Source-git

Source-git documentation #

We initiated source-git as an alternative to working in dist-git. The need was especially amplified after we realized that many projects already do this. The problem was that everyone created a workflow and tooling for themselves which made it really hard to share knowledge and on-board new projects. Our intention is to create an implementation which is highly configurable, easy to start with and provides both command-line tooling and a service which is tightly integrated with GitHub and GitLab.

Before you continue reading, we assume you are familiar with the Fedora Maintenance Guide.

Source-git is a git repository which contains upstream sources and downstream-specific files: for packaging and code changes specific to a distribution, in our case Fedora Linux, CentOS Stream and Red Hat Enterprise Linux.

The basic requirement on a source-git repo is that one needs to be able to create an SRPM out of the current checkout. We need several files to achieve such a thing:

  • RPM Spec file

  • Archive with sources

  • Downstream patches to be applied in the spec file

Source-git example

Spec file #

We are often being asked a question about spec files: “Do we really need to have a spec file in our upstream repo or a source-git repo?!” No, you don’t, though storing the spec file directly in the repository will save you a lot of trouble - when you disconnect the repository from the build recipe, the spec file:

  • Adding or changing a build dependency will require changing the spec file in the other location as well.
  • Someone may change the remote spec file and make builds fail suddenly.
  • Debugging build failures will be a nightmare because you simply can’t edit the spec file easily.

Archive #

Packit downloads the archive defined as Source or Source0 in your spec file to produce a SRPM.

Downstream patch generation process #

The important part of the patch generation process is for Packit to know, from which commit it should start generating patches. The theory here is that you mark a certain commit and it should resemble the content of the archive. All the code changes on top are treated as downstream changes and the respective patch files are generated. The marked commit usually represents a specific upstream release. This means that all the generated patches need to be able to be applied during the build process.

upstream_ref is this border between upstream commits, usually matching an upstream release tag, and downstream commits. Packit iterates over every git commit sitting on top off upstream_ref.

Packit distinguishes between 2 types of git commits:

  1. Those that change code (usually pulled from upstream) - Packit converts these into patch files so they can be applied to the source tree in the %prep section during the build process.
  2. Changes to packaging (spec file, other sources, changes to .packit.yaml).

Downstream patches to be applied in the spec file #

In the downstream, it’s a common workflow to pick up upstream patches from the main branch, backport them on top of the current version in Fedora, CentOS Stream or RHEL and apply the patches during the build process.

Source-git is perfect for this because patches are stored as git commits and patch files are generated on the fly.

This was just a gentle intro to source-git, for more info you can continue with: