---
phase: 9.1-fix-flaky-test-when-ran-in-parallel
plan: 01
type: execute
---
Fix test suite flakiness where 5 tests fail in parallel execution but pass individually.
Purpose: Ensure reliable CI/CD testing by resolving Swift Testing parallel execution state pollution discovered in Phase 9.
Output: All 5 flaky tests pass consistently in full parallel test suite execution.
~/.claude/get-shit-done/workflows/execute-phase.md
~/.claude/get-shit-done/templates/summary.md
@.planning/PROJECT.md
@.planning/ROADMAP.md
@.planning/STATE.md
@.planning/codebase/TESTING.md
# Prior phase context
@.planning/phases/09-trip-planner-modes-tdd/09-01-SUMMARY.md
@.planning/phases/09-trip-planner-modes-tdd/09-02-SUMMARY.md
@.planning/phases/09-trip-planner-modes-tdd/09-03-SUMMARY.md
# Test files with flaky tests
@SportsTimeTests/ScenarioAPlannerSwiftTests.swift
@SportsTimeTests/ScenarioBPlannerTests.swift
@SportsTimeTests/ScenarioCPlannerTests.swift
**Flaky tests identified (5 total):**
From ScenarioAPlannerSwiftTests.swift (3 tests):
- `plan_StopDepartureDate_IsLastGameDate()` (line ~300)
- `plan_ManyGames_HandledEfficiently()` (line ~493)
- `plan_ThreeSameDayGames_PicksFeasibleCombinations()` (line ~944)
From ScenarioBPlannerTests.swift (2 tests):
- `plan_FillerSameDayAsAnchor_Excluded()`
- `plan_MustSeeGamesTooFarApart_Fails()`
From ScenarioCPlannerTests.swift (1 test):
- `corridor_MultipleGamesMixed_FiltersCorrectly()`
**Root cause:** Swift Testing runs tests in parallel by default. Tests likely share mutable state (actor instances, simulator state) causing interference.
**Tech stack available:**
- Swift Testing framework (iOS 26+)
- `#expect()` assertions
- `@Suite` and `@Test` attributes
- `.serialized` trait for controlling parallelization
- `confirmation()` for actor synchronization
**Key observation:** ScenarioAPlannerSwiftTests is missing `@Suite` attribute (other test files have it).
Task 1: Add Swift Testing isolation to flaky tests
SportsTimeTests/ScenarioAPlannerSwiftTests.swift, SportsTimeTests/ScenarioBPlannerTests.swift, SportsTimeTests/ScenarioCPlannerTests.swift
Apply Swift Testing isolation patterns to prevent state pollution:
1. **Add @Suite attribute to ScenarioAPlannerSwiftTests** (currently missing):
- Add `@Suite("ScenarioA Tests")` before `struct ScenarioAPlannerSwiftTests`
- Matches pattern used in ScenarioBPlannerTests and ScenarioCPlannerTests
2. **Apply .serialized trait to flaky tests:**
- Modify the 5 flaky test functions to use `.serialized` trait
- Swift Testing syntax: `@Test(.serialized, "test description")`
- This prevents parallel execution of these specific tests
- Example: Change `@Test("handles many games")` to `@Test(.serialized, "handles many games")`
3. **Verify actor synchronization patterns:**
- Review if tests use `confirmation()` for actor operations (Swift Testing pattern for async/actor testing)
- Add `confirmation()` if tests modify shared actor state and don't already use it
- Only needed if actors are shared across tests (check planner initialization in `plan()` helper)
**Why .serialized:** These tests likely interfere due to parallel execution timing. Running them serially eliminates race conditions while maintaining all other tests' parallel execution benefits.
**What to avoid and WHY:**
- Don't add `.serialized` to ALL tests - only the 5 flaky ones. Other tests benefit from parallel execution speed.
- Don't use XCTest patterns (like `setUp()`/`tearDown()`) - this is Swift Testing, use Swift Testing patterns only.
- Don't modify test assertions - tests validate correct behavior individually, issue is isolation not correctness.
Run full test suite in parallel:
```bash
xcodebuild -project SportsTime.xcodeproj -scheme SportsTime -destination 'platform=iOS Simulator,name=iPhone 17,OS=26.2' test
```
All 5 previously flaky tests must pass:
- `plan_StopDepartureDate_IsLastGameDate()`
- `plan_ManyGames_HandledEfficiently()`
- `plan_ThreeSameDayGames_PicksFeasibleCombinations()`
- `plan_FillerSameDayAsAnchor_Excluded()`
- `plan_MustSeeGamesTooFarApart_Fails()`
- `corridor_MultipleGamesMixed_FiltersCorrectly()`
Run suite 3 times to confirm consistency (flaky tests would fail at least once in 3 runs).
- @Suite attribute added to ScenarioAPlannerSwiftTests
- .serialized trait applied to all 5 flaky tests
- Full test suite passes 3 consecutive times
- Test output shows serialized tests running (no parallel interference)
Task 2: Verify test isolation doesn't reduce coverage
SportsTimeTests/ScenarioAPlannerSwiftTests.swift, SportsTimeTests/ScenarioBPlannerTests.swift, SportsTimeTests/ScenarioCPlannerTests.swift
Confirm serialization fixes flakiness without compromising test quality:
1. **Run tests individually to confirm behavior unchanged:**
```bash
# Run each flaky test individually
xcodebuild -only-testing:SportsTimeTests/ScenarioAPlannerSwiftTests/plan_StopDepartureDate_IsLastGameDate test
xcodebuild -only-testing:SportsTimeTests/ScenarioAPlannerSwiftTests/plan_ManyGames_HandledEfficiently test
# ... repeat for all 5
```
All must still pass individually (sanity check).
2. **Measure performance impact:**
- Note full suite execution time before changes (from Task 1 baseline)
- Note full suite execution time after .serialized changes
- Document in SUMMARY.md (expected: minimal impact since only 5/180+ tests serialized)
3. **Verify no new flaky tests introduced:**
- Check test output for any other tests now showing intermittent failures
- Run suite 5 times total (2 more runs after Task 1's 3 runs)
- All 180+ tests must pass consistently
**What to avoid and WHY:**
- Don't skip individual test verification - ensures serialization didn't break test logic
- Don't accept new flaky tests - if any appear, investigate whether .serialized needs to be applied more broadly or if there's a different issue
```bash
# Verify all tests pass consistently
for i in {1..5}; do
echo "Run $i/5"
xcodebuild -project SportsTime.xcodeproj -scheme SportsTime -destination 'platform=iOS Simulator,name=iPhone 17,OS=26.2' test | grep "Test Suite 'All tests'"
done
```
Expected output (all 5 runs):
- "Test Suite 'All tests' passed"
- No failures in any of the 5 runs
- All 5 previously flaky tests pass individually
- Full test suite passes 5 consecutive times (100% success rate)
- No new flaky tests detected
- Performance impact documented (serializing 5/180+ tests should add ~1-2 seconds max)
Before declaring phase complete:
- [ ] Full test suite passes consistently (5/5 runs successful)
- [ ] All 5 previously flaky tests now pass in parallel execution
- [ ] No new flaky tests introduced
- [ ] Individual test execution still passes for all 5 tests
- [ ] Test isolation change documented in SUMMARY.md
- All tasks completed
- All verification checks pass
- 5 flaky tests now pass reliably in parallel test suite
- No degradation in test coverage or execution
- CI/CD testing is now reliable