Contribution guidelines

This guide describes rules for how to get your contributions into Anaconda. However, if you seek help with implementing changes in Anaconda, please follow our blog series or an addon guide to create Anaconda addon.

Setting up development container

The anaconda team uses a containerized development environment using toolbx. If you can install toolbx or distrobox on your system, it is highly recommended to do that:

  • It is known to work and gives you reproducible results.

  • It avoids having to install development packages on your main machine.

If you are not interested in dealing with containers, just skip this part and continue on the next one:

sudo dnf install toolbox

To create and enter a development toolbx for Anaconda just run these commands:

toolbox create
toolbox enter

Installing dependencies

If you are using cockpit/tasks container for Web UI development only, you can skip this part.

To get all the dependencies and prepare the environment in the container or on your system just run these commands:

sudo ./scripts/testing/install_dependencies.sh

How to run make commands

Anaconda uses autotools so there are familiar ./configure script and Makefile targets. To prepare Anaconda sources, you need to run these commands:

./autogen.sh && ./configure

How to Contribute to the Anaconda Installer (the short version)

  1. I want to contribute to the upstream Anaconda Installer (used in Fedora):

  • base and test your changes on a clone of the fedora-<next Fedora number> branch.

  • open a pull request for the fedora-<next Fedora number> branch (fedora-38, etc.)

  • check the Commit Messages section below for how to format your commit messages

  • check the Release Notes section below for how to provide a release note

  1. I want to contribute to the RHEL Anaconda installer:

  • base and test your changes on a clone of the rhel-<RHEL number> branch.

  • open a pull request for the rhel-<RHEL number> branch (rhel-9, etc.)

  • check the Commits for RHEL Branches section below for how to format your commit messages

  • check the Release Notes section below for how to provide a release note

If you want to contribute a change to both the upstream and RHEL Anaconda then follow both 1) and 2) separately.

Which is my target git branch?

Depending on where you want to make your contribution please choose your correct branch based on the table below.

Fedora Rawhide

main

Fedora XX

fedora-XX

RHEL-X / CentOS Stream X

rhel-X

All of these branches are independent, never merged into each another, so if you want to put your changes into multiple branches, you have to open multiple pull requests.

Finding Bugs to Fix

The development team can mark bugs with specific keywords to show that they belong to a specific category. You can quickly list these by searching the Red Hat bugzilla for bugs in the anaconda component with specific keywords in Whiteboard:

(A single issue could potentially have more than one of these keywords.)

Patches for bugs without keywords are welcome, too!

Testing Anaconda changes

To test changes in Anaconda you have a few options based on what you need to do.

Backend and TUI development

There are two options to develop and test changes which are not yet released.

To find out more information about quick way to propagate your changes into the existing installation ISO image see this blogpost.

Another way is to build the boot.iso directly (takes more time but it’s easier to do). See the next section to find out how to build the ISO.

Building installation images

Building the ISO is the most precise way to find the behavior of Anaconda in the installation environment. However, it needs a lot of HW resources and time to build. During the build, you will be ask for sudo password. Unfortunately, it is required to run the build as root because the build process needs to work with /dev/loop devices. Please do not use toolbx or distrobox because the commands below are calling podman under root which is hard to achieve from inside of other container.

Follow these steps to build the ISO you need.

First build Anaconda RPM files with our container:

make -f ./Makefile.am container-rpms-scratch

Then build an image containing those RPMs.

NOTE: Do not run this in the Anaconda toolbox - it will not work due to the need for root privileges.

To build a regular boot.iso from these RPMs use (loop device mounting requires root privileges):

make -f ./Makefile.am anaconda-iso-creator-build # to build the container if it doesn't exists already
make -f ./Makefile.am container-iso-build

To build a Web UI boot.iso run:

make -f ./Makefile.am anaconda-iso-creator-build # to build the container if it doesn't exists already
make -f ./Makefile.am container-webui-iso-build

To build a Web UI in Live image run:

make -f ./Makefile.am anaconda-live-iso-creator-build # to build the container if it doesn't exists already
make -f ./Makefile.am container-live-iso-build

The resulting ISO will be stored in ./result/iso directory.

Note: You can put additional RPMs to ./result/build/01-rpm-build and these will be automatically used for the ISO build.

Local development workflow

This workflow makes it possible to test changes to the Anaconda source code locally on your machine without any dependencies on external infrastructure. It uses two scripts, one called scripts/testing/rebuild_iso to build a fresh bootable installation image from Anaconda source code on the given branch and corresponding Fedora/CentOS Stream packages. The second script, called scripts/testing/update_iso uses the Anaconda updates image mechanism together with the mkksiso command provided by the Lorax project to very quickly create an updated version of the boot.iso when Anaconda code is changed. The updated boot.iso can then be booted on a VM or bare metal.

The rebuild_iso script

This is just a simple script that rebuilds the boot.iso from Anaconda source code on the current branch & corresponding Fedora (on Fedora branches) or CentoOS Stream (on RHEL branches) packages. The script makes sure to remove the old images first and also records Anaconda Git revision that was used to build the image.

This should take about 15 minutes on modern hardware.

See –help for further information.

The update_iso script

This is the main script that enables local development by quickly updating a boot iso with local changes. This should take a couple seconds on modern hardware.

For the most common use case (“I have changed the Anaconda source and want to see what it does.”) just do this:

  1. run scripts/testing/rebuild_iso first, this creates result/iso/boot.iso

  2. change the Anaconda source code

  3. run scripts/testing/update_iso which creates the result/iso/updated_boot.iso

  4. start the result/iso/updated_boot.iso in a VM or on bare metal

The script also has a few command line options that might come handy:

  • -b, --boot-options makes it possible to add additional boot options to the boot.iso boot menu

  • -k, --ks-file add the specified kickstart file to the updated boot.iso and use it for installation

  • -v, --virt-install boot the updated iso in a temporary VM for super fast & simple debugging

  • -t, --tag use a specific Git revision when generating the updates image

  • You can specify custom ISO image (requirement for Live ISO usage) as optional positional parameter.

Running the updated boot.iso

The updated_boot.iso is just a regular bootable image, but there are a couple things to note:

  • Due to how mkksiso works the image will fail the image checksum test - so always use the first option in the image boot menu that skips the checksum verification.

  • Make sure to shut down VMs before booting them again after re-generating the updated_boot.iso file. Otherwise the VM software might continue using the previous file version & your changes might not be visible. There is also a dummy boot options added to updated_boot.iso called build_time that records when the currently running image has been updated. You can check this boot option either in the image boot menu or by checking /proc/cmdline on a running system.

Anaconda Installer Branching Policy (the long version)

The basic premise is that there are the following branches:

  • main

  • fedora-<next fedora number>

The main branch never waits for any release-related processes to take place and is used for Fedora Rawhide Anaconda builds.

Concerning current RHEL branches, they are too divergent to integrate into this scheme. Thus, commits are merged onto, and builds are done on the RHEL branches. In this case, multiple pull requests will very likely be needed:

  • one for the rhel<number>-branch

  • one for the main branch, if the change is not RHEL only

  • one for the fedora-<number> branch, if change should apply to branched Fedora too

Releases

The release process is as follows, for both Fedora Rawhide and branched Fedora versions:

  • a release commit is made (which bumps version in spec file) & tagged on the fedora-XX or main branch

Concerning the <next Fedora number> branches (which could also be called next stable release if we wanted to decouple our versioning from Fedora in the future):

  • work which goes into the next Fedora goes to fedora-<next Fedora number> and must have another PR for main, too

  • stuff we don’t want to go to the next Fedora (too cutting edge, etc.) goes only to main branch

  • commits specific to a given Fedora release (temporary fixes, etc.) go only to the fedora-<next Fedora number> branch

  • this way we can easily see what was developed in which Fedora timeframe and possibly due to given Fedora testing phase feedback (bugfixes, etc.)

Example for the F38 and F39 cycle

Once Fedora 38 is branched, we have these branches in the repository:

  • main

  • fedora-38

This would continue until f38 is released, after which we:

  • keep the fedora-38 branch as an inactive record of the f38 cycle

  • work on the main branch only

After a while, Fedora 39 is branched and we start the fedora-39 branch off the main branch.

This will result in the following branches for the f39 cycle:

  • main

  • fedora-39

Guidelines for Commits

Commit Messages

The first line should be a succinct description of what the commit does, starting with capital and ending without a period (‘.’). If your commit is fixing a bug in Red Hat’s bugzilla instance, you should add (#123456) to the end of the first line of the commit message. The next line should be blank, followed (optionally) by a more in-depth description of your changes. Here’s an example:

Stop kickstart when space check fails

Text mode kickstart behavior was inconsistent, it would allow an installation to continue even though the space check failed. Every other install method stops, letting the user add more space before continuing.

Commits for RHEL Branches

If you are submitting a patch for any rhel-branch, the last line of your commit must identify the JIRA issue id it fixes, using the Resolves, Related or Reverts keyword, e.g.: Resolves: RHEL-11111

or

Related: RHEL-12345

or

Reverts: RHEL-22222

Use Resolves if the patch fixes the core issue which caused the bug. Use Related if the patch fixes an ancillary issue that is related to, but might not actually fix the bug. Use Reverts if this patch reverts changes introduced by linked bug.

Release Notes

If you are submitting a patch that should be documented in the release notes, create a copy of the docs/release-notes/template.rst file, modify its content and add the new file to your patch, so it can be reviewed and merged together with your changes.

After a final release (for example, Fedora GA), we will remove all release notes from docs/release-notes/ of the release branch and add the content into the docs/release-notes.rst file.

This change will be ported to upstream to remove the already documented release notes from docs/release-notes/ of the upstream branch. In a case of RHEL, port only the new release file.

Pull Request Review

Please note that there is a minimum review period of 24 hours for any patch. The purpose of this rule is to ensure that all interested parties have an opportunity to review every patch. When posting a patch before or after a holiday break it is important to extend this period as appropriate.

All subsequent changes made to patches must be force-pushed to the PR branch before merging it into the main branch.

Code conventions

It is important to have consistency across the codebase. This won’t necessarily make your code work better, but it might help to make the codebase more understandable, easier to work with, and more pleasant to go through when doing a code review.

We use a set of linters (e.g., ruff, pylint) to automatically enforce code quality and style guidelines. These tools are used to gate changes, so it is highly recommended that you run the linters locally before submitting a pull request (PR) to catch any issues early.

You can run the ruff checks locally with:

make TESTS=ruff/run_ruff.sh check

You can run the pylint checks locally with:

make TESTS=pylint/runpylint.py check

In general, we aim to stay as close as possible to PEP8, while extending or adjusting minor rules to suit the context of our project. The following conventions supplement the rules enforced by our linters:

  • Format strings with .format() instead of % (https://pyformat.info/)

  • Follow docstring conventions. See PEP257.

  • Use Enum instead of constants is recommended.

  • Use super() instead of super(ParentClass, self).

  • Use ParentClass.method(self) only in case of multiple inheritance.

  • Instance variables are preferred, class variables should be used only with a good reason.

  • Global instances and singletons should be used only with a good reason.

  • Never do wildcard (from foo import *) imports with the exception when all Anaconda developers agree on that.

  • Use raise & return in the doc string. Do not use raises or returns.

  • Methods that return a task should have the suffix ‘_with_task’ (for example discover_with_task and DiscoverWithTask).

  • Prefer to use pyanaconda.util.join_paths over os.path.join. See documentation for more info.

  • Never call upper() on translated strings. See the bug 1619530

  • Names of signal handlers defined in .glade files should have the on_ prefix.

Merging examples

Merging a GitHub pull request

(Fedora 38 is used as an example, don’t forget to use appropriate Fedora version.)

Press the green Merge pull request button on the pull request page.

Then you are done.

Merging a topic branch manually

(Fedora 38 is used as an example, don’t forget to use appropriate Fedora version.)

Let’s say that there is a topic branch called “fix_foo_with_bar” that should be merged to a given Anaconda non-topic branch.

Checkout the given target branch, pull it and merge your topic branch into it:

git checkout <target branch>
git pull
git merge --no-ff fix_foo_with_bar

Then push the merge to the remote:

git push origin <target branch>

If the pull request has been opened for the fedora-38 branch, then you also need to check if the same change should go to the main branch in another PR.

Pure community features

The pure community features are features which are part of the Anaconda code base but they are maintained and extended mainly by the community. These features are not a priority for the Anaconda project.

In case of issues in pure community features, the Anaconda team will provide only sanity checking. It is the responsibility of the community (maintainers of the feature) to provide fix for the issue. If the issue will have bigger impact on other parts of the Anaconda project or if it will block a release or another priority feature and the fix won’t be provided in a reasonable time the Anaconda team reserves the rights to remove or disable this feature from the Anaconda code base.

Below is a list of pure community features, their community maintainers, and maintainers contact information:

/boot on btrfs subvolume

Enable boot of the installed system from a BTRFS subvolume.

systemd-boot as a bootloader

Enable boot using systemd-boot rather than grub2.