AutoQA Release Process

This page describes the process for tagging, building and deploying a new version of autoqa. This page assumes a basic understanding of rpm file syntax and commands such as,  and.

Numbering scheme
Each release has X.Y.Z identification denoting a major, a minor and a revision number:


 * Major number is increased when AutoQA makes incompatible changes in its test API. (Not used currently, since no stable public API has been offered yet.)
 * Minor number is increased when AutoQA adds new features.
 * Revision number is increased when AutoQA adds new hotfixes, but no new features.

Pre-requisites
You must have AutoQA source code checked out with write access (ssh:// protocol, requires gitautoqa membership):

git clone ssh://git.fedorahosted.org/git/autoqa.git cd autoqa

= Stable release = The following process concerns a stable public release of AutoQA. We use this procedure when we want to officially release a new version of AutoQA for all our users.

Switch Git branch
All the changes must be done in a correct git branch.

 If you want to create a new X.Y.0 major or minor release, create a new release branch: git checkout -b release-X.Y master  If you want to create a new revision release X.Y.Z, switch to a corresponding existing release branch: git checkout release-X.Y and apply all hotfixes you have prepared. 

Update and
Every new release must be mentioned in the rpm file. We also want to provide high-level changes overview in the file.


 * 1) Edit  by incrementing the   and updating the
 * 2) Edit  file and add a section for the new release. Document only the most important end-user visible changes.
 * 3) Locally commit the changes git commit autoqa.spec NEWS

Tag the release
The last commit (changing the file) must be tagged to mark the release.


 * 1) Tag the commit: git tag vX.Y.Z
 * 2) Check that the tag was applied to the correct commit by inspecting: git log --decorate=full

Pull changes back to master
After you tag the release on the release branch, you need to pull the release-related changes (now tagged with vX.Y.Z tag) into the master branch. Merging is a preferred approach, because it ensures the master branch will contain the vX.Y.Z commit. But it is desirable only for a fast-forward merge, otherwise some other changes (we might not want) may be pulled in as well. In that case (e.g. the release branch was used for stabilization and contain several previous commits) we need to use cherry-picking.


 * 1) Change to the master branch git checkout master
 * 2) If you created the release branch immediately before tagging a new release, merge the changes using fast-forward approach: git merge --ff-only release-X.Y
 * 3) If you can't use fast-forward merge or it failed, cherry pick the appropriate commit instead: git cherry-pick vX.Y.Z

Push changes to remote repository
It's time now to push all the changes to the shared remote origin repository.


 * 1) Push the release branch changes: git push --tags origin release-X.Y
 * 2) If you modified the master branch (i.e. for major or minor releases), push it too: git push --tags origin master

Upload tarball
Like many projects, the appropriate method to release a new version is by tarball. Once you have tagged the release, upload a new tarball using the following commands.


 * 1) Check-out the correct tag git checkout vX.Y.Z
 * 2) Upload a new release tarball make upload

Build a source RPM
With the tarball uploaded, it's time to package the new release as an RPM.


 * 1) Check-out the correct tag git checkout vX.Y.Z
 * 2) Build a source package make srpm

Build for applicable releases
With a source RPM created, it's time to build updated packages for any existing stable repositories. This includes, and, depending on the time of release, potentially. Traditionally, this step would be handled by running the command. However, since is not yet packaged and available in Fedora repositories, updates are built locally using.


 * 1) Build packages using mock for Fedora, specify version using   variable.  For possible values for , consult Packaging:DistTag
 * 2) Repeat the build procedure for all desired releases

Create updates
With packages built, it's time to submit them as updates. Traditionally, this step would be handled by using the bodhi update tool. However, since is not yet packaged and available in official Fedora repositories, a custom package repository is used to deliver updates.

  Mirror the autoqa package repository locally rsync -avz fedorapeople.org:/srv/repos/fedora-qa/autoqa ~/public_html/ ; cd autoqa/   Add locally built packages to the desired repositories ./move-pkgs.sh path/to/autoqa.git/build/MOCK/*/*.rpm   Update the yum repo metadata ./update-repos.sh   Update remote repository with changes rsync -avz ~/public_html/autoqa fedorapeople.org:/srv/repos/fedora-qa/   Change permissions on the server so that anyone in gitautoqa FAS group can update the repos chgrp -R gitautoqa /srv/repos/fedora-qa/autoqa chmod g+w -R /srv/repos/fedora-qa/autoqa </li> </ol>

Purging old release branches
If we are sure we will no longer work on older releases (e.g. before X.Y), we can also delete older release branches. If W < Y, then we can delete release-X.W:

<ol> Make sure there's a tag at the tip of the branch and therefore no work is lost: git tag --contains origin/release-X.W This must output some tag (like vX.W.3). Otherwise the branch contains commits which are not part of any tag.</li> Delete the remote branch: git push origin :release-X.W </li> Prune the local branches (that no longer exist on the remote server): git remote prune origin </li> </ol>

The same goes for deleting older major branches (branches release-Q.Y where Q < X).

= Testing release = The following process concerns a testing release of AutoQA. We use this procedure when we want to deploy a testing version of AutoQA onto our staging server for thorough testing.

Build and install RPM

 * 1) SSH to the staging server and go to the AutoQA checkout directory (create one if you don't have one).
 * 2) Switch to the git branch you want to test, probably master: git pull; git checkout master
 * 3) Build an RPM from current HEAD using the same approach as in, but with appending   option: make mock DEVEL=1
 * 4) * Your user must be in the mock group to be able to run this.
 * 5) Install the RPM created in.
 * 6) * Since we use  to generate package versions and sometimes we build and install package from other branches than master, you can actually end up with a package version that is lower than the one currently installed. Instead of using   use.
 * 7) * Be sure to inspect all and  files (if they are created, yum will tell you so) and merge the changes.