Git Usage

Also see Git and SVN combined workflow.

Which git repository to clone

The Gitlab Mirror

git@gitlab.com:tikiwiki/tiki.git
An automated mirror. This is what you should clone if you want to do pull requests on https://gitlab.com/tikiwiki/tiki

No need to ask anyone, but you DO need to upload you ssh key to https://gitlab.com/profile/keys

git.tiki.org

git@git.tiki.org:tiki.git
The main repository is git@git.tiki.org:tiki.git. You need to have your public ssh-rsa key registered to that server if you are going to merge the accepted pull requests from the Gitlab mirror. Ask luciash <luci@tiki.org> or Fabio <fabio.montefuscolo@gmail.com> if you need that direct access.

Getting the source

Shallow clone

Tiki development has been going on for nearly 20 years, and the code base is huge. For these two reasons the repository has a huge size and a regular clone will take a lot of time and disk space.

But it is possible to create a shallow clone, which will bring just the last source code version.

git clone --depth 1 --no-single-branch git@gitlab.com:tikiwiki/tiki.git tiki

This is quite fast, and uses about 292M of disk space. Probably best for all developers.
git branch -a works like most people expect and it's easy to switch branch.

If the purpose of the clone is Tiki development, including master and other branches, it is not needed to clone the entire repository, buy you my want to use --depth 100 so you have some useful history. (this will inflate the size to about 455M however.

git clone --depth 1 --branch master git@gitlab.com:tikiwiki/tiki.git tiki

The smallest option, but you can't use git branch -a like you normally would. instead you need to use git ls-remote.
To add another branch, you need to use the variant below.

This option is best to checkout on production servers. It uses about 187M of disk space

git clone --depth 1 --branch master git@gitlab.com:tikiwiki/tiki.git tiki
git remote set-branches --add origin '21.x'
git fetch --depth 1  #REALLY important on the first fetch, otherwise that whole branch will be fetched with all history.  That will use up 1.4G  and take a really long time.
#For normal operations you can just git fetch normally, it won't un-shallow the branch.

It now uses about 187M of disk space, which is barely more than a single branch. This option is best if you rely on expensive data connections and/or have very limited disk space.

Full clone

git clone git@gitlab.com:tikiwiki/tiki.git

This is quite slow, and uses about 3.1G of disk space. However subsequent operations are not any slower, and you have full local history. This is best if you have the time for the initial clone, and you need to work completely offline.

Full clone with folders per branch

git clone --reference

Full git repository is 3.1G. That's because the whole history is stored locally. Since Tiki setup depends on a lot of libraries, changing branches in a single repository might be impractical, and having one repository for each directory would consume a lot of redundant disk space.

To solve that, start by having a local mirror repository that will be used as reference for all other repositories. This way you can have several repositories that share the local object database. It's like doing a pre-cache of the repository:

git clone  --mirror git@gitlab.com:tikiwiki/tiki.git tiki-reference

(This will take some time, have a coffee).

Supposing a developer wants to clone also the branch 21.x, but without cloning the entire repository, the following command should be used.

git clone --branch=21.x --reference=tiki-reference git@gitlab.com:tikiwiki/tiki.git tiki21x

Repeat for other branches.

Now just commit and push normally, and everything should work fine.

Key issues

In practice, git is often based on ssh, which may be confusing if you've never used it before. Among other things, it will ask to accept the key of the remote server the first time you use it.

clone --mirror git@git.tiki.org:tiki.git tiki.reference
Cloning into bare repository 'tiki.reference'...
The authenticity of host 'git.tiki.org (172.104.234.244)' can't be established.
ECDSA key fingerprint is SHA256:bsX1VZ1KCXiUlequpZc6f+JGBG0xO4HpUSyYnVrXbMs.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'git.tiki.org,172.104.234.244' (ECDSA) to the list of known hosts.
git@git.tiki.org's password:

Repository status

git status

You can read about changed files, untracked files and how far is your local repository from the remote repository by using the git status command. More information can be found by typing git help status.

git status

Status with ignored files

git status --ignored

The command below will list all ignored files present on the working copy. This is useful when trying to commit a file that should be on repository, but it is ignored by a rule present in .gitignore file.

git status --ignored

Status of untracked files

git status -u

Sometimes, depending the task, it is possible that working copy contains a lot of untracked files. But it is possible to check working copy status excluding those files.

git status -u no

There are other occasions when there is a need to check all untracked files, including files on sub-directories.

git status -u

Reading history

To read the commit log, container authors, data and commit messages, you can use the command git log.

git log

Filtering history

git log --author --since

It is possible to filter logs by author and date, like the snippets below demonstrate.

git log --author=jonnybradley

# --author can be used multiple times
git log --author=jonnybradley --author=drsassafras

git log --since=2018-03-01

# --since is the start date, and --until is the end date
git log --since=2018-03-01 --until=2018-04-31

git log --since=2018-03-01 --until=2018-04-31 --author=jonnybradley --author=drsassafras

For more options, try git help log

Adding changes and files

git add

Before committing, the changes you did should be marked as added. So, when you tell git to commit your changes, only files you marked as added will be committed.

echo 'My new super file' > src/foo.txt
echo 'Other super file' > src/bar.txt
git add src/foo.txt src/bar.txt

It is possible to add all changes from a folder at once. It also works to add changes from entire working copy, but it is not a good practice.

echo 'My new super file' > src/foo.txt
echo 'Other super file' > src/bar.txt
git add src
git commit -m '[NEW] Add new super feature to Tiki'

Removing files from Git

git rm

git rm src/foo.txt
git commit -m '[FIX] Removed unused file'

Setting up global .gitignore

This from Stack Overflow:

git config --global core.excludesfile '~/.gitignore'

the to ignore all PhpStorm project settings, fo instance add this to ~.gitignore

/.idea/

Excluding personal files from a specific git repository

Sometimes you want git to ignore files on a specific local repository, not on all of your local repositories (where you would use a global .gitignore). But you also don't want to add those files into the repository .gitignore, because those are personal files (like files related to your code editor, for example). In that case, you can edit the .git/info/exclude file of your repository. This file works like a .gitignore file, but will not be committed to the repository.

See for more information: https://stackoverflow.com/a/22906950/751791

Committing changes

git commit

After all new or changed files are marked to be committed by using git add, or files were removed by using git rm, it is time to commit these changes to your local repository.

git status
git commit -m '[NEW] Descriptive message about my changes'

There are occasions a developer commits changes from someone else. In this case, it is possible to give the commit the correct author identity.

git commit --author=someoneelse@example.com  -m '[FIX] Options list for editable icon preference'

When there are changes just on versioned files, there is shortcut to commit these changes at once, by adding -a parameter on git commit.

git commit -a  -m '[NEW] Descriptive message about my changes'

Working with branches

Apart from master (trunk in SVN) Tiki has several branches. Some are permanent, used in parallel (for LTS and stable still supported Tiki versions), some are experimental and some are temporarily created by developers/contributors for merge request.
You can check the available branches by using the branch command;

git branch -r

On a development environment you may use several Tikis and several branches.
You can check on which branch you are by executing the branch command from inside a Tiki directory;

git branch

Creating branches

git branch

Creating branches is easy and fast on Git. It is a good idea to have different branches for tasks, features or experiments and keep master branch clean, so you can fetch updates regularly without dealing with conflicts.

# create the branch
git branch task-icon-picker
# and then switch to branch
git checkout task-icon-picker

There is a shortcut to create a branch and then switch to it.

git checkout -b task-icon-picker

If this local branch is needed for other team mates, it is possible to make it available on remote repository.

git push -u origin task-icon-picker

Let's suppose you are working on branch *bugfix-foo*, and you want update your branch with branch master changes.

git checkout bugfix-foo   # switch to branch
git merge master          # bring master changes to current branch

Switching branches

If you checked out a single shallow branch (e.g. 21.x) as described above you may get errors like this if you try, for instance git checkout origin/22.x:

Git: cannot checkout branch - error: pathspec '…' did not match any file(s) known to git

Then try this (note, i did something else and ended up checking out the whole 3.6GB tiki repo! )

git remote set-branches --add origin '22.x'
git fetch --depth 1

Sending work to remote repository

git push

To send your new commits from your local repository to the remote repository, just use the command git push. The command below will send all changes on local branch to the remote branch on remote repository.

git push

Fetching updates

git pull

It is a good practice to stay always up to date with remote repository your team is using. This way you avoid a lot of conflicts. To get remote changes and bring to local repository and working copy, use the command git pull.

The command below will fetch updates from remote repository.

git fetch

After fetching changes, the local repository and branch should be updated with it. Supposing the branch to be updated is the master, and the current local branch is also master, it is possible to apply changes with the command below.

git merge origin/master

There is a shortcut to fetch and apply changes on current branch, like below.

git pull

Backporting fixes to previous releases

See: Backport changes using Git

Git undo

See:
https://stackoverflow.com/questions/927358/how-do-i-undo-the-most-recent-local-commits-in-git