conda-forge Maintenance

The end goal of GOATS is eventual distribution as a conda-forge package. To achieve this, all dependencies must first be available on conda-forge.

As part of this process, the GOATS project contributes to the broader Python astronomy ecosystem by maintaining and co-maintaining several conda-forge feedstocks.

Currently, the project maintains:

  • 5 conda-forge packages with sole maintainership,

  • 2 conda-forge packages with shared maintainership, and

  • 1 package (DRAGONS) that is awaiting conda-forge support and currently prevents full inclusion of GOATS on conda-forge.

Sole-Maintained Packages

Feedstocks fully maintained under the GOATS organization:

Note

Feedstock repositories are automatically created and hosted under the conda-forge GitHub organization. Changes to recipes are submitted as pull requests to these repositories (for example, django-dramatiq-feedstock).

Co-Maintained Packages

Feedstocks with shared maintainership:

The TOMToolkit team typically manages releases and feedstock updates for these packages. If a release requires attention or coordination, contact with the TOMToolkit maintainers is recommended.

Pending Packages

The remaining dependency that currently prevents GOATS from becoming a fully supported conda-forge package is DRAGONS and its related sub-dependencies:

During recent meetings, it was discussed that DRAGONS is being considered for addition to conda-forge. This topic will be revisited following the release of version 4.1.0.

Tracking and Coordination

All work related to maintaining and supporting the conda-forge effort for GOATS is organized under the Epic GOATS-952: conda-forge Maintenance. Issues, tasks, and subtickets related to feedstock updates, conda maintenance, or package onboarding should be created under this epic.

Note

All new issues or updates related to any feedstock should be logged as subtasks under GOATS-952 in Jira.

Maintaining a conda-forge Feedstock

The official conda-forge maintainer documentation provides the authoritative reference for feedstock maintenance and update procedures.

Note

To perform any maintenance actions on a conda-forge feedstock, such as merging pull requests, rerendering recipes, or triggering rebuilds, the maintainer must first be listed in the recipe-maintainers section of the feedstock’s meta.yaml file.

When a new release of a maintained package is published, conda-forge automation tags maintainers on the corresponding pull request in the feedstock repository. Responsibilities include:

  1. Reviewing the tagged pull request and verifying that the upstream release matches the feedstock version and metadata.

  2. Confirming that dependency versions are correct and compatible.

  3. Ensuring that continuous integration builds (Linux, macOS, Windows) complete successfully.

  4. Merging the pull request once all checks have passed.

In most cases, dependencies remain stable between releases. When upstream dependencies change, the meta.yaml recipe must be updated accordingly, local builds verified if necessary, and tests confirmed to pass before merging.

Note

When troubleshooting CI build failures, check the conda-forge build logs for each platform (Linux, macOS, Windows). These logs often indicate missing dependencies, version mismatches, or syntax issues in meta.yaml.

Internal Package Hosting

Because GOATS cannot currently be distributed through conda-forge until all dependencies are supported there, an internal build and hosting mechanism is used to maintain package availability.

All internal builds are managed through the goats-infra repository. This repository contains the build recipes, GitHub Actions workflows, and configuration for producing and publishing GOATS and related packages as Conda artifacts.

Completed builds are automatically published to the goats-infra GitHub Pages Conda channel.

This channel is configured as a Conda repository and can be added to a local environment with:

conda config --add channels https://gemini-hlsw.github.io/goats-infra/conda/

This workflow ensures that GOATS and its ecosystem can continue to be installed, tested, and deployed while work toward full conda-forge inclusion is ongoing.