Functional testing 

Running functional tests 

# Run TYPO3 functional tests
Build/Scripts/runTests.sh -s functional

# Alternative: Via Composer script
composer ci:test:php:functional

With the database configured, unexpected TYPO3 initialization failures and teardown integrity failures fail the suite. The shared Tests/Functional/AbstractFunctionalTestCase.php skips only when its database-configuration check finds no configuration; a failed bootstrap is not a successful test run. Parent teardown runs after setup completes and its failures remain visible.

The Unit suite checks this behavior using bounded child-PHPUnit executions: an invalid package fails initialization, a leaked error-reporting setting fails teardown, a clean lifecycle executes its assertion, and absent database configuration produces the deliberate skip. The deliberately failing child fixtures live outside the normal Functional discovery directory. Each child uses an owned instance identifier through the testing framework's protected instance seam. Concurrent success and failed-bootstrap children are checked together; the parent stops any unfinished child and removes only its owned instance directories, including after a child fails.

Functional test example 

<?php

namespace Netresearch\NrLlm\Tests\Functional\Repository;

use Netresearch\NrLlm\Domain\Model\Provider;
use Netresearch\NrLlm\Domain\Repository\ProviderRepository;
use TYPO3\TestingFramework\Core\Functional\FunctionalTestCase;

class ProviderRepositoryTest extends FunctionalTestCase
{
    protected array $testExtensionsToLoad = [
        'netresearch/nr-vault',
        'netresearch/nr-llm',
    ];

    private ProviderRepository $repository;

    protected function setUp(): void
    {
        parent::setUp();
        $this->repository = $this->get(ProviderRepository::class);
    }

    public function testFindOneByIdentifierReturnsProvider(): void
    {
        $this->importCSVDataSet(__DIR__ . '/Fixtures/providers.csv');

        $provider = $this->repository->findOneByIdentifier('openai-test');

        $this->assertInstanceOf(Provider::class, $provider);
        $this->assertEquals('OpenAI Test', $provider->getName());
    }
}

Test fixtures 

"tx_nrllm_provider"
"uid","pid","identifier","name","adapter_type","is_active"
1,0,"openai-test","OpenAI Test","openai",1

{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1677652288,
  "model": "gpt-5",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Test response"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 5,
    "total_tokens": 15
  }
}

Best practices 

1. Isolate tests: Each test should be independent.
2. Mock external APIs: Never call real APIs in unit tests.
3. Use data providers: For testing multiple scenarios.
4. Test edge cases: Empty inputs, null values, boundaries.
5. Descriptive names: Test method names should describe behavior.
6. Arrange-Act-Assert: Follow AAA pattern.
7. Fast tests: Unit tests should complete in milliseconds.
8. Coverage goals: Aim for >80% line coverage.