Introducing modulemd-tools

01. 07. 2020 | Jakub Kadlčík | EN dev copr fedora modularity

If you have been following news and innovations in the Fedora world for the past few years, the Fedora Modularity project couldn’t escape your eye. It has been a subject of many discussions, opinions, and critique. I am pleased to tell you, that the biggest pain points (at least for me) are finally being addressed.

Background

A lot of teams are involved in the development of Fedora Modularity and vastly more people are affected by it as packagers and end-users. It is obvious, that each group has its own priorities, use-cases and therefore different opinions on what is good or bad about the current state of the project. Personally, I was privileged (or maybe doomed) to represent yet another, often forgotten, group of users - third-party build systems.

Our team is directly responsible for the development and maintenance of Copr and a few years ago we decided to support building modules alongside building just regular packages. We stumbled upon many frustrating pitfalls that I don’t want to discuss right now but the major one was definitely not enough tools for working with modules. That was understandable in the early stages of the development process but it has been years and we still don’t have the right tools for building modules on our own, without relying on the Fedora infrastructure. You may recall me expressing the need for them at the Flock 2019 conference.

Well, brace yourselves, the change is coming. We are now introducing modulemd-tools.

Modulemd-tools

The simple fact is, that module YAML definitions are meant to be (partially) generated, and up until now there was no tool for doing so. That affected everyone building modules on their own. We are now introducing a project called modulemd-tools which is a collection of small tools for parsing and generating modulemd YAML files. It is packaged for Fedora now so you can easily install it.

dnf install modulemd-tools

So far it provides two scripts and it is up to you to pick which suits you best.

dir2module

This script takes essential module information such as its name, stream, version, architecture, etc via the command line parameters. Then it either recursively finds all RPM packages within a specified directory or reads a list of them from a text file. As a result, it generates a module definitions to a YAML file or to the standard output.

A simple usage might look like this.

$ dir2module foo:devel:123:f32:x86_64 -m "Summary and stuff" --dir .

Created foo:devel:123:f32:x86_64.modulemd.yaml

The first positional parameter is a string in NSVCA format, for more information please see NSVCA vs NEVRA. Now, the more complicated example specifying a license and requirements and what not can look like this.

$ dir2module foo:devel:123:f32:x86_64 -m "Summary and stuff" --dir ~/rpmbuild/RPMS/ -l GPLv2 -r mod1:stream1 -r mod2:stream2 --force --stdout

---
document: modulemd
version: 2
data:
  name: foo
  stream: devel
# ...

repo2module

The point of repo2module is the same as for the previous script - it generates a module YAML definition based on some input parameters. The difference is that it doesn’t work on a directory-level but rather takes an existing YUM repository as its input.

$ repo2module --module-name=testmodule --module-stream=stable . modules.yaml

What’s next?

That’s up to you. What specific tasks regarding modulemd YAML generation or parsing do you do? Share them with us. Besides that, on our current roadmap is to implement modularity support into createrepo_c and then replace all the custom code in Copr by using these tools and help other build system developers to do the same. In the endgame, all Copr, OBS, and MBS will internally use the same tooling, and therefore it is going to be safe for you, users, to start incorporating them into your workflows as well because they will become de-facto a standard.