Parallel Documentation

Appearance

Test Stability Fixes ​

Issue ​

Random test failures in the screenflow service where tests would be killed intermittently during execution. The process would show "Killed" without completing all test suites.

Root Causes Identified ​

1. Memory Leaks from Unclosed NestJS Test Modules ​

Problem: Test files using Test.createTestingModule() were not calling module.close() in afterEach hooks, causing memory leaks that accumulated over time.

Impact: As tests ran, memory usage increased until the OS killed the process to prevent system instability.

Affected Files:

  • All use-case test files in services/screenflow/src/modules/application-layer/use-cases/
  • Infrastructure service tests in services/screenflow/src/modules/infrastructure-layer/guacamole/
  • Package tests in packages/nest/src/framework/

2. Unhandled setTimeout Calls ​

Problem: Some tests used setTimeout() without properly awaiting them, causing assertions to run after the test had already completed.

Impact: Tests would pass or fail randomly depending on timing, and could leave pending timers that interfere with subsequent tests.

Affected Files:

  • services/screenflow/src/modules/domain-layer/aggregates/resource/resource.aggregate.spec.ts
  • services/screenflow/src/modules/domain-layer/entities/instance-definition/instance-definition.entity.spec.ts

Solutions Applied ​

Fix 1: Added Proper Test Module Cleanup ​

Added afterEach hooks to properly close NestJS test modules:

typescript
describe('SomeUseCase', () => {
    let useCase: SomeUseCase
    let module: TestingModule // Added module variable
    
    beforeEach(async () => {
        module = await Test.createTestingModule({ // Changed from const
            providers: [...]
        }).compile()
        
        useCase = module.get<SomeUseCase>(SomeUseCase)
    })
    
    afterEach(async () => {
        await module?.close() // Added cleanup
    })
})

Files Fixed:

  • services/screenflow/src/modules/application-layer/use-cases/aggregates/resource/start-instance/start-instance.use-case.spec.ts
  • services/screenflow/src/modules/application-layer/use-cases/aggregates/resource/delete-instance/delete-instance.use-case.spec.ts
  • services/screenflow/src/modules/application-layer/use-cases/aggregates/resource/delete-instance-definition/delete-instance-definition.use-case.spec.ts
  • services/screenflow/src/modules/application-layer/use-cases/aggregates/resource/update-instance/update-instance.use-case.spec.ts
  • services/screenflow/src/modules/application-layer/use-cases/aggregates/resource/update-instance-definition/update-instance-definition.use-case.spec.ts
  • services/screenflow/src/modules/infrastructure-layer/guacamole/input-classifier/guacamole-input-classifier.service.spec.ts
  • services/screenflow/src/modules/infrastructure-layer/guacamole/guacamole-parser/guacamole-parser.service.spec.ts
  • packages/nest/src/framework/log/logger.service.test.ts
  • packages/nest/src/framework/langchain/langchain-hub.service.spec.ts

Fix 2: Properly Awaited setTimeout Calls ​

Changed unhandled setTimeout to properly awaited promises:

Before:

typescript
it('should handle multiple delete calls', () => {
    resource.deleteInstanceDefinition()
    const firstDeleteTime = resource.instanceDefinition.getDeletedAt()
    
    setTimeout(() => {
        resource.deleteInstanceDefinition()
        const secondDeleteTime = resource.instanceDefinition.getDeletedAt()
        expect(secondDeleteTime?.getTime()).toBeGreaterThan(firstDeleteTime?.getTime() || 0)
    }, 1)
})

After:

typescript
it('should handle multiple delete calls', async () => {
    resource.deleteInstanceDefinition()
    const firstDeleteTime = resource.instanceDefinition.getDeletedAt()
    
    await new Promise((resolve) => setTimeout(resolve, 10))
    
    resource.deleteInstanceDefinition()
    const secondDeleteTime = resource.instanceDefinition.getDeletedAt()
    expect(secondDeleteTime?.getTime()).toBeGreaterThan(firstDeleteTime?.getTime() || 0)
})

Files Fixed:

  • services/screenflow/src/modules/domain-layer/aggregates/resource/resource.aggregate.spec.ts (line 524)
  • services/screenflow/src/modules/domain-layer/entities/instance-definition/instance-definition.entity.spec.ts (line 159)

Test Results ​

Before Fixes ​

PASS screenflow src/modules/domain-layer/aggregates/procedure/procedure.aggregate.spec.ts (8.974 s)
PASS screenflow src/modules/domain-layer/aggregates/resource/resource.aggregate.spec.ts
PASS screenflow src/modules/domain-layer/entities/instance-definition/instance-definition.entity.spec.ts
PASS screenflow src/modules/domain-layer/entities/instance/instance.entity.spec.ts
PASS screenflow src/modules/infrastructure-layer/guacamole/guacamole-parser/guacamole-parser.service.spec.ts
Killed
Warning: command "jest" exited with non-zero status code

After Fixes ​

Test Suites: 12 passed, 12 total
Tests:       176 passed, 176 total
Time:        2.451 s - 7.237 s (varies)
Ran all test suites.

✓ Ran 3 consecutive times without failures
✓ All 4 projects (screenflow, coding, nest, valuation) passing

Best Practices for Future Tests ​

1. Always Clean Up NestJS Modules ​

typescript
let module: TestingModule

beforeEach(async () => {
    module = await Test.createTestingModule({...}).compile()
})

afterEach(async () => {
    await module?.close()
})

2. Never Use Unhandled Timers in Tests ​

typescript
// ❌ Bad
setTimeout(() => { /* test assertions */ }, 100)

// ✅ Good
await new Promise((resolve) => setTimeout(resolve, 100))
/* test assertions */

3. Use Async/Await for All Async Tests ​

typescript
// ❌ Bad
it('test name', () => {
    setTimeout(...)
})

// ✅ Good
it('test name', async () => {
    await new Promise(...)
})

The Jest configuration already had good settings to help prevent issues:

services/screenflow/jest.config.ts:

  • testTimeout: 60000 - 60 second timeout for individual tests
  • maxWorkers: 1 - Run tests sequentially to avoid memory issues

jest.preset.js:

  • testTimeout: 30000 - 30 second default timeout
  • maxWorkers: '50%' - Use half of available cores

Impact ​

  • Stability: Tests now run consistently without random failures
  • Performance: Faster test runs (2-7s vs previously hanging/killed)
  • Memory: No memory leaks from unclosed modules
  • Reliability: Can be run multiple times consecutively without issues

Verification ​

Run tests multiple times to verify stability:

bash
# Run screenflow tests 3 times
for i in {1..3}; do nx run screenflow:test --skip-nx-cache; done

# Run all project tests
nx run-many --target=test --skip-nx-cache --projects=screenflow,coding,nest,valuation