Implement unit tests for Test2::Harness core lifecycle (12 files)

[Koan-Bot]

Context

The t/unit/Test2/Harness/ directory contains 12 test stubs for the core harness lifecycle — the instance orchestrator, run management, runner execution, and test settings. These are the heart of the 2.0 multi-process architecture.

Modules to test

Harness entry:

• t/unit/Test2/Harness.t → Test2::Harness

Instance (orchestrator):

• t/unit/Test2/Harness/Instance.t → Test2::Harness::Instance
• t/unit/Test2/Harness/Instance/Message.t, Instance/Request.t, Instance/Response.t

Run & Job:

• t/unit/Test2/Harness/Run.t → Test2::Harness::Run
• t/unit/Test2/Harness/Run/Job.t → Test2::Harness::Run::Job

Runner:

• t/unit/Test2/Harness/Runner.t → Test2::Harness::Runner
• t/unit/Test2/Harness/Runner/Preloading.t, Runner/Preloading/Stage.t

Test configuration:

• t/unit/Test2/Harness/TestFile.t → Test2::Harness::TestFile
• t/unit/Test2/Harness/TestSettings.t → Test2::Harness::TestSettings

Suggested test scenarios

Instance (orchestrator)

• run(cb) — callback execution, error handling
• stop() — graceful shutdown
• handle_request(request) — request routing and response
• terminate(reason) — termination with reason tracking (first-reason-wins semantics)
• api_abort() / api_kill() — cascading abort/kill to scheduler + runner

Run & Job

• Run: set_ipc() / ipc(), run_id generation, abort_on_bail attribute (default true), halt state transitions
• Job: job data construction, environment setup, test file association

Runner

• ready() — readiness check
• start() / stop() / abort() / reload() — lifecycle methods
• job_launch_data() — job execution data preparation
• skip_job() — skip logic
• launch_job() — job forking and execution
• is_daemon attribute behavior

Runner::Preloading

• Stage management, module preloading, stage-based test routing

TestFile & TestSettings

• TestFile: test file metadata extraction, category/duration detection
• TestSettings: settings composition, per-test overrides

Notes

These are high-value tests — the Instance/Runner/Scheduler triad is the core of yath 2.0's multi-process architecture. Mock the IPC layer (use simple in-process message passing) to test orchestration logic without actual forking.

The terminate() method's first-reason-wins semantics is critical — test concurrent termination attempts.

[atoomic]

@Koan-Bot fix