Quick GIT SCM Howto

This is just a quick rundown that I put here so I can refer to when I forget various things ;-). I hope it is of use to someone.

Creating your local repository

Git is distributed remember so all the history is copied to every repository. I am currently working on a distributed file system, it is called distfs and I will be using that in my examples.

In your project directory I am sure you will have some directory structure with all your working files in... often things like src/ and bin/.

git init

From the top project directory
mine is /home/ben/distfs

Now... you want to exclude all those binary files don't you? If not you are doing something wrong! For this we edit a file in the .git directory we just created.

cd .git
nano info/exclude

It may be better to write a .gitignore file and include the same information. A .gitignore file can be checked in and then will be taken with your project so others do not check in compiled files... which always annoys me.

My project is a C project and I want to ignore object files everywhere and all files in the bin/ directory so this is what my file looks like:

# git-ls-files --others --exclude-from=.git/info/exclude
# Lines that start with '#' are comments.
# For a project mostly in C, the following would be a good set of
# exclude patterns (uncomment them if you want to use them):
*.[oa]
*~
bin/*

Now you need to add your files to version control. You can add them all individually but because we want to add them all and we have written rules on which we would rather exclude it is good enough to just add them with a directory definition. Once you have done that you can check the files that will be under version control by running git status.

git add .
git status

Output for me is (notice no bin/ prefix):

# On branch master
#
# Initial commit
#
# Changes to be committed:
#   (use "git rm --cached <file>..." to unstage)
#
#       new file:   src/blockfunctions.c
#       new file:   src/blockfunctions.h
#       new file:   src/compile.sh
#       new file:   src/configfunctions.c
#       new file:   src/configfunctions.h
#       new file:   src/directoryfunctions.c
#       new file:   src/directoryfunctions.h
#       new file:   src/fstest.c
#       new file:   src/hello.c
#       new file:   src/sha256.c
#       new file:   src/sha256sum.c
#       new file:   src/sha512.c
#       new file:   testfs/.config
#

Not happy? use git rm and take out the files you don't want. You will probably have to force it (e.g. git rm -f testfs/.config), then you can check again with git status.

Once you are happy, commit the current set of files.

git commit -m "Initial Commit"

Creating the Remote Bare Repository

mkdir distfs.git
cd distfs.git
git init --bare

The .git is optional but seems to be convention

Now you have a bare git repository (that would probably be located on your server). Once done you should start the daemon, I usually start it from the place where all the other git repositories are but you don't have to.

git daemon --verbose --base-path=/home/git --enable=receive-pack

To be able to clone the repository with something like: git clone git://192.168.10.3/distfs.git you must export the repository. To do this make a file called git-daemon-export-ok in the distfs.git directory or add --export-all to your git daemon command. --enable=receive-pack is to allow the git push command later, try without and see what happens.

touch git-daemon-export-ok

git daemon --verbose --export-all --base-path=/home/git

Done... you are now able to push a project to your server using the native git protocol.

git push git://192.168.10.3/distfs.git HEAD

Push the project from the client machine

And clone/pull on any other machine:

git clone git://192.168.10.3/distfs.git

Clone the project from the server machine

Using SSH

You already have an ssh daemon running, why not use this to clone your repository? It is easy, using the example above just do a git push/clone once you have a bare repository on your server.

git push ssh://192.168.10.3/home/git/distfs
git clone ssh://192.168.10.3/home/git/distfs

Branching

Easy... to branch from current tree, first commit:

git commit -a -m "My Commit"

then:

git checkout -b mybranchname

or longhand:

git branch mybranchname
git checkout mybranchname

or to branch from a specific point in history (a previous commit):

git checkout -b mybranchname 342b192

where 342b192 is the start of the hash you want to branch from. Or longhand:

git branch mybranchname 342b192
git checkout mybranchname

To merge switch to master or whatever then:

git merge mybranchname

Don't care about a branch anymore? then delete it :-)

git branch -d mybranchname

Remember that branches are not pushed automatically to the server and so you must add the branch to your config or append the name to git push.

References: 1 2 3: Git Branch man page.

Helpful Links

http://help.github.com/line-endings/

More information to come.

git ls-files
git help ignore
cat .git/info/exclude 
nano .git/info/exclude 
nano .git/info/exclude 
git status
git add nanoxml/src/
git status
git add model/src/com/cheesmo/nzb/model/impl/NanoXMLNzbParser.java
git add 
nzbclient/src/com/cheesmo/nzb/client/DefaultSSLSocketFactory.java
git add nzbclient/src/com/cheesmo/nzb/client/DownloadSegment.java
git add nzbclient/src/com/cheesmo/nzb/client/SegmentQueue.java
git status
nano .git/info/exclude 
git status
git add nzbclient/README.txt
git status
man gitadd
git help add
git add -u
git status
git commit
git push
git diff

Techie Stuff