Add Docker-based integration test suite

[Copilot]

• Explore repo structure and understand existing tests and module
• Create Tests/Integration/docker-compose.yml — Snipe-IT + MySQL stack for testing
• Create Tests/Integration/setup.php — PHP bootstrap script to create admin user with API token
• Create Tests/Integration/Start-TestEnvironment.ps1 — start Docker, wait for health, provision admin, return {Url, Token}
• Create Tests/Integration/Stop-TestEnvironment.ps1 — tear-down Docker stack with --volumes --remove-orphans
• Create Tests/Integration/SnipeIT.Integration.Tests.ps1 — 52-test Pester file covering CRUD + checkout/checkin
• Create .github/workflows/integration.yml — GitHub Actions workflow for integration tests
• Fix unit CI failures: changed path from ./Tests to ./Tests/Unit; improved docker-compose.yml snipeit healthcheck to check API returns 401
• Fix integration test timeout:
  • docker-compose.yml: changed MySQL healthcheck from -h localhost (Unix socket) to -h 127.0.0.1 -P 3306 (TCP) — ensures MySQL is accepting TCP connections (not just socket) before Snipe-IT starts, preventing Connection refused during migrations
  • integration.yml: replaced docker-health-status poll with a direct HTTP poll for Apache (http://localhost:8000/), followed by an explicit php artisan migrate --force step (idempotent — fixes startup if migration failed), then a second poll for the API returning 401

[djeborn]

@copilot - Can you go fix the errors there?

[Copilot]

@copilot - Can you go fix the errors there?

Fixed in commit 2fc8ad6. Two issues were causing CI failures:

1. CI workflow picked up integration tests — $config.Run.Path was set to ./Tests which includes Tests/Integration/. On Ubuntu runners Docker is available, so the integration tests tried to start the stack. On Windows they crashed the test container entirely. Fixed by changing the path to ./Tests/Unit.

2. Docker healthcheck passed too early — the original check (curl -sf http://localhost/) passed as soon as Apache started, before DB migrations finished. The provision step would then run setup.php against a DB with no tables, producing the Table 'snipeit.users' doesn't exist errors. Fixed by checking that the /api/v1/hardware API endpoint returns 401 (Unauthenticated), which only happens once Laravel migrations have completed and the app is fully ready.

[djeborn]

@copilot - Fix the integration test error.

[Copilot]

@copilot - Fix the integration test error.

Fixed in commit a12b89b. Two root causes:

1. MySQL healthcheck used Unix socket, not TCP — the check used -h localhost which verifies the socket inside the mysql container, but Snipe-IT connects via TCP to mysql:3306. The socket can be ready before TCP, so Snipe-IT got Connection refused during migrations. Fixed by changing to -h 127.0.0.1 -P 3306 (explicit TCP).

2. Migrations only ran once at startup and were never retried — when migrations failed, Apache started anyway with an empty DB. The health check waited for the API to return 401, but with no tables it would never 401, causing a 6-minute timeout. Fixed the workflow with three distinct steps:

   • Wait for Apache to respond on port 8000 (HTTP poll, no longer relying on docker health state)
   • Explicitly run php artisan migrate --force (idempotent — no-op if migrations already ran, fix if they failed)
   • Wait for /api/v1/hardware to return 401 before provisioning