feat: add xtask for parameterized task workflows

[krishicks]

Summary

This commit adds a cargo xtask framework for orchestrating complex development workflows with typed parameter validation and tested composition.

The initial implementation provides two tasks:

• e2e runs Podman end-to-end suites locally or in a Lima-managed Linux machine.
• release-smoke-test installs a supplied Debian package in a Lima machine, creates a sandbox, and verifies a policy-denied curl request.

Machine-backed e2e testing supports Ubuntu 24.04, Ubuntu 26.04, and CentOS Stream 10. Each target is prepared with the OpenShell development environment and rootless Podman, with target-specific setup for packages and container configuration. CentOS Stream coverage runs under enforcing SELinux, labels shared bind mounts correctly, and checks the audit log for unexpected denials.

The Lima provider manages machine lifecycle, persistent ext4 build disks, prepared snapshots for faster reruns, relevant GitHub credentials, and cleanup of incomplete machines. Explicitly retained machines remain available for inspection.

Rust models operating systems, suites, providers, resources, snapshot identities, and machine lifecycle, while focused shell scripts handle guest setup. This structure allows a single parameterized command to exercise an expanding Linux matrix without encoding conditional behavior in the mise task graph.

GitHub Actions does not invoke these tasks as part of this change.

Additional background

There is a desire to run tasks against a matrix of environments, for example we should be able to run the e2e tests for Podman against a variety of Linux environments: Ubuntu with Podman 4.x, 5.x, and 6.x, CentOS Stream 10 with SELinux enabled, etc.

Our current mise task graph models many variants as separate named tasks, with the actual parameter handling and conditional behavior pushed into shell scripts. Although mise supports parameterized tasks, implementing this matrix there would encode conditional behavior in the mise task graph. Encoding it as an xtask lets us model that behavior with reusable Rust types and unit tests.

cargo xtask adds orchestration, parameter validation, and composition in typed and tested Rust, while retaining focused shell scripts for guest setup and existing suite execution. This gives us the ability to have a single, parameterized task that can handle, for example, spinning up an Ubuntu VM, preparing it appropriately, running a test in it, and shutting it down, with optional behavior like enabling snapshots for faster feedback loops.

Usage

Details

E2E tests

# Run the Podman e2e suite locally
cargo xtask e2e --suite podman

# Run it in a Lima machine using the default Ubuntu 26.04 guest
cargo xtask e2e --suite podman --provider lima

# Run one test on Ubuntu 24.04 with a reusable snapshot
cargo xtask e2e \
--suite podman \
--test sandbox_mounts_existing_driver_config_volume \
--provider lima \
--os ubuntu-24.04 \
--snapshot

Release smoke tests

# Test a package using Lima and Ubuntu 26.04; architecture is inferred from its filename
cargo xtask release-smoke-test \
--deb ~/Downloads/openshell_0.0.80-1_arm64.deb

# Test explicitly on Ubuntu 24.04 and ARM64
cargo xtask release-smoke-test \
--deb ~/Downloads/openshell_0.0.80-1_arm64.deb \
--os ubuntu-24.04 \
--arch arm64

# Reuse a prepared Ubuntu 26.04 machine snapshot across runs
cargo xtask release-smoke-test \
--deb ~/Downloads/openshell_0.0.80-1_arm64.deb \
--snapshot

--keep-machine can be added to either command to leave the machine available for inspection.

Related Issue

• Use a unified mechanism for build, test, packaging, and release #2204

Changes

Testing

• mise run pre-commit passes
• Unit tests added/updated
• E2E tests added/updated (if applicable)

Checklist

• Follows Conventional Commits
• Commits are signed off (DCO)
• Architecture docs updated (if applicable)

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.