ADR-006: Test Infrastructure Overhaul

Status: Accepted Date: 2026-05-20 Updated: 2026-07-14 Deciders: (to be assigned)

Implementation Status

Phases 1 through 3 were completed on 2026-07-13:

• Every installed concrete project model has a domain-grouped factory.
• A live registry contract detects missing and duplicate factory adapters.
• Each factory creates two independently valid, persisted instances and passes full_clean().
• Core-only, optional multitenancy, and swappable-user host configurations are covered.
• Factory Boy remains outside the base runtime dependencies, and test support is excluded from wheels.
• Discovery, deduplication, hardware, location, performer, alert, and connection services have direct behavioral coverage at 90% or higher per targeted module.
• Query budgets protect discovery batching, alert fanout, connection health, and connection statistics from data-dependent query growth.
• Request-level smoke tests cover tenant-scoped chassis administration, atomic discovery approval, the monitoring dashboard, settings diffs, and HTMX channel fragments.

The coverage expansion completed on 2026-07-14. CI now enforces a synchronized 95% branch-coverage floor across every distributable Python module, backed by service, integration, command, admin, and host-configuration contracts.

Context

At the time of this decision, the test suite at tests/ had 13 files (11 test modules plus configuration). Coverage analysis showed:

• 33 of 38 service files have zero tests.
• No test factories exist (raw model instances constructed inline).
• No integration tests (service + model + DB).
• No E2E tests (admin views, HTMX flows, discovery pipeline).
• No fixture factories — tests construct model instances manually with inline kwargs.
• The flat tests/ directory does not mirror the domain structure of the app.

This means:

• Service-layer changes have no automated safety net.
• Refactoring (ADR-001 through ADR-005) is high-risk without test coverage.
• CI provides no regression detection.

Decision

1. Adopt factory_boy for model factories. Create one factory per model under tests/factories/.
2. Restructure tests from flat layout to domain-aligned:

   tests/
     conftest.py
     settings.py
     urls.py
     factories/
       __init__.py
       hardware.py
       discovery.py
       monitoring.py
       settings.py
       ...
     services/
       test_discovery_service.py
       test_hardware_service.py
       test_alert_service.py
       ...
     models/
       test_wireless_chassis.py
       test_rf_channel.py
       ...
     admin/
       test_dashboard_views.py
       ...
     integration/
       test_discovery_pipeline.py
       test_sync_lifecycle.py
       ...

3. Target: service-layer coverage ≥80% per module before refactoring. Write tests for every service method that has non-trivial branching.
4. Add integration tests for the discovery pipeline (scan → queue → adopt), sync lifecycle (poll → update → alert), and settings resolution chain.
5. Add smoke/E2E tests for critical admin flows using Django’s test client (chassis CRUD, discovery approval, monitoring dashboard).

Consequences

• Positive: Refactoring becomes safe. New contributors have visible patterns for writing tests. CI catches regressions. The factory layer reduces test boilerplate by ~70%.
• Negative: Adding ~300-400 tests requires dedicated sprints. Factory maintenance adds overhead when model schemas change.
• Migration: Phase 1 — factories for all models (complete). Phase 2 — service tests for discovery, hardware, monitoring domains (complete). Phase 3 — admin E2E smoke tests (complete). Phase 4 — discovery and poll-to-alert integration pipelines (complete).

Compliance

• CI enforces a synchronized 95% branch-coverage floor and rejects omitted distributable modules.
• New service methods require tests in the same PR.