit 54196cc2
tree 92b8b694ffb1675e5975148e1121810081dbdffe
author J. Bruce Fields <[email protected]> 1143414668 -0500
committer J. Bruce Fields <[email protected]> 1143414668 -0500

initial commit

A tree can refer to one or more "blob" objects, each corresponding to a file. In addition, a tree can also refer to other tree objects, thus creating a directory hierarchy. You can examine the contents of any tree using ls-tree (remember that a long enough initial portion of the SHA-1 will also work):

$ git ls-tree 92b8b694
100644 blob 3b18e512dba79e4c8300dd08aeb37f8e728b8dad    file.txt

Thus we see that this tree has one file in it. The SHA-1 hash is a reference to that file's data:

$ git cat-file -t 3b18e512
blob

A "blob" is just file data, which we can also examine with cat-file:

$ git cat-file blob 3b18e512
hello world

Note that this is the old file data; so the object that Git named in its response to the initial tree was a tree with a snapshot of the directory state that was recorded by the first commit.

All of these objects are stored under their SHA-1 names inside the Git directory:

$ find .git/objects/
.git/objects/
.git/objects/pack
.git/objects/info
.git/objects/3b
.git/objects/3b/18e512dba79e4c8300dd08aeb37f8e728b8dad
.git/objects/92
.git/objects/92/b8b694ffb1675e5975148e1121810081dbdffe
.git/objects/54
.git/objects/54/196cc2703dc165cbd373a65a4dcf22d50ae7f7
.git/objects/a0
.git/objects/a0/423896973644771497bdc03eb99d5281615b51
.git/objects/d0
.git/objects/d0/492b368b66bdabf2ac1fd8c92b39d3db916e59
.git/objects/c4
.git/objects/c4/d59f390b9cfd4318117afde11d601c1085f241

and the contents of these files is just the compressed data plus a header identifying their length and their type. The type is either a blob, a tree, a commit, or a tag.

The simplest commit to find is the HEAD commit, which we can find from .git/HEAD:

$ cat .git/HEAD
ref: refs/heads/master

As you can see, this tells us which branch we're currently on, and it tells us this by naming a file under the .git directory, which itself contains a SHA-1 name referring to a commit object, which we can examine with cat-file:

$ cat .git/refs/heads/master
c4d59f390b9cfd4318117afde11d601c1085f241
$ git cat-file -t c4d59f39
commit
$ git cat-file commit c4d59f39
tree d0492b368b66bdabf2ac1fd8c92b39d3db916e59
parent 54196cc2703dc165cbd373a65a4dcf22d50ae7f7
author J. Bruce Fields <[email protected]> 1143418702 -0500
committer J. Bruce Fields <[email protected]> 1143418702 -0500

add emphasis

The "tree" object here refers to the new state of the tree:

$ git ls-tree d0492b36
100644 blob a0423896973644771497bdc03eb99d5281615b51    file.txt
$ git cat-file blob a0423896
hello world!

and the "parent" object refers to the previous commit:

$ git cat-file commit 54196cc2
tree 92b8b694ffb1675e5975148e1121810081dbdffe
author J. Bruce Fields <[email protected]> 1143414668 -0500
committer J. Bruce Fields <[email protected]> 1143414668 -0500

initial commit

The tree object is the tree we examined first, and this commit is unusual in that it lacks any parent.

Most commits have only one parent, but it is also common for a commit to have multiple parents. In that case the commit represents a merge, with the parent references pointing to the heads of the merged branches.

Besides blobs, trees, and commits, the only remaining type of object is a "tag", which we won't discuss here; refer to Section G.3.134, “git-tag(1)” for details.

So now we know how Git uses the object database to represent a project's history:

Note, by the way, that lots of commands take a tree as an argument. But as we can see above, a tree can be referred to in many different ways--by the SHA-1 name for that tree, by the name of a commit that refers to the tree, by the name of a branch whose head refers to that tree, etc.--and most such commands can accept any of these names.

In command synopses, the word "tree-ish" is sometimes used to designate such an argument.

The primary tool we've been using to create commits is git-commit -a, which creates a commit including every change you've made to your working tree. But what if you want to commit changes only to certain files? Or only certain changes to certain files?

If we look at the way commits are created under the cover, we'll see that there are more flexible ways creating commits.

Continuing with our test-project, let's modify file.txt again:

$ echo "hello world, again" >>file.txt

but this time instead of immediately making the commit, let's take an intermediate step, and ask for diffs along the way to keep track of what's happening:

$ git diff
--- a/file.txt
+++ b/file.txt
@@ -1 +1,2 @@
 hello world!
+hello world, again
$ git add file.txt
$ git diff

The last diff is empty, but no new commits have been made, and the head still doesn't contain the new line:

$ git diff HEAD
diff --git a/file.txt b/file.txt
index a042389..513feba 100644
--- a/file.txt
+++ b/file.txt
@@ -1 +1,2 @@
 hello world!
+hello world, again

So git diff is comparing against something other than the head. The thing that it's comparing against is actually the index file, which is stored in .git/index in a binary format, but whose contents we can examine with ls-files:

$ git ls-files --stage
100644 513feba2e53ebbd2532419ded848ba19de88ba00 0       file.txt
$ git cat-file -t 513feba2
blob
$ git cat-file blob 513feba2
hello world!
hello world, again

So what our git add did was store a new blob and then put a reference to it in the index file. If we modify the file again, we'll see that the new modifications are reflected in the git diff output:

$ echo 'again?' >>file.txt
$ git diff
index 513feba..ba3da7b 100644
--- a/file.txt
+++ b/file.txt
@@ -1,2 +1,3 @@
 hello world!
 hello world, again
+again?

With the right arguments, git diff can also show us the difference between the working directory and the last commit, or between the index and the last commit:

$ git diff HEAD
diff --git a/file.txt b/file.txt
index a042389..ba3da7b 100644
--- a/file.txt
+++ b/file.txt
@@ -1 +1,3 @@
 hello world!
+hello world, again
+again?
$ git diff --cached
diff --git a/file.txt b/file.txt
index a042389..513feba 100644
--- a/file.txt
+++ b/file.txt
@@ -1 +1,2 @@
 hello world!
+hello world, again

At any time, we can create a new commit using git commit (without the "-a" option), and verify that the state committed only includes the changes stored in the index file, not the additional change that is still only in our working tree:

$ git commit -m "repeat"
$ git diff HEAD
diff --git a/file.txt b/file.txt
index 513feba..ba3da7b 100644
--- a/file.txt
+++ b/file.txt
@@ -1,2 +1,3 @@
 hello world!
 hello world, again
+again?

So by default git commit uses the index to create the commit, not the working tree; the "-a" option to commit tells it to first update the index with all changes in the working tree.

Finally, it's worth looking at the effect of git add on the index file:

$ echo "goodbye, world" >closing.txt
$ git add closing.txt

The effect of the git add was to add one entry to the index file:

$ git ls-files --stage
100644 8b9743b20d4b15be3955fc8d5cd2b09cd2336138 0       closing.txt
100644 513feba2e53ebbd2532419ded848ba19de88ba00 0       file.txt

And, as you can see with cat-file, this new entry refers to the current contents of the file:

$ git cat-file blob 8b9743b2
goodbye, world

The "status" command is a useful way to get a quick summary of the situation:

$ git status
On branch master
Changes to be committed:
  (use "git reset HEAD <file>..." to unstage)

        new file:   closing.txt

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:   file.txt

Since the current state of closing.txt is cached in the index file, it is listed as "Changes to be committed". Since file.txt has changes in the working directory that aren't reflected in the index, it is marked "changed but not updated". At this point, running "git commit" would create a commit that added closing.txt (with its new contents), but that didn't modify file.txt.

Also, note that a bare git diff shows the changes to file.txt, but not the addition of closing.txt, because the version of closing.txt in the index file is identical to the one in the working directory.

In addition to being the staging area for new commits, the index file is also populated from the object database when checking out a branch, and is used to hold the trees involved in a merge operation. See Section G.2.3, “gitcore-tutorial(7)” and the relevant man pages for details.

At this point you should know everything necessary to read the man pages for any of the git commands; one good place to start would be with the commands mentioned in Section G.2.5, “giteveryday(7)”. You should be able to find any unknown jargon in Section G.4.16, “gitglossary(7)”.

The Git User's Manual provides a more comprehensive introduction to Git.

Section G.2.4, “gitcvs-migration(7)” explains how to import a CVS repository into Git, and shows how to use Git in a CVS-like way.

For some interesting examples of Git use, see the howtos.

For Git developers, Section G.2.3, “gitcore-tutorial(7)” goes into detail on the lower-level Git mechanisms involved in, for example, creating a new commit.

Section G.2.1, “gittutorial(7)”, Section G.2.4, “gitcvs-migration(7)”, Section G.2.3, “gitcore-tutorial(7)”, Section G.4.16, “gitglossary(7)”, Section G.3.58, “git-help(1)”, Section G.2.5, “giteveryday(7)”, The Git User's Manual

Part of the Section G.3.1, “git(1)” suite.

gitcore-tutorial - A Git core tutorial for developers

git *

This tutorial explains how to use the "core" Git commands to set up and work with a Git repository.

If you just need to use Git as a revision control system you may prefer to start with "A Tutorial Introduction to Git" (Section G.2.1, “gittutorial(7)”) or the Git User Manual.

However, an understanding of these low-level tools can be helpful if you want to understand Git's internals.

The core Git is often called "plumbing", with the prettier user interfaces on top of it called "porcelain". You may not want to use the plumbing directly very often, but it can be good to know what the plumbing does for when the porcelain isn't flushing.

Back when this document was originally written, many porcelain commands were shell scripts. For simplicity, it still uses them as examples to illustrate how plumbing is fit together to form the porcelain commands. The source tree includes some of these scripts in contrib/examples/ for reference. Although these are not implemented as shell scripts anymore, the description of what the plumbing layer commands do is still valid.

Creating a new Git repository couldn't be easier: all Git repositories start out empty, and the only thing you need to do is find yourself a subdirectory that you want to use as a working tree - either an empty one for a totally new project, or an existing working tree that you want to import into Git.

For our first example, we're going to start a totally new repository from scratch, with no pre-existing files, and we'll call it git-tutorial. To start up, create a subdirectory for it, change into that subdirectory, and initialize the Git infrastructure with git init:

$ mkdir git-tutorial
$ cd git-tutorial
$ git init

to which Git will reply

Initialized empty Git repository in .git/

which is just Git's way of saying that you haven't been doing anything strange, and that it will have created a local .git directory setup for your new project. You will now have a .git directory, and you can inspect that with ls. For your new empty project, it should show you three entries, among other things:

In particular, the refs subdirectory will contain two other subdirectories, named heads and tags respectively. They do exactly what their names imply: they contain references to any number of different heads of development (aka branches), and to any tags that you have created to name specific versions in your repository.

One note: the special master head is the default branch, which is why the .git/HEAD file was created points to it even if it doesn't yet exist. Basically, the HEAD link is supposed to always point to the branch you are working on right now, and you always start out expecting to work on the master branch.

However, this is only a convention, and you can name your branches anything you want, and don't have to ever even have a master branch. A number of the Git tools will assume that .git/HEAD is valid, though.

You have now created your first Git repository. Of course, since it's empty, that's not very useful, so let's start populating it with data.

We'll keep this simple and stupid, so we'll start off with populating a few trivial files just to get a feel for it.

Start off with just creating any random files that you want to maintain in your Git repository. We'll start off with a few bad examples, just to get a feel for how this works:

$ echo "Hello World" >hello
$ echo "Silly example" >example

you have now created two files in your working tree (aka working directory), but to actually check in your hard work, you will have to go through two steps:

The first step is trivial: when you want to tell Git about any changes to your working tree, you use the git update-index program. That program normally just takes a list of filenames you want to update, but to avoid trivial mistakes, it refuses to add new entries to the index (or remove existing ones) unless you explicitly tell it that you're adding a new entry with the --add flag (or removing an entry with the --remove) flag.

So to populate the index with the two files you just created, you can do

$ git update-index --add hello example

and you have now told Git to track those two files.

In fact, as you did that, if you now look into your object directory, you'll notice that Git will have added two new objects to the object database. If you did exactly the steps above, you should now be able to do

$ ls .git/objects/??/*

and see two files:

.git/objects/55/7db03de997c86a4a028e1ebd3a1ceb225be238
.git/objects/f2/4c74a2e500f5ee1332c86b94199f52b1d1d962

which correspond with the objects with names of 557db... and f24c7... respectively.

If you want to, you can use git cat-file to look at those objects, but you'll have to use the object name, not the filename of the object:

$ git cat-file -t 557db03de997c86a4a028e1ebd3a1ceb225be238

where the -t tells git cat-file to tell you what the "type" of the object is. Git will tell you that you have a "blob" object (i.e., just a regular file), and you can see the contents with

$ git cat-file blob 557db03

which will print out "Hello World". The object 557db03 is nothing more than the contents of your file hello.

Anyway, as we mentioned previously, you normally never actually take a look at the objects themselves, and typing long 40-character hex names is not something you'd normally want to do. The above digression was just to show that git update-index did something magical, and actually saved away the contents of your files into the Git object database.

Updating the index did something else too: it created a .git/index file. This is the index that describes your current working tree, and something you should be very aware of. Again, you normally never worry about the index file itself, but you should be aware of the fact that you have not actually really "checked in" your files into Git so far, you've only told Git about them.

However, since Git knows about them, you can now start using some of the most basic Git commands to manipulate the files or look at their status.

In particular, let's not even check in the two files into Git yet, we'll start off by adding another line to hello first:

$ echo "It's a new day for git" >>hello

and you can now, since you told Git about the previous state of hello, ask Git what has changed in the tree compared to your old index, using the git diff-files command:

$ git diff-files

Oops. That wasn't very readable. It just spit out its own internal version of a diff, but that internal version really just tells you that it has noticed that "hello" has been modified, and that the old object contents it had have been replaced with something else.

To make it readable, we can tell git diff-files to output the differences as a patch, using the -p flag:

$ git diff-files -p
diff --git a/hello b/hello
index 557db03..263414f 100644
--- a/hello
+++ b/hello
@@ -1 +1,2 @@
 Hello World
+It's a new day for git

i.e. the diff of the change we caused by adding another line to hello.

In other words, git diff-files always shows us the difference between what is recorded in the index, and what is currently in the working tree. That's very useful.

A common shorthand for git diff-files -p is to just write git diff, which will do the same thing.

$ git diff
diff --git a/hello b/hello
index 557db03..263414f 100644
--- a/hello
+++ b/hello
@@ -1 +1,2 @@
 Hello World
+It's a new day for git

Now, we want to go to the next stage in Git, which is to take the files that Git knows about in the index, and commit them as a real tree. We do that in two phases: creating a tree object, and committing that tree object as a commit object together with an explanation of what the tree was all about, along with information of how we came to that state.

Creating a tree object is trivial, and is done with git write-tree. There are no options or other input: git write-tree will take the current index state, and write an object that describes that whole index. In other words, we're now tying together all the different filenames with their contents (and their permissions), and we're creating the equivalent of a Git "directory" object:

$ git write-tree

and this will just output the name of the resulting tree, in this case (if you have done exactly as I've described) it should be

8988da15d077d4829fc51d8544c097def6644dbb

which is another incomprehensible object name. Again, if you want to, you can use git cat-file -t 8988d... to see that this time the object is not a "blob" object, but a "tree" object (you can also use git cat-file to actually output the raw object contents, but you'll see mainly a binary mess, so that's less interesting).

However -- normally you'd never use git write-tree on its own, because normally you always commit a tree into a commit object using the git commit-tree command. In fact, it's easier to not actually use git write-tree on its own at all, but to just pass its result in as an argument to git commit-tree.

git commit-tree normally takes several arguments -- it wants to know what the parent of a commit was, but since this is the first commit ever in this new repository, and it has no parents, we only need to pass in the object name of the tree. However, git commit-tree also wants to get a commit message on its standard input, and it will write out the resulting object name for the commit to its standard output.

And this is where we create the .git/refs/heads/master file which is pointed at by HEAD. This file is supposed to contain the reference to the top-of-tree of the master branch, and since that's exactly what git commit-tree spits out, we can do this all with a sequence of simple shell commands:

$ tree=$(git write-tree)
$ commit=$(echo 'Initial commit' | git commit-tree $tree)
$ git update-ref HEAD $commit

In this case this creates a totally new commit that is not related to anything else. Normally you do this only once for a project ever, and all later commits will be parented on top of an earlier commit.

Again, normally you'd never actually do this by hand. There is a helpful script called git commit that will do all of this for you. So you could have just written git commit instead, and it would have done the above magic scripting for you.

Remember how we did the git update-index on file hello and then we changed hello afterward, and could compare the new state of hello with the state we saved in the index file?

Further, remember how I said that git write-tree writes the contents of the index file to the tree, and thus what we just committed was in fact the original contents of the file hello, not the new ones. We did that on purpose, to show the difference between the index state, and the state in the working tree, and how they don't have to match, even when we commit things.

As before, if we do git diff-files -p in our git-tutorial project, we'll still see the same difference we saw last time: the index file hasn't changed by the act of committing anything. However, now that we have committed something, we can also learn to use a new command: git diff-index.

Unlike git diff-files, which showed the difference between the index file and the working tree, git diff-index shows the differences between a committed tree and either the index file or the working tree. In other words, git diff-index wants a tree to be diffed against, and before we did the commit, we couldn't do that, because we didn't have anything to diff against.

But now we can do

$ git diff-index -p HEAD

(where -p has the same meaning as it did in git diff-files), and it will show us the same difference, but for a totally different reason. Now we're comparing the working tree not against the index file, but against the tree we just wrote. It just so happens that those two are obviously the same, so we get the same result.

Again, because this is a common operation, you can also just shorthand it with

$ git diff HEAD

which ends up doing the above for you.

In other words, git diff-index normally compares a tree against the working tree, but when given the --cached flag, it is told to instead compare against just the index cache contents, and ignore the current working tree state entirely. Since we just wrote the index file to HEAD, doing git diff-index --cached -p HEAD should thus return an empty set of differences, and that's exactly what it does.

However, our next step is to commit the change we did, and again, to understand what's going on, keep in mind the difference between "working tree contents", "index file" and "committed tree". We have changes in the working tree that we want to commit, and we always have to work through the index file, so the first thing we need to do is to update the index cache:

$ git update-index hello

(note how we didn't need the --add flag this time, since Git knew about the file already).

Note what happens to the different git diff-* versions here. After we've updated hello in the index, git diff-files -p now shows no differences, but git diff-index -p HEAD still does show that the current state is different from the state we committed. In fact, now git diff-index shows the same difference whether we use the --cached flag or not, since now the index is coherent with the working tree.

Now, since we've updated hello in the index, we can commit the new version. We could do it by writing the tree by hand again, and committing the tree (this time we'd have to use the -p HEAD flag to tell commit that the HEAD was the parent of the new commit, and that this wasn't an initial commit any more), but you've done that once already, so let's just use the helpful script this time:

$ git commit

which starts an editor for you to write the commit message and tells you a bit about what you have done.

Write whatever message you want, and all the lines that start with # will be pruned out, and the rest will be used as the commit message for the change. If you decide you don't want to commit anything after all at this point (you can continue to edit things and update the index), you can just leave an empty message. Otherwise git commit will commit the change for you.

You've now made your first real Git commit. And if you're interested in looking at what git commit really does, feel free to investigate: it's a few very simple shell scripts to generate the helpful (?) commit message headers, and a few one-liners that actually do the commit itself (git commit).

While creating changes is useful, it's even more useful if you can tell later what changed. The most useful command for this is another of the diff family, namely git diff-tree.

git diff-tree can be given two arbitrary trees, and it will tell you the differences between them. Perhaps even more commonly, though, you can give it just a single commit object, and it will figure out the parent of that commit itself, and show the difference directly. Thus, to get the same diff that we've already seen several times, we can now do

$ git diff-tree -p HEAD

(again, -p means to show the difference as a human-readable patch), and it will show what the last commit (in HEAD) actually changed.

More interestingly, you can also give git diff-tree the --pretty flag, which tells it to also show the commit message and author and date of the commit, and you can tell it to show a whole series of diffs. Alternatively, you can tell it to be "silent", and not show the diffs at all, but just show the actual commit message.

In fact, together with the git rev-list program (which generates a list of revisions), git diff-tree ends up being a veritable fount of changes. You can emulate git log, git log -p, etc. with a trivial script that pipes the output of git rev-list to git diff-tree --stdin, which was exactly how early versions of git log were implemented.

In Git, there are two kinds of tags, a "light" one, and an "annotated tag".

A "light" tag is technically nothing more than a branch, except we put it in the .git/refs/tags/ subdirectory instead of calling it a head. So the simplest form of tag involves nothing more than

$ git tag my-first-tag

which just writes the current HEAD into the .git/refs/tags/my-first-tag file, after which point you can then use this symbolic name for that particular state. You can, for example, do

$ git diff my-first-tag

to diff your current state against that tag which at this point will obviously be an empty diff, but if you continue to develop and commit stuff, you can use your tag as an "anchor-point" to see what has changed since you tagged it.

An "annotated tag" is actually a real Git object, and contains not only a pointer to the state you want to tag, but also a small tag name and message, along with optionally a PGP signature that says that yes, you really did that tag. You create these annotated tags with either the -a or -s flag to git tag:

$ git tag -s <tagname>

which will sign the current HEAD (but you can also give it another argument that specifies the thing to tag, e.g., you could have tagged the current mybranch point by using git tag <tagname> mybranch).

You normally only do signed tags for major releases or things like that, while the light-weight tags are useful for any marking you want to do -- any time you decide that you want to remember a certain point, just create a private tag for it, and you have a nice symbolic name for the state at that point.

Git repositories are normally totally self-sufficient and relocatable. Unlike CVS, for example, there is no separate notion of "repository" and "working tree". A Git repository normally is the working tree, with the local Git information hidden in the .git subdirectory. There is nothing else. What you see is what you got.

This has two implications:

Note that the second point is true even across machines. You can duplicate a remote Git repository with any regular copy mechanism, be it scp, rsync or wget.

When copying a remote repository, you'll want to at a minimum update the index cache when you do this, and especially with other peoples' repositories you often want to make sure that the index cache is in some known state (you don't know what they've done and not yet checked in), so usually you'll precede the git update-index with a

$ git read-tree --reset HEAD
$ git update-index --refresh

which will force a total index re-build from the tree pointed to by HEAD. It resets the index contents to HEAD, and then the git update-index makes sure to match up all index entries with the checked-out files. If the original repository had uncommitted changes in its working tree, git update-index --refresh notices them and tells you they need to be updated.

The above can also be written as simply

$ git reset

and in fact a lot of the common Git command combinations can be scripted with the git xyz interfaces. You can learn things by just looking at what the various git scripts do. For example, git reset used to be the above two lines implemented in git reset, but some things like git status and git commit are slightly more complex scripts around the basic Git commands.

Many (most?) public remote repositories will not contain any of the checked out files or even an index file, and will only contain the actual core Git files. Such a repository usually doesn't even have the .git subdirectory, but has all the Git files directly in the repository.

To create your own local live copy of such a "raw" Git repository, you'd first create your own subdirectory for the project, and then copy the raw repository contents into the .git directory. For example, to create your own copy of the Git repository, you'd do the following

$ mkdir my-git
$ cd my-git
$ rsync -rL rsync://rsync.kernel.org/pub/scm/git/git.git/ .git

followed by

$ git read-tree HEAD

to populate the index. However, now you have populated the index, and you have all the Git internal files, but you will notice that you don't actually have any of the working tree files to work on. To get those, you'd check them out with

$ git checkout-index -u -a

where the -u flag means that you want the checkout to keep the index up-to-date (so that you don't have to refresh it afterward), and the -a flag means "check out all files" (if you have a stale copy or an older version of a checked out tree you may also need to add the -f flag first, to tell git checkout-index to force overwriting of any old files).

Again, this can all be simplified with

$ git clone git://git.kernel.org/pub/scm/git/git.git/ my-git
$ cd my-git
$ git checkout

which will end up doing all of the above for you.

You have now successfully copied somebody else's (mine) remote repository, and checked it out.

Branches in Git are really nothing more than pointers into the Git object database from within the .git/refs/ subdirectory, and as we already discussed, the HEAD branch is nothing but a symlink to one of these object pointers.

You can at any time create a new branch by just picking an arbitrary point in the project history, and just writing the SHA-1 name of that object into a file under .git/refs/heads/. You can use any filename you want (and indeed, subdirectories), but the convention is that the "normal" branch is called master. That's just a convention, though, and nothing enforces it.

To show that as an example, let's go back to the git-tutorial repository we used earlier, and create a branch in it. You do that by simply just saying that you want to check out a new branch:

$ git checkout -b mybranch

will create a new branch based at the current HEAD position, and switch to it.

You can always just jump back to your original master branch by doing

$ git checkout master

(or any other branch-name, for that matter) and if you forget which branch you happen to be on, a simple

$ cat .git/HEAD

will tell you where it's pointing. To get the list of branches you have, you can say

$ git branch

which used to be nothing more than a simple script around ls .git/refs/heads. There will be an asterisk in front of the branch you are currently on.

Sometimes you may wish to create a new branch without actually checking it out and switching to it. If so, just use the command

$ git branch <branchname> [startingpoint]

which will simply create the branch, but will not do anything further. You can then later -- once you decide that you want to actually develop on that branch -- switch to that branch with a regular git checkout with the branchname as the argument.

One of the ideas of having a branch is that you do some (possibly experimental) work in it, and eventually merge it back to the main branch. So assuming you created the above mybranch that started out being the same as the original master branch, let's make sure we're in that branch, and do some work there.

$ git checkout mybranch
$ echo "Work, work, work" >>hello
$ git commit -m "Some work." -i hello

Here, we just added another line to hello, and we used a shorthand for doing both git update-index hello and git commit by just giving the filename directly to git commit, with an -i flag (it tells Git to include that file in addition to what you have done to the index file so far when making the commit). The -m flag is to give the commit log message from the command line.

Now, to make it a bit more interesting, let's assume that somebody else does some work in the original branch, and simulate that by going back to the master branch, and editing the same file differently there:

$ git checkout master

Here, take a moment to look at the contents of hello, and notice how they don't contain the work we just did in mybranch -- because that work hasn't happened in the master branch at all. Then do

$ echo "Play, play, play" >>hello
$ echo "Lots of fun" >>example
$ git commit -m "Some fun." -i hello example

since the master branch is obviously in a much better mood.

Now, you've got two branches, and you decide that you want to merge the work done. Before we do that, let's introduce a cool graphical tool that helps you view what's going on:

$ gitk --all

will show you graphically both of your branches (that's what the --all means: normally it will just show you your current HEAD) and their histories. You can also see exactly how they came to be from a common source.

Anyway, let's exit gitk (^Q or the File menu), and decide that we want to merge the work we did on the mybranch branch into the master branch (which is currently our HEAD too). To do that, there's a nice script called git merge, which wants to know which branches you want to resolve and what the merge is all about:

$ git merge -m "Merge work in mybranch" mybranch

where the first argument is going to be used as the commit message if the merge can be resolved automatically.

Now, in this case we've intentionally created a situation where the merge will need to be fixed up by hand, though, so Git will do as much of it as it can automatically (which in this case is just merge the example file, which had no differences in the mybranch branch), and say:

        Auto-merging hello
        CONFLICT (content): Merge conflict in hello
        Automatic merge failed; fix conflicts and then commit the result.

It tells you that it did an "Automatic merge", which failed due to conflicts in hello.

Not to worry. It left the (trivial) conflict in hello in the same form you should already be well used to if you've ever used CVS, so let's just open hello in our editor (whatever that may be), and fix it up somehow. I'd suggest just making it so that hello contains all four lines:

Hello World
It's a new day for git
Play, play, play
Work, work, work

and once you're happy with your manual merge, just do a

$ git commit -i hello

which will very loudly warn you that you're now committing a merge (which is correct, so never mind), and you can write a small merge message about your adventures in git merge-land.

After you're done, start up gitk --all to see graphically what the history looks like. Notice that mybranch still exists, and you can switch to it, and continue to work with it if you want to. The mybranch branch will not contain the merge, but next time you merge it from the master branch, Git will know how you merged it, so you'll not have to do that merge again.

Another useful tool, especially if you do not always work in X-Window environment, is git show-branch.

$ git show-branch --topo-order --more=1 master mybranch
* [master] Merge work in mybranch
 ! [mybranch] Some work.
--
-  [master] Merge work in mybranch
*+ [mybranch] Some work.
*  [master^] Some fun.

The first two lines indicate that it is showing the two branches with the titles of their top-of-the-tree commits, you are currently on master branch (notice the asterisk * character), and the first column for the later output lines is used to show commits contained in the master branch, and the second column for the mybranch branch. Three commits are shown along with their titles. All of them have non blank characters in the first column (* shows an ordinary commit on the current branch, - is a merge commit), which means they are now part of the master branch. Only the "Some work" commit has the plus + character in the second column, because mybranch has not been merged to incorporate these commits from the master branch. The string inside brackets before the commit log message is a short name you can use to name the commit. In the above example, master and mybranch are branch heads. master^ is the first parent of master branch head. Please see Section G.4.12, “gitrevisions(7)” if you want to see more complex cases.

Now, let's pretend you are the one who did all the work in mybranch, and the fruit of your hard work has finally been merged to the master branch. Let's go back to mybranch, and run git merge to get the "upstream changes" back to your branch.

$ git checkout mybranch
$ git merge -m "Merge upstream changes." master

This outputs something like this (the actual commit object names would be different)

Updating from ae3a2da... to a80b4aa....
Fast-forward (no commit created; -m option ignored)
 example | 1 +
 hello   | 1 +
 2 files changed, 2 insertions(+)

Because your branch did not contain anything more than what had already been merged into the master branch, the merge operation did not actually do a merge. Instead, it just updated the top of the tree of your branch to that of the master branch. This is often called fast-forward merge.

You can run gitk --all again to see how the commit ancestry looks like, or run show-branch, which tells you this.

$ git show-branch master mybranch
! [master] Merge work in mybranch
 * [mybranch] Merge work in mybranch
--
-- [master] Merge work in mybranch

It's usually much more common that you merge with somebody else than merging with your own branches, so it's worth pointing out that Git makes that very easy too, and in fact, it's not that different from doing a git merge. In fact, a remote merge ends up being nothing more than "fetch the work from a remote repository into a temporary tag" followed by a git merge.

Fetching from a remote repository is done by, unsurprisingly, git fetch:

$ git fetch <remote-repository>

One of the following transports can be used to name the repository to download from:

Once you fetch from the remote repository, you merge that with your current branch.

However -- it's such a common thing to fetch and then immediately merge, that it's called git pull, and you can simply do

$ git pull <remote-repository>

and optionally give a branch-name for the remote end as a second argument.

It is likely that you will be pulling from the same remote repository from time to time. As a short hand, you can store the remote repository URL in the local repository's config file like this:

$ git config remote.linus.url http://www.kernel.org/pub/scm/git/git.git/

and use the "linus" keyword with git pull instead of the full URL.

Examples.

the above are equivalent to:

We said this tutorial shows what plumbing does to help you cope with the porcelain that isn't flushing, but we so far did not talk about how the merge really works. If you are following this tutorial the first time, I'd suggest to skip to "Publishing your work" section and come back here later.

OK, still with me? To give us an example to look at, let's go back to the earlier repository with "hello" and "example" file, and bring ourselves back to the pre-merge state:

$ git show-branch --more=2 master mybranch
! [master] Merge work in mybranch
 * [mybranch] Merge work in mybranch
--
-- [master] Merge work in mybranch
+* [master^2] Some work.
+* [master^] Some fun.

Remember, before running git merge, our master head was at "Some fun." commit, while our mybranch head was at "Some work." commit.

$ git checkout mybranch
$ git reset --hard master^2
$ git checkout master
$ git reset --hard master^

After rewinding, the commit structure should look like this:

$ git show-branch
* [master] Some fun.
 ! [mybranch] Some work.
--
*  [master] Some fun.
 + [mybranch] Some work.
*+ [master^] Initial commit

Now we are ready to experiment with the merge by hand.

git merge command, when merging two branches, uses 3-way merge algorithm. First, it finds the common ancestor between them. The command it uses is git merge-base:

$ mb=$(git merge-base HEAD mybranch)

The command writes the commit object name of the common ancestor to the standard output, so we captured its output to a variable, because we will be using it in the next step. By the way, the common ancestor commit is the "Initial commit" commit in this case. You can tell it by:

$ git name-rev --name-only --tags $mb
my-first-tag

After finding out a common ancestor commit, the second step is this:

$ git read-tree -m -u $mb HEAD mybranch

This is the same git read-tree command we have already seen, but it takes three trees, unlike previous examples. This reads the contents of each tree into different stage in the index file (the first tree goes to stage 1, the second to stage 2, etc.). After reading three trees into three stages, the paths that are the same in all three stages are collapsed into stage 0. Also paths that are the same in two of three stages are collapsed into stage 0, taking the SHA-1 from either stage 2 or stage 3, whichever is different from stage 1 (i.e. only one side changed from the common ancestor).

After collapsing operation, paths that are different in three trees are left in non-zero stages. At this point, you can inspect the index file with this command:

$ git ls-files --stage
100644 7f8b141b65fdcee47321e399a2598a235a032422 0       example
100644 557db03de997c86a4a028e1ebd3a1ceb225be238 1       hello
100644 ba42a2a96e3027f3333e13ede4ccf4498c3ae942 2       hello
100644 cc44c73eb783565da5831b4d820c962954019b69 3       hello

In our example of only two files, we did not have unchanged files so only example resulted in collapsing. But in real-life large projects, when only a small number of files change in one commit, this collapsing tends to trivially merge most of the paths fairly quickly, leaving only a handful of real changes in non-zero stages.

To look at only non-zero stages, use --unmerged flag:

$ git ls-files --unmerged
100644 557db03de997c86a4a028e1ebd3a1ceb225be238 1       hello
100644 ba42a2a96e3027f3333e13ede4ccf4498c3ae942 2       hello
100644 cc44c73eb783565da5831b4d820c962954019b69 3       hello

The next step of merging is to merge these three versions of the file, using 3-way merge. This is done by giving git merge-one-file command as one of the arguments to git merge-index command:

$ git merge-index git-merge-one-file hello
Auto-merging hello
ERROR: Merge conflict in hello
fatal: merge program failed

git merge-one-file script is called with parameters to describe those three versions, and is responsible to leave the merge results in the working tree. It is a fairly straightforward shell script, and eventually calls merge program from RCS suite to perform a file-level 3-way merge. In this case, merge detects conflicts, and the merge result with conflict marks is left in the working tree.. This can be seen if you run ls-files --stage again at this point:

$ git ls-files --stage
100644 7f8b141b65fdcee47321e399a2598a235a032422 0       example
100644 557db03de997c86a4a028e1ebd3a1ceb225be238 1       hello
100644 ba42a2a96e3027f3333e13ede4ccf4498c3ae942 2       hello
100644 cc44c73eb783565da5831b4d820c962954019b69 3       hello

This is the state of the index file and the working file after git merge returns control back to you, leaving the conflicting merge for you to resolve. Notice that the path hello is still unmerged, and what you see with git diff at this point is differences since stage 2 (i.e. your version).

So, we can use somebody else's work from a remote repository, but how can you prepare a repository to let other people pull from it?

You do your real work in your working tree that has your primary repository hanging under it as its .git subdirectory. You could make that repository accessible remotely and ask people to pull from it, but in practice that is not the way things are usually done. A recommended way is to have a public repository, make it reachable by other people, and when the changes you made in your primary working tree are in good shape, update the public repository from it. This is often called pushing.

Publishing the changes from your local (private) repository to your remote (public) repository requires a write privilege on the remote machine. You need to have an SSH account there to run a single command, git-receive-pack.

First, you need to create an empty repository on the remote machine that will house your public repository. This empty repository will be populated and be kept up-to-date by pushing into it later. Obviously, this repository creation needs to be done only once.

Your private repository's Git directory is usually .git, but your public repository is often named after the project name, i.e. <project>.git. Let's create such a public repository for project my-git. After logging into the remote machine, create an empty directory:

$ mkdir my-git.git

Then, make that directory into a Git repository by running git init, but this time, since its name is not the usual .git, we do things slightly differently:

$ GIT_DIR=my-git.git git init

Make sure this directory is available for others you want your changes to be pulled via the transport of your choice. Also you need to make sure that you have the git-receive-pack program on the $PATH.

Your "public repository" is now ready to accept your changes. Come back to the machine you have your private repository. From there, run this command:

$ git push <public-host>:/path/to/my-git.git master

This synchronizes your public repository to match the named branch head (i.e. master in this case) and objects reachable from them in your current repository.

As a real example, this is how I update my public Git repository. Kernel.org mirror network takes care of the propagation to other publicly visible machines:

$ git push master.kernel.org:/pub/scm/git/git.git/

Earlier, we saw that one file under .git/objects/??/ directory is stored for each Git object you create. This representation is efficient to create atomically and safely, but not so convenient to transport over the network. Since Git objects are immutable once they are created, there is a way to optimize the storage by "packing them together". The command

$ git repack

will do it for you. If you followed the tutorial examples, you would have accumulated about 17 objects in .git/objects/??/ directories by now. git repack tells you how many objects it packed, and stores the packed file in .git/objects/pack directory.

If you are paranoid, running git verify-pack command would detect if you have a corrupt pack, but do not worry too much. Our programs are always perfect ;-).

Once you have packed objects, you do not need to leave the unpacked objects that are contained in the pack file anymore.

$ git prune-packed

would remove them for you.

You can try running find .git/objects -type f before and after you run git prune-packed if you are curious. Also git count-objects would tell you how many unpacked objects are in your repository and how much space they are consuming.

If you run git repack again at this point, it will say "Nothing new to pack.". Once you continue your development and accumulate the changes, running git repack again will create a new pack, that contains objects created since you packed your repository the last time. We recommend that you pack your project soon after the initial import (unless you are starting your project from scratch), and then run git repack every once in a while, depending on how active your project is.

When a repository is synchronized via git push and git pull objects packed in the source repository are usually stored unpacked in the destination. While this allows you to use different packing strategies on both ends, it also means you may need to repack both repositories every once in a while.

Although Git is a truly distributed system, it is often convenient to organize your project with an informal hierarchy of developers. Linux kernel development is run this way. There is a nice illustration (page 17, "Merges to Mainline") in http://www.xenotime.net/linux/mentor/linux-mentoring-2006.pdf[Randy Dunlap's presentation].

It should be stressed that this hierarchy is purely informal. There is nothing fundamental in Git that enforces the "chain of patch flow" this hierarchy implies. You do not have to pull from only one remote repository.

A recommended workflow for a "project lead" goes like this:

A recommended work cycle for a "subsystem maintainer" who works on that project and has an own "public repository" goes like this:

A recommended work cycle for an "individual developer" who does not have a "public" repository is somewhat different. It goes like this:

If you are coming from CVS background, the style of cooperation suggested in the previous section may be new to you. You do not have to worry. Git supports "shared public repository" style of cooperation you are probably more familiar with as well.

See Section G.2.4, “gitcvs-migration(7)” for the details.

It is likely that you will be working on more than one thing at a time. It is easy to manage those more-or-less independent tasks using branches with Git.

We have already seen how branches work previously, with "fun and work" example using two branches. The idea is the same if there are more than two branches. Let's say you started out from "master" head, and have some new code in the "master" branch, and two independent fixes in the "commit-fix" and "diff-fix" branches:

$ git show-branch
! [commit-fix] Fix commit message normalization.
 ! [diff-fix] Fix rename detection.
  * [master] Release candidate #1
---
 +  [diff-fix] Fix rename detection.
 +  [diff-fix~1] Better common substring algorithm.
+   [commit-fix] Fix commit message normalization.
  * [master] Release candidate #1
++* [diff-fix~2] Pretty-print messages.

Both fixes are tested well, and at this point, you want to merge in both of them. You could merge in diff-fix first and then commit-fix next, like this:

$ git merge -m "Merge fix in diff-fix" diff-fix
$ git merge -m "Merge fix in commit-fix" commit-fix

Which would result in:

$ git show-branch
! [commit-fix] Fix commit message normalization.
 ! [diff-fix] Fix rename detection.
  * [master] Merge fix in commit-fix
---
  - [master] Merge fix in commit-fix
+ * [commit-fix] Fix commit message normalization.
  - [master~1] Merge fix in diff-fix
 +* [diff-fix] Fix rename detection.
 +* [diff-fix~1] Better common substring algorithm.
  * [master~2] Release candidate #1
++* [master~3] Pretty-print messages.

However, there is no particular reason to merge in one branch first and the other next, when what you have are a set of truly independent changes (if the order mattered, then they are not independent by definition). You could instead merge those two branches into the current branch at once. First let's undo what we just did and start over. We would want to get the master branch before these two merges by resetting it to master~2:

$ git reset --hard master~2

You can make sure git show-branch matches the state before those two git merge you just did. Then, instead of running two git merge commands in a row, you would merge these two branch heads (this is known as making an Octopus):

$ git merge commit-fix diff-fix
$ git show-branch
! [commit-fix] Fix commit message normalization.
 ! [diff-fix] Fix rename detection.
  * [master] Octopus merge of branches 'diff-fix' and 'commit-fix'
---
  - [master] Octopus merge of branches 'diff-fix' and 'commit-fix'
+ * [commit-fix] Fix commit message normalization.
 +* [diff-fix] Fix rename detection.
 +* [diff-fix~1] Better common substring algorithm.
  * [master~1] Release candidate #1
++* [master~2] Pretty-print messages.

Note that you should not do Octopus because you can. An octopus is a valid thing to do and often makes it easier to view the commit history if you are merging more than two independent changes at the same time. However, if you have merge conflicts with any of the branches you are merging in and need to hand resolve, that is an indication that the development happened in those branches were not independent after all, and you should merge two at a time, documenting how you resolved the conflicts, and the reason why you preferred changes made in one side over the other. Otherwise it would make the project history harder to follow, not easier.

Section G.2.1, “gittutorial(7)”, Section G.2.2, “gittutorial-2(7)”, Section G.2.4, “gitcvs-migration(7)”, Section G.3.58, “git-help(1)”, Section G.2.5, “giteveryday(7)”, The Git User's Manual

Part of the Section G.3.1, “git(1)” suite.

gitcvs-migration - Git for CVS users

Git differs from CVS in that every working tree contains a repository with a full copy of the project history, and no repository is inherently more important than any other. However, you can emulate the CVS model by designating a single shared repository which people can synchronize with; this document explains how to do that.

Some basic familiarity with Git is required. Having gone through Section G.2.1, “gittutorial(7)” and Section G.4.16, “gitglossary(7)” should be sufficient.

Suppose a shared repository is set up in /pub/repo.git on the host foo.com. Then as an individual committer you can clone the shared repository over ssh with:

$ git clone foo.com:/pub/repo.git/ my-project
$ cd my-project

and hack away. The equivalent of cvs update is

$ git pull origin

which merges in any work that others might have done since the clone operation. If there are uncommitted changes in your working tree, commit them first before running git pull.

You can update the shared repository with your changes by first committing your changes, and then using the git push command:

$ git push origin master

to "push" those commits to the shared repository. If someone else has updated the repository more recently, git push, like cvs commit, will complain, in which case you must pull any changes before attempting the push again.

In the git push command above we specify the name of the remote branch to update (master). If we leave that out, git push tries to update any branches in the remote repository that have the same name as a branch in the local repository. So the last push can be done with either of:

$ git push origin
$ git push foo.com:/pub/project.git/

as long as the shared repository does not have any branches other than master.

We assume you have already created a Git repository for your project, possibly created from scratch or from a tarball (see Section G.2.1, “gittutorial(7)”), or imported from an already existing CVS repository (see the next section).

Assume your existing repo is at /home/alice/myproject. Create a new "bare" repository (a repository without a working tree) and fetch your project into it:

$ mkdir /pub/my-repo.git
$ cd /pub/my-repo.git
$ git --bare init --shared
$ git --bare fetch /home/alice/myproject master:master

Next, give every team member read/write access to this repository. One easy way to do this is to give all the team members ssh access to the machine where the repository is hosted. If you don't want to give them a full shell on the machine, there is a restricted shell which only allows users to do Git pushes and pulls; see Section G.3.121, “git-shell(1)”.

Put all the committers in the same group, and make the repository writable by that group:

$ chgrp -R $group /pub/my-repo.git

Make sure committers have a umask of at most 027, so that the directories they create are writable and searchable by other group members.

First, install version 2.1 or higher of cvsps from http://www.cobite.com/cvsps/ and make sure it is in your path. Then cd to a checked out CVS working directory of the project you are interested in and run Section G.3.34, “git-cvsimport(1)”:

$ git cvsimport -C <destination> <module>

This puts a Git archive of the named CVS module in the directory <destination>, which will be created if necessary.

The import checks out from CVS every revision of every file. Reportedly cvsimport can average some twenty revisions per second, so for a medium-sized project this should not take more than a couple of minutes. Larger projects or remote repositories may take longer.

The main trunk is stored in the Git branch named origin, and additional CVS branches are stored in Git branches with the same names. The most recent version of the main trunk is also left checked out on the master branch, so you can start adding your own changes right away.

The import is incremental, so if you call it again next month it will fetch any CVS updates that have been made in the meantime. For this to work, you must not modify the imported branches; instead, create new branches for your own changes, and merge in the imported branches as necessary.

If you want a shared repository, you will need to make a bare clone of the imported directory, as described above. Then treat the imported directory as another development clone for purposes of merging incremental imports.

Git allows you to specify scripts called "hooks" to be run at certain points. You can use these, for example, to send all commits to the shared repository to a mailing list. See Section G.4.6, “githooks(5)”.

You can enforce finer grained permissions using update hooks. See Controlling access to branches using update hooks.

It is also possible to provide true CVS access to a Git repository, so that developers can still use CVS; see Section G.3.35, “git-cvsserver(1)” for details.

CVS users are accustomed to giving a group of developers commit access to a common repository. As we've seen, this is also possible with Git. However, the distributed nature of Git allows other development models, and you may want to first consider whether one of them might be a better fit for your project.

For example, you can choose a single person to maintain the project's primary public repository. Other developers then clone this repository and each work in their own clone. When they have a series of changes that they're happy with, they ask the maintainer to pull from the branch containing the changes. The maintainer reviews their changes and pulls them into the primary repository, which other developers pull from as necessary to stay coordinated. The Linux kernel and other projects use variants of this model.

With a small group, developers may just pull changes from each other's repositories without the need for a central maintainer.

Section G.2.1, “gittutorial(7)”, Section G.2.2, “gittutorial-2(7)”, Section G.2.3, “gitcore-tutorial(7)”, Section G.4.16, “gitglossary(7)”, Section G.2.5, “giteveryday(7)”, The Git User's Manual

Part of the Section G.3.1, “git(1)” suite.

giteveryday - A useful minimum set of commands for Everyday Git

Everyday Git With 20 Commands Or So

Git users can broadly be grouped into four categories for the purposes of describing here a small set of useful command for everyday Git.

A standalone individual developer does not exchange patches with other people, and works alone in a single repository, using the following commands.

A developer working as a participant in a group project needs to learn how to communicate with others, and uses these commands in addition to the ones needed by a standalone developer.

A fairly central person acting as the integrator in a group project receives changes made by others, reviews and integrates them and publishes the result for others to use, using these commands in addition to the ones needed by participants.

This section can also be used by those who respond to git request-pull or pull-request on GitHub (www.github.com) to integrate the work of others into their history. An sub-area lieutenant for a repository will act both as a participant and as an integrator.

A repository administrator uses the following tools to set up and maintain access to the repository by developers.

update hook howto has a good example of managing a shared central repository.

In addition there are a number of other widely deployed hosting, browsing and reviewing solutions such as:

Part of the Section G.3.1, “git(1)” suite