Git is a source code version control system. Such a system is most useful when you work in a team, but even when you're working alone, it's a very useful tool to keep track of the changes you have made to your code.
In this class, you are required to use Git for doing your homework assignments (we call them labs). You will use Git not only for coding your labs, but also for retrieving the skeleton code I provide, submitting your code, and downloading solutions.
This tutorial covers not only the basic git operations that you need, but also the workflow between you, me, and the TAs--from my preparation of an assignment all the way to the grading of your submissions by the TAs. Even if you are already familiar with git, you may find the description of the workflow interesting.
Type "echo $EDITOR". If the shell does not respond with the name of your editor--vim, emacs, or nano--add the following line at the end of the .bashrc file in your home directory:
export EDITOR=your_choice_of_editor
Log out and log in again. Type "echo $EDITOR" to make sure that your modification to .bashrc has taken effect.
Tell git your name and email:
git config --global user.name "Your Full Name"
git config --global user.email [email protected]
git stores this information in ~/.gitconfig
Let's create a new directory, ~/tmp/test1, for our first git project.
cd
mkdir tmp
cd tmp
mkdir test1
cd test1
Put the directory under git revision control:
git init
If you type ll (I'll assume that ll is an alias for ls -alF), you will see that there is a .git directory. The git repository for the current directory is stashed in the .git directory.
Let's start our programming project. Write hello.c with your editor: #include <stdio.h> int main() { printf("%s\n", "hello world"); return 0; }
Compile and run it:
gcc hello.c
./a.out
Let's see what git thinks about what we're doing:
git status
The git status command reports that hello.c and a.out are "Untracked". We can have git track hello.c by adding it to the "staging" area (more on this later):
git add hello.c
Run git status again. It now reports that hello.c is "a new file to be committed." Let's commit it:
git commit
Git opens up your editor for you to type a commit message. A commit message should succinctly describe what you're committing in the first line. If you have more to say, follow the first line with a blank line, and then with a more through multi-line description. For now, type in the following one-line commit message, save, and exit the editor.
Added hello-world program.
Run git status again. It now reports that only a.out is untracked. It has no mention of hello.c. When git says nothing about a file, it means that it is being tracked, and that it has not changed since it has been last committed.
We have successfully put our first coding project under git revision control.
Modify hello.c to print "bye world" instead, and run git status. It reports that the file is "Changed but not updated." This means that the file has been modified since the last commit, but it is still not ready to be committed because it has not been moved into the staging area. In git, a file must first go to the staging area before it can be committed.
Before we move it to the staging area, let's see what we changed in the file:
git diff
Or, if your terminal supports color,
git diff --color
The output should tell you that you took out the "hello world" line, and added a "bye world" line, like this:
- printf("%s\n", "hello world");
+ printf("%s\n", "bye world");
We move the file to the staging area with git add command:
git add hello.c
In git, "add" means this: move the change you made to the staging area. The change could be a modification to a tracked file, or it could be a creation of a brand new file. This is a point of confusion for those of you who are familiar with other version control systems such as Subversion.
At this point, "git diff" will report no change. Our change--from hello to bye--has been moved into staging already. So this means that "git diff" reports the difference between the staging area and the working copy of the file.
To see the difference between the last commit and the staging area, add "--cached" option:
git diff --cached
Let's commit our change. If your commit message is a one-liner, you can skip the editor by giving the message directly as part of the git commit command:
git commit -m "changed hello to bye"
To see your commit history:
git log
You can add a brief summary of what was done at each commit:
git log --stat --summary
Or you can see the full diff at each commit:
git log -p
And in color:
git log -p --color
A file in a directory under git revision control is either tracked or untracked. A tracked file can be unmodified, modified but unstaged, or modified and staged. Confused? Let's try again. There are four possibilities for a file in a git-controlled directory:
-
Untracked
Object files and executable files that can be rebuilt are usually not tracked.
-
Tracked, unmodified
The file is in the git repository, and it has not been modified since the last commit. "git status" says nothing about the file.
-
Tracked, modified, but unstaged
You modified the file, but didn't "git add" the file. The change has not been staged, so it's not ready for commit yet.
-
Tracked, modified, and staged
You modified the file, and did "git add" the file. The change has been moved to the staging area. It is ready for commit.
The staging area is also called the "index".
Here are some more git commands that you will find useful.
To rename a tracked file:
git mv old-filename new-filename
To remove a tracked file from the repository:
git rm filename
The mv or rm actions are automatically staged for you, but you still need to "git commit" your actions.
Sometimes you make some changes to a file, but regret it, and want to go back to the version last committed. If the file has not been staged yet, you can do:
git checkout -- filename
If the file has been staged, you must first unstage it:
git reset HEAD filename
There are two ways to display a manual page for a git command. For example, for the "git status" command, you can type one of the following two commands:
git help status
man git-status
Lastly, "git grep" searches for specified patterns in all files in the repository. To see all places you called printf():
git grep printf
You created a brand new project in the test1 directory, added a file, and modified the file. But more often than not, a programmer starts with an existing code base. When the code base is under git version control, you can clone the whole repository. This is in fact what you will do to start your lab assignments from my skeleton code. Let's move up one directory, clone test1 into test2, and cd into the test2 directory:
cd ..
git clone test1 test2
cd test2
Type ll to see that your hello.c file is cloned here. Moreover, if you run "git log", you will see that the whole commit history is replicated here. "git clone" not only copies the latest version of the files, but also copies the entire repository, including the entire commit history. After cloning, the two repositories are indistinguishable.
Let's make some changes--and let's be bad. Edit hello.c to replace "printf" with "printf%^&", save and commit:
vim hello.c
git add hello.c
git commit -m "hello world modification - work in progress"
Now run "git log" to see your recent commit carrying on the commit history that was cloned. If you want to see only the commits after cloning:
git log origin..
Of course you can add -p and --color to see the full diff in color:
git log -p --color origin..
Let's make one more modification. Fix the printf, and perhaps change the "bye world" to "rock my world" while we're there.
vim hello.c
git add hello.c
git commit -m "fixed typo & now prints rock my world"
Run "git log -p --color origin.." again to see the two commits you have made after cloning.
You can save the full details of everything you did after cloning with the following command:
git format-patch --stdout origin > mywork.mbox
If you open mywork.mbox with your editor (be careful not to modify the file), you will see that the file contains the full diffs of all commits you made after cloning. A diff is also called a "patch". The mywork.mbox file, therefore, contains a set of patches, one for each commit after cloning.
When you were looking at the mywork.mbox file in your editor, you might have noticed that each patch kinda looks like an email message. That's because they are. In fact, the mywork.mbox file is in the UNIX mailbox format (hence the extension .mbox), so you can view it using an email application.
You can use mutt, a terminal-based email app, to open this file as if it were your mailbox. Type 'q' to exit out of mutt.
mutt -f mywork.mbox
If you want to read the file with the diffs in color, run the following command. Use arrow keys to scroll and type 'q' to exit.
cat mywork.mbox | colordiff | less -R
This patch file is in fact what you submit when you complete your lab assignment. You will first clone my repository that contains some skeleton code, add and modify files to complete the lab, and then generate and submit a patch file that contains all your commits since you cloned my repository. I will supply a script that automates the patch generation and submission.
The TAs will apply your patch to my repository (the one you started with by cloning it) to recreate your repository at the time of your submission, and grade your work.
Let's go through that process. Move up one directory and clone test1 into test3:
cd ..
git clone test1 test3
cd test3
Run "git log" to verify that you have the commits made in test1, but not the ones made in test2.
Now let's bring in the commits made in test2 by applying the patch file that you generated in test2:
git am ../test2/mywork.mbox
You should see the following messages:
Applying: hello world modification - work in progress
Applying: fixed typo & now prints rock my world
Run "git log" to verify that now you have all the commits you made both in test1 and test2.
After the lab deadline, I will publish the solution by adding a "solution" subdirectory to my repository. Let's simulate that process. Go into the original test1 directory, and make the solution subdirectory and create two files in it:
cd ../test1
mkdir solution
cd solution
cp ../hello.c .
echo 'hello:' > Makefile
Type ll to see that two files--Makefile and hello.c--have been created in the solutions directory. (hello.c was copied from the parent directory, and Makefile was created directly on the command line using the echo command. BTW, the Makefile contains a single line, "hello:". Can you see why this is a legitimate Makefile?)
Now, move out of the solution directory, and git add & git commit the solution directory:
cd ..
git add solution
git commit -m "added solution"
Note that "git add solution" stages all files in the directory.
Once you hear that the lab solution is available, you will want to retrieve it and take a look. You do that by "pulling" the changes in my repository into your repository. Let's pull the changes we just made in test1 into test2:
cd ../test2/
git pull
The "git pull" command looks at the original repository that you cloned from, fetches all the changes made since the cloning, and merges the changes into the current repository. You now have the solution right in your repository.
This tutorial covers everything you need to do your lab assignments.
Git is an extremely powerful tool and a beautifully designed piece of software. If you want to learn more about it, start with the official git tutorial:
man gittutorial
There are pointers to further documentations at the end of the tutorial.
The documentation page of the Git web site has many links as well:
http://git-scm.com/documentation