A bunch of entertaining git tricks and reminders for myself.

Configure your user name

For the current repository only.

git config user.email 'max.mustermann@example.com'
git config user.name 'Max Mustermann'

Rename the author in all previous commits

When you start a project not caring about user names and emails, you can get into trouble once you want to publish it. :)

git filter-branch --commit-filter 'GIT_AUTHOR_NAME="Gaga" git-commit-tree "$@"'

Interesting variables to update:

• GIT_AUTHOR_NAME
• GIT_AUTHOR_EMAIL
• GIT_COMMITTER_NAME
• GIT_COMMITTER_EMAIL

You shouldn't do this when these commits are already otherwise published. (No, I don't know what will happen.)

Git with server-side repository

Make a bare-bones repo

On the server side, create a bare-bones repository (without working copy) containing your project.

Option A: make an empty bare-bones repo

cd
mkdir -p repos/proj.git
cd repos/proj.git
git init --bare

Option B: from a clone

cd
mkdir repos
cd repos
git clone --bare $ORIGINALREPO proj.git

Hook up client and server

Using SSH on the server

This is my strong preference for personal projects.

If you have a machine with a shared ssh account, you can also easily use that with the ssh protocol, e.g. ssh://username@hostname/full/path/to/repo.git or username@hostname:path/relative/to/homedir/repo.git. (Note: No ssh:// prefix when using relative paths.)

git remote add origin ssh://username@hostname/full/path/to/repo.git
git push  # works with an empty repo on server side.

Using git daemon on the server

Allow the repository to be exported by git daemon:

touch repos proj.git/git-daemon-export-ok

Run daemon and allow pushes.

git daemon --base-path=repos --enable=receive-pack

On the client side, clone the repository:

git clone git://$HOSTNAME/proj
cd proj
# Modify, commit
git push

Open source / Work workflow

This is big, so it's on a separate page.