Source Code Management
Version Control System!
Revision Control System
! because we need to keep track of changes:
what was the content? !
who? (very important, right? :))
! centralized SCMs: Subversion, CVS
! distributed SCMs: GIT, Mercurial
Centralized vs. Distributed
master copy in one
everyone has a full
copy - distributed!
single point of failure!
works offline or
disconnected from a
considers data as a list
of file -based changes
(initial version + deltas) !
a commit operates over
the central repository
saves snapshots of data
it manages !
a commit operates over
the local repository
Client 1 !
British slang for "pig headed, think they are always
2005, Linux kernel maintenance (35000 files)!
written in C
"So I'm sorry, but for something like git, where efficiency was a primary !
objective, the "advantages" of C++ is just a huge mistake. The fact that !
we also piss off people who cannot see that is just a big additional !
modern, robust, fast, distributed, great branching
system, complex merges, open source!
available on Mac, Windows, Linux
a git repository is a....graph, Directed Acyclic Graph!
node = commit!
edge = pointers from child nodes to parent node(s)!
nodes are identified by SHA-1 hashes from: content change +
parent(s) commit identifier => speed!
a lot of the work done by git involves moving references
(labels) around => speed
git references = labels assigned to commits !
several types: branch (local/remote), tag, stash!
references are stored in .git/refs/heads ( the SHA-1
identifier of the commit it points to, 41 bytes)!
there are references to references -> HEAD, for example,
which points to the current branch!
references make commits reachable
git manages its content via 3 states: modified, staged,
Workspace (working directory)
one version of the content, extracted from git database to be
used or modified
Staging area (index)
allows you to elect files/content that is going to participate in
the next commit!
you can see how the next snapshot would look like before
actually creating it - avoid mistakes!
allows you to break changes in workspace into more than 1
where object database and metadata is stored!
the .git folder inside your repository root folder
git manages its content via 3 4 states: modified,
staged, committed, stashed!
here you can save changes you want to keep but
not ready to commit yet!
stashes are in fact labeled nodes on the
git repository located on another machine, usually!
the local repository may be a mirror of the remote one!
write into it via push, read from it via pull & fetch
blobs - contents of a file!
tree - directories of blobs or other trees!
commits - SHA-1 of a tree, parent commit(s), message!
Create a repository
initializes a new, empty repository, inside the current directory
git clone https://firstname.lastname@example.org/
clones the remote repository on the local machine, inside a
newly created git-scm-remote folder
Prepare a commit (stage)
git add <file> | <pattern>
marks the file(s) as belonging to the next commit, i.e. changes
git reset <file>
removes the file from the staging area
git commit -m "Commit message"
records the commit in the repository history
git commit --amend
modifies the last commit by adding the staged changes
git checkout -- <file>
the <file> contents in working directory is replaced with the
last committed one
git reset --hard <branch/commit/tag>
discards commits until the one pointed to by the reference
git revert <commit>
reverses the specified commit by creating a new one
saves the working directory state and pushes it on the stashes
git stash apply
gets the most recent stash and applies it to the working
git push <server> <branch>
pushes your changes on the specified branch of the remote
git format-patch HEAD --stdout > mypatch.patch
git fetch <server>
pulls down all the data from the remote repo that you don't
have yet on tracked branches
does a git pull and merges those on your current branch
git branch <branch name>
creates a new branch starting from your last commit
git checkout -b <branch name>
creates a new branch starting from your last commit and
switches to it
git branch -d <branch name>
deletes the specified branch
git merge <branch name>
merges the specified branch into the current one
git rebase <branch name>
rebases the specified branch onto the current one
GIT --@work : merge
GIT --a branching model
prints out the commit tree from the current HEAD
shows info like: current branch, working directory status,
git stash show
shows the most recent stash
shows where you are in repository history relatively to the latest
tag, i.e a revision number. Eg: v1.0.0-50-g1f8115b
git diff <commit>
shows the diff between working directory and specified commit
git branch [-r]
lists either local or remote (-r) branches
git tag v1.0.0 -m "My first release"
creates a "lightweight" tag, i.e. an immovable reference on the
git tag v1.0.0 -a -m "My first release"
creates an "annotated" tag, i.e. an immovable reference on the
current head + info like: tagger, email, date which are stored in
the git database
git cherry-pick <commit>
integrates changes in specified commit into the current branch
given a start and an end commit reference it will checkout
commits within that range allowing to find a regression
annotates each line in a file with that last commit to change it
.gitignore file to express ignore patterns, can be put in
git repos smaller than SVN ones, 30x in case of Mozilla!
steep learning curve (debatable)!
can't checkout part of a repository!
oriented towards content rather than files !
for transport it supports: HTTP, SSH, GIT protocols!
objects are generally added not deleted, reason for git gc
GIT --how to
latest release http:/
works from CLI!
good GUI clients:
Github for Mac/Win, free ("optimized" for github.com remotes,
but works with others too)!
SourceTree (Mac & Win), free!
(Em)Powered by GIT
local, flexible, fast
non-linear, facilitates collaboration