DANGER WILL ROBINSON!!! THIS IS AN INCOMPLETE DRAFT!!!
This workflow is designed to be as close as possible to the KDE SVN Workflow. It is only recommended to be used for the first week or two of using Git with KDE while you become familiar with the basic Git commands. Once comfortable with the basic commands you should then move on to the Feature Branch Workflow.
In particular this workflow will not use git branches, merging or any remote features, all feature development work will be in the local master branch (the Git name for trunk).
This section documents how to set up Git and your code repository for development.
Follow all the instructions on the KDE Git Configuration page, the instructions given below assume you are using the standard configuration.
You need to copy your central code repository from git.kde.org into your local KDE source directory. In Git this process is called cloning.
To clone your project repository:
cd your/source/dir git clone kde:<project> cd <project>
In our KFoo example:
git clone kde:kfoo cd kfoo
See the KDE Build Environment page for advice on structuring your source directory.
If you have a slow or intermittent internet connection then you may prefer to download a snapshot tarball to bootstrap your clone. You can copy the required command from the projects.kde.org Repository page for your project, but it will be of the form:
wget -c http://anongit.kde.org/<project>/<project>-latest.tar.gz
This section documents the basic workflow for developing a new feature for the master unstable branch (aka trunk in subversion). You will first make local code changes and commit them in your local repository before pushing them to the central code repository.
Changing and building code itself is no different from when using subversion, but how changed files are managed and committed is very different, as are the names and commands used. There are two key differences:
Files in a git repository can be in various states:
Note that different versions of a file can be in the Untracked and Tracked states at the same time.
This difference between staged and unstaged changes allows you to choose what changes you want in each commit.
To see what the current status of your local repository is:
This will show you the status as follows;
# On branch master # Your branch is ahead of 'origin/master' by 1 commit. # # Changes to be committed: # (use "git reset HEAD <file>..." to unstage) # # modified: src/kstagedfile.h # # Changed but not updated: # (use "git add <file>..." to update what will be committed) # (use "git checkout -- <file>..." to discard changes in working directory) # # modified: src/kchangedfile.h # # Untracked files: # (use "git add <file>..." to include in what will be committed) # # src/kuntrackedfile.txt
To add a changed file to the staging area, or to add a new file to be tracked by the repository:
git add <filename>
To delete a file from the repository and add the deletion to the staging area:
git rm <filename>
To reset a staged file to be unstaged but without losing your changes:
git reset HEAD <filename>
To reset an unstaged file to the currently committed version and irretrievably throw away your changes:
git checkout <filename>
The git commit command differs from the svn commit command in only creating a commit in your local clone of the repository, it does not merge the changes into the central repository. This allows you to queue up a series of small changes before merging them into the central repository. This local commit history can also be modified before you finally do push the changes.
The git commit command adds all files in the staged state into a new commit:
If you have only changed tracked files then you can automatically add all tracked files that have been changed into staging and commit all the staged files in one command:
git commit -a
To see a list of what files have been changed but not yet committed:
To see what code changes have been made but not yet staged:
git diff git diff <filename>
To see what code changes have been staged but not committed:
git diff --staged git diff --staged <filename>
To see all the commits made:
To see what changes were made in a commit:
git show <sha5>
The Git diff, log and show commands have many options for changing what is shown. In particular diff can show the difference in a file between any two commits.
When you are ready to push all your local commits to the central repository, you first need to update to the latest version of the master code in the central repository:
git pull --rebase
This will update the underlying master code and apply your commits on the top of the latest version of the central repository, this is known as a rebase. Normally the rebase should apply cleanly, but sometimes there will be a conflict which you will need to resolve before you can complete applying your commits. There will be instructions given by git on how to complete the commit if required [TODO: get a copy of the output text].
Note that you cannot rebase if you have uncommitted changes. If you need to rebase but have changes that you don't want to commit, then you will need to stash the changes.
Depending on how much the underlying master code has changed you may want to rebuild and test the code before doing a final rebase. It is recommended to do regular rebase's between commits to keep the possibility of conflicts to a minimum.
Once your local commits have been cleanly reapplied you can push your changes to the central repository:
git push origin master:master
Git will report if the push was successful or explain why it failed.
If your bug fix is only for unstable master then no special actions are required, just follow the same steps as the feature development workflow above.
If your bug fix is for a stable branch then this cannot be done without using git branches. The steps required will be given below but not explained in any detail. If possible it is recommended you wait until you are familiar with using the Feature Branch Workflow.
The steps detailed below are very inefficient as they use the same build tree and environment for unstable and stable which may cause a lot of rebuilding. A more efficient method is detailed in the Feature Branch Workflow.
First, find out the name of the stable branch you want to bug fix:
git branch -r
This will list all the branches on the central repository, for example kdelibs includes the following branches:
origin/HEAD -> origin/master origin/KDE/4.5 origin/KDE/4.6 origin/kdecore/klocale-win origin/kdeui/kdatetimeedit origin/master
Here origin/KDE/4.6 is the 4.6 release of kdelibs which we will use for this example.
Next, create a local branch pointing to the stable release branch you want to bug fix:
git branch --track KDE/4.6 origin/KDE/4.6 git checkout KDE/4.6
If you have previously created the stable branch and want to do a new bug fix, then simply check it out again and update it:
git checkout KDE/4.6 git pull --rebase
Now make your required changes to the code, build and test. Once done commit the code changes to your local repository, remembering to include the BUG: and FIXED-IN: keywords in the commit message:
git commit -a
Make a note of the sha5 of the commit as you will need it later to copy the commit to the unstable master branch.
Now you can push your changes into the central code repository:
git pull --rebase git push origin KDE/4.6:KDE/4.6
The next step is to apply the bug fix to the unstable master branch. First check the master branch back out and update it:
git checkout master git pull --rebase
Next, Git makes it easy to apply the existing bug fix from the stable branch by using the cherry-pick command:
git cherry-pick -x -e <sha5 of stable commit>
This will allow you to edit the commit message as required before the change is committed. If the patch does not apply cleanly a message will appear and the changes will not be committed. You will then have to edit the changes and finish the commit manually.
Once cleanly applied and committed you can build and test the change, then merge them into the central code repository:
git pull --rebase git push origin master:master