Integration testing and edge case handling #11777

Parent Epic

Part of #11762 (Ink-based TUI Rewrite)

Objective

Comprehensive testing of the Ink implementation, especially edge cases that caused issues in the previous implementation.

Critical Test Scenarios

1. Concurrent Updates (The Main Problem)

• Prose streaming + tool status updates simultaneously
• Multiple tools running in parallel
• Task with child tool + prose streaming
• Rapid state changes don't cause flicker

2. Child Session Events

• Events arrive before task metadata (buffering)
• Events arrive after task completes (ignored)
• Child session ID mapping works correctly
• Multiple tasks with children simultaneously

3. Message Lifecycle

• User input → streaming → completion → Static
• Cancel during streaming
• Error during streaming
• Empty response handling

4. Session Management

• Create new session
• Switch between sessions
• Session state preserved on switch
• Agent/model changes persist

5. Input Handling

• Long input lines
• Multi-line input (if supported)
• Special characters
• History navigation
• Rapid typing during streaming

6. Terminal Edge Cases

• Very narrow terminal (< 40 cols)
• Very wide terminal (> 200 cols)
• Terminal resize during operation
• Non-TTY environment (piped input)

Test Structure

// test/cli/ink/integration.test.tsx
import { render } from 'ink-testing-library'
import App from '../src/cli/ink/App'

describe('Ink TUI Integration', () => {
  describe('Concurrent Updates', () => {
    it('handles prose + tool updates without crash', async () => {
      const { lastFrame } = render(<App />)
      // Simulate concurrent events
      // Assert no crash, correct display
    })
  })
  
  describe('Child Session Events', () => {
    it('buffers events arriving before mapping', async () => {
      // ...
    })
  })
})

Tasks

• Set up ink-testing-library
• Create test fixtures for SDK events
• Write concurrent update tests
• Write child session event tests
• Write message lifecycle tests
• Write session management tests
• Write input handling tests
• Write terminal edge case tests
• Manual testing checklist

Manual Testing Checklist

• Fresh install and run
• /start command with real agent
• Multiple message exchanges
• Task spawning (e.g., ask PM to research something)
• Cancel mid-stream
• Session switching
• Agent cycling
• Long-running operations

Acceptance Criteria

• All automated tests pass
• Manual testing checklist complete
• No crashes in normal usage
• No crashes in edge cases
• Performance acceptable (no visible lag)

Estimated Effort

3.5 days

Risk

Medium - Testing concurrent behavior is inherently complex

Dependencies

• All other implementation issues complete