分类: LINUX
2010-09-13 10:44:45
This isn't meant to be an all-encompassing guide by any means - it is meant to be a really quick walk-through on how to do some basic operations on the pacman codebase in git, such as submitting a patch.
For more extensive tutorials, check out the following:
In addition, all git commands have decent manpages to refer to. They can be reached one of two ways - for the git-add command, type 'man git-add' or 'git help add'.
The first step with git is cloning a remote repository. This is known in the CVS and SVN camps as checking out. GIT checkout has a different purpose, but that will be covered later.
To grab the pacman source into a new directory named 'pacman', run the following:
git clone git://projects.archlinux.org/pacman.git pacman
This will check out a local copy of the repository for you. This means you have the FULL history of the project on your computer, not just the most recent revision. This allows you to get work done even when offline, for example.
The first steps after cloning may be just to look around. If you have read the tutorials mentioned above, even if you do not understand everything in them, you will be much better off.
You will probably want to set up your name and email address for use in commit logs:
git config user.name "Your Name" git config user.email "me@example.com"
If you pass the '--global' flag to the above commands, the name and email will be stored in ~/.gitconfig, so will be used for all git projects unless overridden by a setting in the individual project.
To update your local repository with any new branches, run 'git pull'.
'git branch' will show you a list of branches. Initially, master is the only branch. However, if you pulled from a remote repo, you may have grabbed other branches- these can be seen with 'git branch -r'. Read the manpage for details.
When working with git, it is good practice to never do your work on the master branch. This should stay clean to allow you to run 'git pull' and ensure that conflicts do not happen on the update.
To create your own working branch, do the following (naming it whatever your heart desires):
git branch working git checkout working
Or compress the above into one command:
git checkout -b working
To switch back to the master branch use:
git checkout master
but you can only leave a branch if there are no pending changes to commit. Find out what changes have not been committed using:
$ git status # On branch working # Changed but not updated: # (use "git add..." to update what will be committed) # # modified: lib/libalpm/util.c # no changes added to commit (use "git add" and/or "git commit -a") $
Either commit the changes or revert them before switching branches.
I highly recommend you read the man page for these commands.
Each developer usually has his own repository, it might be interesting to have them all available locally. For example:
git remote add toofishes
You can see all available branches with 'git branch -r', and as above, create your own based on one of these remote branches:
git checkout -b toofishes-working toofishes/working
If you use CVS or SVN and are used to 'co' being equivalent to 'checkout', try the following:
git config --global alias.co "checkout"
You can set any other aliases you like. For example:
git config --global alias.chp "cherry-pick" git config --global alias.b "branch" git config --global alias.rf "checkout HEAD"
'git status' is highly helpful, it is recommended to read up on that.
Woo! You found a bug in pacman (what a surprise) and know how to fix it. Ensure you have your working branch checked out ('git checkout working'). Then edit the file(s) you need in order to make your changes. Compiling is a good idea to ensure your patch didn't break anything, and if it is a big change, running 'make check' is highly recommended.
So what do you do now? First, run 'git status'. You should see a list or even a few lists of files. The descriptions by each are a bit confusing, but you should be able to figure it out. GIT takes a different approach than CVS or SVN to committing changes- it doesn't commit a thing by default. You have to tell it what to commit, usually by running 'git add
To commit your patch to your branch:
git addgit commit -s
You will then be prompted for a commit message. When writing the message, keep the following in mind. The first line is used as a patch summary- keep it short and concise. Next, skip a line and type out a full description of what your patch does. By full, I don't mean long- if you described everything in the summary line, then don't even bother with a message. Finally, skip one more line and you will have your Signed-off-by. This should have been automatically added by passing the '-s' parameter to 'git commit'.
There is one more important step before submission. Because git is distributed, you don't have the most current version of the repository unless you go out and get it. In the easiest case, this is just running 'git pull'.
You also want to make sure your patches are based off the most recent revision, known as the 'head'. To do this, checkout your branch with your patches, and use the following command:
git rebase master
To visualize what the above command did, qgit can be very helpful.
To format a patch for email submission and review:
git format-patch master
This command will format all patches that make up the difference between your working branch and the master branch. They will be saved in the local directory; to store them elsewhere read up on the '-o' option.
So you sent off your patch to the ML and you got a few suggestions back. How does one fix it? Hopefully you did it on a branch and not the master branch, otherwise you are going to have a much tougher time. :)
If it was the last patch on a branch:
(edit the required files) git addgit commit --amend
If it was deeper in your patch tree, use git rebase -i. Use git log to find the sha1 of the commit just before the one you wish to edit (or the unique prefix), and then:
git rebase -i(edit the text file that appears so 'edit' appears next to the commit you wish to modify) (edit the required files) git add -u git commit --amend git rebase --continue
has good information on the above, that is where most of this came from. After fixing your patch, you will probably want to rebase it as described above, and then use format-patch to submit it again.
A nice way to send patches is to use git send-email, but it requires some initial setup, especially for the smtp client. If you follow the instructions carefully, it should go fine :
Then you just need to tell git to use msmtp:
git config --global sendemail.smtpserver "/usr/bin/msmtp"
For each git repo, you can specify the email address where the patches should be sent:
git config sendemail.to "pacman-dev@archlinux.org"
Then simply send your patches generated by git format-patch:
git-send-email 0001-amazing-new-feature
Commands you will definitely want to be knowledgeable on:
clone (only once!), branch, checkout, status, pull, fetch, diff, add, commit, rebase, format-patch
qgit in extra is a great GUI viewer for git repositories. In addition, read up on 'git-instaweb'.
Used to CVS or SVN-like behavior on commits, where all changes in the local tree are committed? Try using 'git-commit -a'.
Not running a black and white console? Then you probably want color in a lot of GIT's console output.
git config --global color.branch auto git config --global color.diff auto git config --global color.status auto