Github

From 太極
Revision as of 18:37, 6 April 2016 by Brb (talk | contribs) (→‎Gitlab)
Jump to navigation Jump to search

Git

git config --global user.name "Your Name Here"
git config --global user.email [email protected]

OR to have a project specific git identity (this will write your identify in .git/config file) 

git config user.name "Your Name Here"
git config user.email [email protected]
git config --global credential.helper cache

My example of working on a new repository called 'network'.

  1. Follow https://help.github.com/articles/create-a-repo to create a new repository. For convenience, I also check the button to create README file.
  2. Click 'GitHub' icon on Windows Desktop. Look at the LHS and click on the word 'github'. Click 'clone' button (This can be accomplished by git clone https://github.com/arraytools/network.git). The new repository will appear under C:\Users\USERNAME\Documents\Github\ directory. Now Click 'Git Shell' icon on the Windows Desktop and go to C:\Users\USERNAME\Documents\Github\network directory where 'network' is my repository's name.
git config --global color.ui auto  # colorize the output of git
git init

git add client.c
git add server.c
git add server2.c
git commit -m 'first commit'

git status
# On branch master
# Your branch is ahead of 'origin/master' by 1 commit.

# git remote add origin https://github.com/arraytools/network.git
# git push <remote> <branch>
git push origin master

# If we modify any file, we need to go through git add, git commit and git push 3 commands.

# get commit_id
git log
# get a specific version
git checkout commit_id

# after an examination, we want to get the latest version
git pull origin master
# If you do not want to merge the upstream changes wit your local repository, run git fetch to 
# fetch the changes and then git merge to merge the changes. 
# git pull is just a combination of fetch and merge.

# To rename a file
git mv originafile newfile
git commit -m "rename orginalfile"
git push

# To remove a file
rm myfile
git add . -A
git commit -m "remove a file"
git push

# Revert to origin's master branch's version of file
# http://stackoverflow.com/questions/1817766/revert-to-origins-master-branchs-version-of-file
# 1. Assuming you did not commit the file, or add it to the index, then:
git checkout filename
# 2. Assuming you added it to the index, but did not commit it, then:
git reset HEAD filename
git checkout filename

Check the https://github.com/arraytools/network. The commit goes to the repository!

In summary: add and commit are local operations, push, pull and fetch are operations that interact with a remote.

If we want to checkout a specific commit on a new computer, we can use (here we use Qt repository as an example) 

git clone https://xxxxxx/xxxxxx/Qt.git
cd Qt       
git log --oneline
git checkout SHA1  (7 digits obtained from git log --oneline commandis enough)

After that we can run 

# move HEAD to origin
git checkout origin/master

# Visualize using text mode
git log --graph --oneline --date-order --decorate --color --all

Using GUI client: gitg 

sudo apt-get install gitg

To switch among different branch (eg. github project page is located in gh-pages branch of a repository), 

git checkout gh-pages
git checkout master

Tip: To avoid being asking the username each time when you use 'git push', you can modify the .git/config file of your local repository. This file contains a section called 'remote' with an entry called 'url'. The 'url' entry should contains the https link of repository you're talking about. When you prefix the host url with your username, git shouldn't be asking for your username anymore. Here's an example: url = https://[email protected]

GUI version of Git software

Windows

Go to http://git-scm.com/download. The Windows version contain 'Git Bash', 'Git CMD', and 'Git GUI'. The 'Git GUI' software (based on Tcl/Tk) works pretty cool. It can 'Rescan' the project, compare the changed filefs and visualize the master's history too. The Git comes with 2 built-in tools: Git-gui is for committing and gitk is for browsing. Screenshots of gitk can be found below.

Note that Git-gui cannot run git pull directly. We have to go through two steps: Fetch and merge. See this message.

Linux

Use gitk. For example, 

gitk-all

Set up a new local/remote repository

mkdir /path/to/your/project
cd /path/to/your/project
git init
git remote add origin https://[email protected]/arraytools/REPOSITORYNAME.git
# git remote add origin ssh://[email protected]/home/git/REPOSITORYNAME.git
# git clone             ssh://[email protected]/home/git/REPOSITORYNAME.git

git config --global user.name "YOUR NAME"
git config --global user.email "YOUR EMAIL ADDRESS"
git config --list # confirm

echo "arraytools" >> contributors.txt
git add contributors.txt
git commit -m 'Initial commit with contributors'
git push -u origin master

Already has a git repository on my computer

cd /path/to/my/repo
git remote add origin https://[email protected]/arraytools/REPOSITORYNAME.git
git push -u origin --all # pushes up the repo and its refs for the first time
git push -u origin --tags # pushes up any tags

'master' and 'origin'

  • master is a branch name. You can use git branch to find out all branches. The current branch has a asterisk in the command line output and has a bold font in the gitk program.
  • origin is a repository name. You are free to create a new one and delete origin especially in situation that you are working with multiple remotes.

https://lostechies.com/joshuaflanagan/2010/09/03/use-gitk-to-understand-git/

Setup editor

See man git-commit. 

git config --global core.editor "nano"
# OR
export GIT_EDITOR=nano
# OR for other programs to use too
export EDITOR=nano

.gitignore file

git help gitignore
# or
man gitignore

A leading slash indicates that the ignore entry is only to be valid with respect to the directory in which the .gitignore file resides. Specifying *.o would ignore all .o files in this directory and all subdirs, while /*.o would just ignore them in that dir, while again, /foo/*.o would only ignore them in /foo/*.o.

branch - not the same as in CVS

https://lostechies.com/joshuaflanagan/2010/09/03/use-gitk-to-understand-git/

Gitk shows all of the commits as a single straight line. In git, a branch is a label for a commit. The label moves to new commits as they are created. When you create a git branch, you are not changing anything in the structure of the repository or the source tree. You are just creating a new label.

Below are some info borrowed from Pro Git.

Create a new branch

git branch testing
git log --oneline --decorate

Pay attention to the keywords 'HEAD', 'master' and 'testing' in this case.

Switch branches

git checkout testing
nano test.rb
git commit -a -m 'made a change'
git checkout master
nano test.rb
git commit -a -m 'made other changes'
git log --oneline --decorate --graph --all

Basic merging

git checkout -b iss53 
# shorthand for 
# git branch iss53
# git checkout iss53

nano index.html
git commit -a -m 'added a new footer [issue 53]'
git checkout master

git checkout -b hotfix
nano index.html
git commit -a -m 'fixed the broken email address'

git checkout master
git merge hotfix
git branch -d hotfix

git checkout iss53
nano index.html
git commit -a -m 'finished the new footer [issue 53]'

Basic Merging: 

git checkout master
git merge iss53

Basic merge conflicts: 

git merge iss53
git status # Look at the standard conflict-resolution markers to the top of files

git mergetool
git status

Branch management

git branch
git branch -v          # see the last commit on each branch
git branch --merged    # Filter the list to branches that you have merged into the branch you're currently on
git branch --no-merged # See the branches that contain work you haven't yet merged in
git branch -d testing

Branch workflows

NA

Remote branches

Remote references are references (pointers) in your remote repositories, including branches, tags, and so on.

Pushing: 

git ls-remote (remoteName)
git remote show (remoteName)

git push origin serverfix   # serverfix is a branch name

Do not type your password every time: you can set up a credential cache. The simplest is to keep it in memory for a few minutes, which you can set up by running 

git config --global credential.helper cache

One collaborator fetches from the server. They will get a reference to where the server's version of serverfix is under the remote branch origin/serverfix: 

git fetch origin
git checkout -b serverfix origin/serverfix

Tracking Branches: 

git checkout --track origin/serverfix

git checkout -b sf origin/serverfix

git branch -u origin/serverfix

Pulling: 

git pull   
# Equivalent to two actions
git fetch origin  # get the contents of the remote repository (origin), but keep them under origin/branch branch
                  # requires the password  
git merge origin/master # merge the master branch of the remote repository (origin) with your current branch
                        # no password required

See git fetch and git-merge.

Deleting Remote Branches: 

git push origin --delete serverfix

Rebasing

discard unstaged changes in git

For a specific file use: 

git checkout path/to/file/to/revert

For all unstaged files use: 

git checkout -- .
git checkout -- path/to/foo  # for a specific file <foo>

Make sure to include the period at the end.


Detached head/HEAD detached from xxxxxxx

I got the above message when I run

  1. Gitgui -> checkout -> Revision Expression (some previous version)
  2. Gitgui -> checkout -> Revision Expression (the latest version)
  3. Git Bash -> git status

At this time, I cannot continue as usual to make changes to files and do commits.

The solution is run 

git checkout master

See this post from stackoverflow.com.

git merge

See an example of merging a temporary branch with the master branch in the local repository.

This is another example. It teaches how to make sure my master branch is in synch with the central repository on github (which I refer to using the remote “origin”) before I merge my changes into master.

git rebase

https://lostechies.com/joshuaflanagan/2010/09/03/use-gitk-to-understand-git-merge-and-rebase/

The simple git merge method can make the history very complicated; see and the following cons.

  • Branching paths in the history can be unnecessarily complicated
  • The extra merge commit.
  • Your branch is now no longer a private, local concern. Everyone now knows that you worked in an issue123 branch. Why should they care?

git rebase can be used to avoid these issues.

The good approach (as in that example) is 

git checkout issue123
git rebase master
git checkout master
git merge issue123
git branch -d issue123
git push origin master

Rename a repository

  1. Go to github.com, open the project and click Settings button on the left-bottom corner. Change the repository name on top.
  2. On local machine, rename the directory. Go to the directory. Issue
git remote -v

to get the ULR for the current working copy. Suppose the url is [email protected]:someuser/someproject.git. Now issue the following command to change to the new repository 

git remote set-url origin [email protected]:someuser/newprojectname.git

Create a git server (command line only)

Follow the instruction on git-scm.com. It works.

I tested it by

  1. create a git account (called 'git') on my host machine.
  2. sudo to create a directory called /opt/git/project.git. Change the owner to 'git'. Cd to the directory and initialize it.
  3. Create two virtual machines (vb1 and vb2). Creating a username 'david' on vb1 and a user name 'joseph' on vb2.
  4. Create ssh key for both 'david' and 'joseph'. Ssh to copy their ssh keys to git account.
  5. Create a new directory on vb1 and initialize it. Run git commands to commit & push files to the server (no password is needed). Note that when we use 'git commit', git will ask to create a username and email by first running 'git config' command.
  6. Switch to vb2 and run git clone (no password is needed). The user can modify the code and commit & push files to the server.
  7. Run 'git log' to check if each user's name/email are shown on the log.

Some important points:

  • There is no daemon to be installed. We only need to install the 'git' program on the server.
  • When a client uses the 'git' command to communicate with the server, it is actually using the 'ssh' to access the server.
  • We still need to create a user for this git server. Then the developers' rsa keys can be saved to the git user's <.ssh/authorized_keys> file.
  • The instruction asks to chmod 700 (rwx------) for the .ssh directory and chmod 600 (-rw------) for the <.ssh/authorized_keys> file.
  • If the git repository directory is not saved under the user's directory, we need to make sure the owner of the directory is the git user.

Another way to create a git server: gitolite (favored by Ubuntu)

This approach involves the following steps

  1. Installing a gitolite server
  2. Gitolite configuration
  3. Managing gitolite users and repositories
  4. Using your server

Not sure about any advantages of this approach?

Create a git server (github like w/ web interface)

If we like to create a github-like web interface, check out GitLab.

Below is my note

  1. https://about.gitlab.com/downloads/ contains steps of setting up Gitlab.
  2. By default, the domain name you have entered in setting up gitlab will be the URL you will use to access gitlab.
  3. Use the recommended method to install gitlab. Nginx will be installed as an http server.
  4. The root username and password is root and 5iveL!fe.
  5. When new users are created by root, we can put a faked email there (eg [email protected]). The root account can create password for the user.
  6. User's password is used to access GitLab web interface only. It is not used for pushing commits.
  7. After a new user is created, log out of root account and log in using the new user account. Click 'Profile setting' icon and then select SSH > Add SSH key. Copy your <id_rsa.pub> content there. To create your ssh key, use the command line "ssh-keygen -t rsa". The <id_rsa.pub> is located under ~/.ssh directory. The title should be auto populated. If ssh key is added successfully to gitLab, we won't get a pop-up asking password when we run 'git push'.
  8. A new project should be created by users (not root). If I create a project by root, I keep getting a permission issue when I run 'git push'.
  9. The username will affect path to all personal projects; e.g. [email protected]:newuser/test2.git.

Gitlab2.png Gitlab1.png Gitlab3.png

How to Install Gitlab with PostgreSQL and Nginx on Ubuntu 15.04 from howtoforge.com.

Graphical tool

gitk and git-gui

sudo apt-get update; sudo apt-get install gitk
git clone [email protected]:joshuaflanagan/gitk-demo.git

cd gitk-demo
gitk --all   # show all refs (branches, tags, etc.)

What I find is gitk will take a little time to rebuild something because after I run git pull, it will show several files (if not all) are uncommitted.

Gitk worked with git-gui program (sudo apt-get install git-gui) which is a gui program to run rescan/stage/commit/push. You launch git-gui from gitk-File-Start git-gui.

It looks these 2 gui tools are sufficient enough.

Others

  • QGit - QGit is a git GUI viewer built on Qt/C++.
  • git-cola ???
  • gitg
  • Giggle - Similar to gitk
  • SmartGit - Support push, pull, fetch

Git Tips

git difftool & meld

git difftool FILENAME -y will launch 'meld' (if it has been installed before) to compare the file between revisions by using custom tools. It has to be run before we call git add. This is quite convenient since you can double check before running git commit. The '-y' argument is used to launch a diff tool without a prompt. See the documentation here.

Monitor/find files that have been changed since last pull

git branch
git fetch origin/master    # or git fetch on Windows
git diff --name-only origin/master

git log                    # local
git log origin/master      # remote repository

$ git status
On branch master
Your branch is behind 'origin/master' by 2 commits, and can be fast-forwarded.
  (use "git pull" to update your local branch)

nothing to commit, working directory clean

We can also use the gitk to view the log. The following is a screenshot from Window's git gui (Windows Start > Git > Git Gui. Then Repository > Visualize All Branch History). As you can see the local repository (master w/ yellow circle) is 2 commits behind the remote repository (remotes/origin/master w/ blue circle).

Gitk2.png Gitk.png

Pull and overwrite local files

If we want to run git pull and also overwrite possibly changed local file, we use (see stackoverflow) 

git fetch --all
git reset --hard origin/master

Your branch is ahead by X commits after running git pull

git status shows Your branch is Ahead by X commits after running git pull. See this post. The solution is to run git fetch after git pull.

This repository currently has approximately XXXX loose objects

http://stackoverflow.com/questions/21457407/git-gui-perpetually-getting-this-repository-currently-has-approximately-320-lo 

git gc --aggressive

git commit -am

works only for simple edits. It does not work for renaming files, etc.

git diff --staged

git checkout -- FileName

to recover a modified/deleted file where -- is related to branches 

rm FileName
git pull   # do nothing
git status # still miss 'FileName'
git checkout FileName

git reset HEAD FileName

to un-stage a file

Rewrite a commit

ps. This is different from the case of Pull and overwrite local files

Suppose someone made a stupid commit. We want to get a previous version of the code and commit that one. 

git checkout XXXXXX -- FileName
git commit -am CommitMessage

origin and nickname

git remote add NickName https://XXX
git remote 
git push -u NickName master

Git hosting services

Github

Gitlab

Compared to Github, Gitlab can

  1. host private projects for free
  2. host static web pages on http://pages.gitlab.io/.

Import from CVS

sudo apt-get install git-cvs  # cvsimport command
tar xzvf tmp.tar.gz
cd DirectoryName

# The next command requires a connection to the CVS server even we have a copy of CVS in the local pc
git cvsimport -C ~/Downloads/tmp ModuleName # SLOW & give up

General resources

Host web site

Some Examples

Jekyll

Install Jekyll

on Ubuntu 14.04 I use (the second command will take a while to start), 

sudo apt-get install ruby1.9.1-dev
sudo gem install jekyll

Example 1

cd /tmp
jekyll new MyNewSite
# It will create a new folder 'MyNewSite' with about.md, _config.yml, css (folder),  
#    _includes (folder), index.html, _layouts (folder), _posts (folder) and _sass (folder).
cd MyNewSite
jekyll serve --watch
# It'll say the Server Address: http://127.0.0.1:4000/
# We can open a browser to see a template of html page created by '''jekyll new''' command.

hexo

Hexo is a fast, simple and powerful blog framework. You write posts in Markdown (or other languages) and Hexo generates static files with a beautiful theme in seconds.

Example

Sphinx

Examples

Markdown

Atom screenshot.png

Github Markdown

Retrieved from "https://wiki.taichimd.us/index.php?title=Github&oldid=9395"