Sysadmin/GitKdeOrgManual

Overview of facilities

• KDE Identity (identity.kde.org)

Account management; notably managing your SSH public keys for read-write developer access.

• git.kde.org

The main git server. Should be used only for pushing new commits to a repository over the SSH protocol.

• anongit.kde.org

Several servers which allow read-only access to the repositories via the git:// and http:// protocols. They are requested to update when anyone pushes to a repo on git.kde.org, so it can be thought of as being always up-to-date.

• KDE Projects (projects.kde.org)

Central project hub and primary repository browser.

• quickgit.kde.org

Alternative repository browser. At present the only way to view personal clones of project repositories and personal scratch repositories (see below), however the former are planned to appear in projects.kde.org in the future.

• ReviewBoard (git.reviewboard.kde.org)

Patch review (account sign-up via KDE Identity). [1].

• commits.kde.org

Provides Git commit "short URLs", redirecting to KDE Projects and gitweb.kde.org pages as appropriate (example).

• commitfilter.kde.org

Sends an email with each commit for the projects you want to watch.

How to get read-write developer access

KDE developer accounts are managed through KDE Identity. If you already have a KDE SVN developer account, it has been imported into KDE Identity and you may use the Password Reset feature to set a password and manage your SSH public keys. If you don't have a developer account yet, you can request Developer Access in the website's menu upon registering and logging into your account.

Information For KDE Developers

You can find general information about using Git as a KDE Developer on the KDE Git page on TechBase.

To configure Git for your KDE Development environment, please see the KDE Git Configuration page on TechBase.

You can find some simple step-by-step recipes for using the KDE Git repositories on the KDE Git Recipes page on TechBase.

Overview of repository URL schemes

URL prefixes

Anonymous read-only access uses the following URL prefix:

 git://anongit.kde.org/

Read-write developer access uses this prefix instead:

 [email protected]:

Repository paths

Following the prefix, here are the path schemes for different types of repositories:

• <project identifier>

A KDE project repository, be it part of the KDE SC, KDE Extragear or KDE Playground.

• websites/<address sans leading www. and dots replaced by dashes>

A KDE website project, e.g. websites/projects-kde-org.

• sysadmin/<repository name>

Non-public repositories used by KDE's sysadmin team.

• clones/<original repository path>/<KDE Identity user name>/<user-chosen repository name>

Personal clones of project repositories, e.g. clones/konversation/hein/morecowbell or clones/websites/projects-kde-org/hein/pluginwork (more below).

• scratch/<KDE identity user name>/<user-chosen repository name>

Personal scratch repositories are a means to start a new project or just to store your favorite .bashrc in a safe location: anything is allowed so long as it is related to KDE or your work for KDE in some way (more below).

Let Git rewrite URL prefixes

Instead of remembering the above URL prefixes, you can also put the following in your ~/.gitconfig:

 [url "git://anongit.kde.org/"]
     insteadOf = kde:
 [url "[email protected]:"]
     pushInsteadOf = kde:

If you are in a corporate environment (e.g. behind a firewall) which only allows Ports 80 (HTTP) and 443 (HTTPS), use the following configuration instead (i.e. substitute git:// for http://):

 [url "http://anongit.kde.org/"]
     insteadOf = kde:

Then, to clone e. g. the Amarok repo, just do

 $ git clone kde:amarok

By using the kde: prefix, read access will automatically happen over Git, and authenticated SSH is only required for pushes. Since commits are mirrored to anongit right when you push them, you will not have to worry about anongit being outdated.

Server-side commands

git.kde.org understands several server-side commands that can be used on the command line via SSH in this fashion:

ssh [email protected] <command> [parameters]

To simplify the first argument to "kde" (i.e. "ssh kde <command> [parameters]"), put the following in ~/.ssh/config:

Host kde
    HostName git.kde.org
    User git

The following is a list of the commands that are currently available, broadly divided into categories according to their purpose.

Commands for information retrieval

• info (link here)

Shows a table of repository paths and path patterns you have the permission to see along with details about your access rights to them.

A brief legend for the permission flags shown in the listing:

• @R - Read permissions.
• @W - Write permissions.
• @C - Create permissions (e.g. the initial push to a newly-created repo).

If you want to list actual repositories corresponding to patterns listed by info, such as your personal scratch repositories, see the expand command described next.

• expand [regex] (link here)

Like info above, but actually walks through the repositories to verify the information. It's much slower as a result, and should be used if info doesn't provide enough information. For example, info will list your personal scratch space only in the form of a pattern while expand can list the actual repositories located there.

The output is limited to about 20 rows. The optional regex parameter allows you to filter the listing.

• who-pushed <repository path> <commit sha1 hash> (link here)

Shows the KDE Identity user name of the contributor who pushed the specified commit to the specified repository.

Commands to manage personal repositories

• clone <path to source repository> <clone name> (link here)

Can be used to make a personal clone of a project repository.

An example:

 ssh [email protected] clone konversation mykonvi

This results in a clone at clones/konversation/<your KDE Identity user name>/mykonvi.

A second example with a longer source repository path:

 ssh [email protected] clone websites/projects-kde-org newtheme

This results in a clone at clones/websites/projects-kde-org/<your KDE Identity user name>/newtheme.

More on personal clones here.

• D rm <repository path> (link here)

Used to delete a personal clone of a project repository or a personal scratch repository. Requires the repository to be unlocked first using the unlock command and will additionally ask for confirmation. See also the D trash command as an alternative to outright and irrevocable deletion.

• D lock <repository path> (link here)

Locks a repository, causing the destroy command to deny deleting it.

Newly-created repositories are locked by default.

• D unlock <repository path> (link here)

Unlocks a repository, making it possible to delete it using the destroy command.

Commands to manage the personal trash area

• D trash <repository path> (link here)

Moves a repository to the personal trash area, creating an entry in the form <repository path>/<timestamp> there. The timestamps, which have second precision, make it possible to have more than one version of a repository in the trash area at the same time.

Note: Entries in the personal trash area are automatically removed after 28 days!

• D restore <trash area entry> (link here)

Restores an entry from the personal trash area (see the list-trash command below for how to list the contents of your personal trash area).

restore will deny restoring an entry if doing so would overwrite an existing repository.

• D list-trash (link here)

Lists all entries in the personal trash area, in the form <repository path>/<timestamp>.

Commands related to repository importing

• hooks enable <project repository path> (link here)

An example:

 ssh [email protected] hooks enable konversation

Available only to repository and system administrators, this command enables several hook scripts that git.kde.org will then execute during a push operation to the specified project repository. Importantly, it also enables write access for non-administrators, which is otherwise disabled along with the hooks scripts.

The hook scripts in question are the ones reponsible for forwarding commits to the kde-commits mailing list and CIA.vc, and for processing commit message keywords (BUG, CCMAIL, etc.) that may interact with KDE Bugzilla or cause further emails to be sent. As these hook scripts are only available to project repositories, and not to personal repositories, the command only applies to them.

After creating a new, empty project repository for you the system administators will initially disable the hook scripts so you can safely import large numbers of old commits.

Commands for system administrators

• sudo <KDE Identity user name> <command> (link here)

Used by system administrators to run one of the above as another user.

• able <en|dis> <@all|repository path> (link here)

Used by system administrators to enable or disable writes to particular repositories or all repositories, for maintenance.

• hooks disable <project repository path> (link here)

Disables the hook scripts git.kde.org normally executes during a push operation to a project repository. While the hook scripts are disabled only repository administrators can push commits to a repository. Both system and repository administrators have the ability to reenable the hook scripts using the hooks enable command.

• ohnoes <show|recover> <repository path> <gitref> (link here)

Used by system administrators to recover deleted branches or mistaken force pushes (rewinds).

Commit hook keywords

When you commit changes to Git you will be asked for a description of your commit. There are several special keywords defined that you can use in this description. These keywords are always in uppercase.

The following keywords are pseudo-headers - they have to appear at the start of a line and be followed by a colon:

• FEATURE: [<bugnumber>]
  Marks the feature as implemented by CC'ing the commit message to <bugnumber>[email protected]. This keyword will also be used to automatically extract entries for the release changelog, so it makes sense to use it for new features even if you don't have a bugnumber for the feature.

• BUG: <bugnumber>
  Marks the bug as fixed by CC'ing the commit message to <bugnumber>[email protected]. This keyword will also be used to automatically extract entries for the release changelog. To use this feature, you need to create a user on kde bugs with a same email and request for bug edit permission through a KDE Sysadmin ticket.

• FIXED-IN: <version>
  If the BUG keyword is used to close a bug, this keyword will update the bug's release version (Version Fixed In field) that the fix appears in. You can only select one version, so prefer the lowest release version that will actually have the fix.

• CCBUG: <bugnumber>
  CC's to the bugreport by sending mail to <bugnumber>@bugs.kde.org. This will not close the bug.

• CCMAIL: <email-address>
  CC's to the given e-mail address. Used to notify other developers, usually the maintainers of your commit.

• REVIEW: <reviewnumber>
  Closes the review on KDE Reviewboard and adds a comment which refers to the commit.

• GUI:
  Indicates a user visible change in the user interface. This is used to make the documentation team aware of such changes.

• DIGEST:
  Notifies the KDE Commit Digest team of interesting/important commits to look at (for easy collection, summarization and presentation of commits to end-users).

These keyword can appear anywhere on a line:
(should the svn stuff be removed from this article?)--Sreich 04:07, 27 December 2011 (UTC)

• SVN_SILENT
  Marks the commit message "silent" by adding "(silent)" to the subject of the mail to allow filtering out trivial commits. Use this tag carefully and only for really uninteresting, uncontroversial commits.

• GIT_SILENT
  Same as SVN_SILENT

• SVN_MERGE
  Special keyword for people that use the svnmerge script. Marks the commit message as being a merge commit by adding "(merge)" to the subject of the mail. This way receivers can filter out mails that are caused by merging the same patch from one branch to another branch. Reviews of those mails is usually not needed. This keyword filters out the endless property changes used by this script.

Personal repositories

git.kde.org currently offers two types of personal repositories: Personal clones of project repositories and personal scratch repositories.

Personal clones of project repositories

A personal clone of a project repository can be created using the server-side clone command on the command line:

 ssh [email protected] clone <path to source repository> <clone name>

This will create a clone of the source repository at clones/<path to source repository>/<KDE Identity user name>/<clone name>. (See more examples of clone in action here.)

This scheme makes it very easy to locate all personal clones of a given project and should be preferred over making one in your personal scratch space. (In fact, the server-side clone command won't allow you to clone a project repository into your personal scratch space, but nothing technically prevents you from taking the detour of a local clone to achieve this.)

Personal clones of project repositories currently do not show up on KDE Projects, but we have plans to change that in the future. Until then, you can use quickgit.kde.org to browse these repositories.

Personal scratch repositories

Personal scratch repositories are a means to start a new project or just to store your favorite .bashrc in a safe location: anything is allowed so long as it is related to KDE or your work for KDE in some way.

For the rest of this instruction, let's assume the following names:

• <kde-id> - Your KDE identity user name
• <repo> - Your (new) git repository name

If you start a new project then you need to create a git repository out of it before you start to push it to the scratch repositories. This is done by going into the the directory which will contain your project and type:

git init

This will convert the local directory to a git repository. This is of course not necessary if you already have a local git repo containing your project.

To make pushing to the repository convenient, it is recommended to add a remote to push to first:

 git remote add origin [email protected]:scratch/<kde-id>/<repo>

Now you need to create something to push. Otherwise git will refuse the next step. So it is recommended that you create a README for your project and then commit that. Afterward, you perform the first push to the repository as follows:

 git push --all origin

Now your new scratch repository is created and initialized. It will take about 30 minutes until the creation of the new repository has propagated to the other tools and is visible there.

git push --all does not push tags. You can push them in a second step with

 git push --tags ...as above...

Personal scratch repositories can be browsed on quickgit.kde.org.

You can set a description on your scratch repository by issuing a command

 ssh [email protected] desc scratch/<kde-id>/<repo> "Application to do magic"

If you feel your new project is ready for the wider world and/or wish to signal that it welcomes outside contributors, you may wish to promote it to the status of a KDE Playground project. KDE Playground project repositories are located at the top-level, i.e. the repository will be moved out of your scratch space and may have to be renamed in the event of a collision with an existing repository name. KDE Playground projects are featured on KDE Projects and covered by the kde-commits mailing list (and thus CommitFilter), LXR, the EBN and CIA, unlike personal scratch repositories.

To request your scratch repository be promoted to the status of a KDE Playground project, you currently need to file a sysadmin repo request. In the future we plan to provide a fully automated facility on KDE Projects.

Note that we have deliberately decided not to allow the direct creation of KDE Playground projects; the path to existence for a KDE Playground repository project always leads through a personal scratch space first. This is to give you the power to decide whether your project is ready, and also to force you to deliberate whether it truly is.

Deleting personal repositories

A personal repository can either be deleted outright and irrevocably by using the destroy command (which requires you to unlock it first to avoid accidental deletion), or you may move it to the personal trash area with the trash command.

 ssh [email protected] D trash scratch/<username>/<project>

Entries in the personal trash area are kept for 28 days, and can be resurrected at any moment during those 28 days by way of the restore command. You can list the current contents of your personal trash area with the list-trash command.

Using Review Board and post-review

A very comfortable way of posting changes for review is Review Board, where every project repository has its own entry. You can read about how to use Git and Review Board on the KDE Review Board page.

Here's an insightful blog post that explains how to use post-review.

Requesting project migrations from KDE SVN or Gitorious.org

To get your project moved from KDE SVN or Gitorious.org to git.kde.org, you have to file a sysadmin request. It will ask you for the following information:

• The name and description of the project.
• The current location of the project.
• Its current or intended module (e.g. playground/utils or extragear/network).
• Which KDE Identity user name(s) should have admin rights to the repository and the entry on KDE Projects.
• The email address that the ReviewBoard group for the project should send emails to.
• The date and time the migration should take place (can be "asap").

When we have completed processing your request, there will be an empty repository at the chosen path (more here) that the repository administrators can push the data into. (When converting from KDE svn to git this typically involves writing a rule set, running svn-all-fast-export, and then pushing the created repository into the new git path.) Once you are done pushing everything to the repository, use the hooks enable command to enable the commit hooks and allow write access to non-administrators.

Advanced Git

Safety Precautions

With these techniques, always work on a disposable copy of the repo with all the remotes removed, so if you screw up, it doesn't really matter. Also, work on a separate branch. That way, you can usually use git reset --hard <original-branch> to get back to the starting state.

Also, make sure there are no grafts around (eg: linking to the old kdelibs history in the case of frameworks). The safest way to do this is to use fresh checkouts.

Merging repositories

The git-merge-repo script in kde-dev-scripts can merge the tree and history of one git repository into another.

First, create a commit in the source repo that removes any files you don't want to copy, and rearrange the remaining files to be as you want them to appear in the target repo. It is important the HEAD of the source and target repositories have completely disjoint trees (so you could copy one tree into the other with no file conflicts).

Then go to the root of the target repository and run

/path/to/kde-dev-scripts/git-merge-repo <path to source repo>

This will preserve commit identities (unless you filter the source repository - see below).

Note: Before pushing such a merge, talk to sysadmin. They can temporarily disable commit hooks (like CCMAIL and BUG) so that people do not get emails about old commits.

Filtering

git filter-branch allows you to edit history. This is useful when you want to merge only a small part of one repository into another. You can trim the tree, and also alter the commit messages to disable the KDE commit hook keywords, so that the merged history does not cause CCMAILs to be sent for ancient commits (although it's probably better to ask sysadmin to disable the commit hooks temporarily while you push).

A combination of --tree-filter, --prune-empty and --msg-filter generally gets what you want. For example,

git filter-branch --prune-empty \
                  --tree-filter "find -type f -\! -path './.git/*' -\! -name foo.\* -delete" \
                  --msg-filter 'sed -e 's/^\(CCMAIL\|REVIEW\|BUG\|CCBUG\|FEATURE\)/(&)/'; echo; echo "Commit $GIT_COMMIT in <source-repo>"' \
                  HEAD

This example will remove everything that does not match foo.*. Note the -path argument to find that makes sure you don't delete any of git's own files. --prune-empty will remove non-merge commits that no longer have any effect on the tree (you can run git rebase after to trim the merge commits if you want). --msg-filter neuters the keywords and adds information about where the commit came from (don't forget to change <source-repo>!)

More complex filters are possible. Have a look at the man page for git-filter-branch.

See mastering git filter-branch: points to extract a subproject for more helpful hints.