fusion/docs/test-speed-audit-FN-5048.md

FN-5048 Test-Speed Audit

Refreshed baseline (2026-06-03): see docs/test-speed-baseline-2026-06-03.md for the U1 machine-readable per-file timing snapshot, refreshed top-10 offenders, and the cold-start/transform-cost probe feeding the U8 vitest-4 gate.

Scope and method

• Related baseline context:
  • docs/test-audit-report.md (prior workspace test audit baseline)
  • docs/testing-suite-quality-prd.md (suite quality taxonomy + priorities)
• Commands run (baseline):
  • pnpm --filter @fusion/core exec vitest run --reporter=verbose --silent=passed-only
  • pnpm --filter @fusion/engine exec vitest run --reporter=verbose --silent=passed-only
  • pnpm --filter @runfusion/fusion exec vitest run --reporter=verbose --silent=passed-only
  • pnpm --filter @fusion/dashboard exec vitest run --reporter=verbose --silent=passed-only
• Durations were derived from verbose per-test timing logs by aggregating milliseconds per test file and top describe groups.
• Classification rubric:
  • delete: low-signal/duplicate patterns from AGENTS.md “What NOT to write”.
  • rewrite: keep coverage, reduce cost via fake timers/seam narrowing/lighter fixtures/shared harness.
  • keep: high-signal regression/integration backstop.

Baseline run totals

• @fusion/core: 26.29s (pass)
• @fusion/engine: 93.29s (pass)
• @runfusion/fusion: 13.90s (pass)
• @fusion/dashboard: 359.52s (vitest run across all dashboard projects; multiple unrelated pre-existing failures surfaced)

Top offenders by package (baseline)

@fusion/core (top 10)

Rank	File	Aggregated runtime	Dominant suite(s)	Class	Proposed action
1	src/__tests__/agent-store.test.ts	8.47s	AgentStore (100%)	keep	Keep (broad state/store coverage).
2	src/__tests__/mission-store.test.ts	6.16s	MissionStore (100%)	keep	Keep; prioritize targeted fixture slimming only if needed later.
3	src/__tests__/db.test.ts	5.64s	Database (~68%)	keep	Keep (real SQLite backstop).
4	src/__tests__/test-project.test.ts	5.13s	test-project fixture (100%)	rewrite	Evaluate fixture-heavy setup duplication, trim low-value cases.
5	src/__tests__/plugin-loader.test.ts	3.48s	PluginLoader (100%)	keep	Keep.
6	src/__tests__/store-watcher.test.ts	1.80s	TaskStore (100%)	rewrite	Replaced slow watch polling waits with FN-2707 fake-timer recipe.
7	src/__tests__/run-audit.test.ts	2.60s	Run Audit (100%)	keep	Keep.
8	src/__tests__/store-comments.test.ts	2.58s	TaskStore (100%)	keep	Keep.
9	src/__tests__/plugin-store.test.ts	2.48s	PluginStore (100%)	keep	Keep.
10	src/__tests__/store-create.test.ts	2.41s	TaskStore (100%)	keep	Keep.

@fusion/engine (top 10)

Rank	File	Aggregated runtime	Dominant suite(s)	Class	Proposed action
1	src/__tests__/merger-overlap-guard.test.ts	31.00s	overlap-aware fallback integration (~74%)	keep	Keep (real-git heavy merge safety).
2	src/__tests__/merger-staging-allowlist.test.ts	20.41s	staging allowlist (~60%)	keep	Keep (explicitly protected real-git backstop).
3	src/__tests__/merger-diff-volume-gate.test.ts	13.96s	integration + gate logic	keep	Keep.
4	src/__tests__/self-healing-already-merged.real-git.test.ts	10.21s	real-git recovery	keep	Keep.
5	src/__tests__/merger-autostash-cleanup.test.ts	9.79s	sweep* paths	keep	Keep.
6	src/__tests__/branch-conflicts-recovery.test.ts	9.39s	branch conflict classification	keep	Keep.
7	src/__tests__/merger-autostash-orphan-surface.test.ts	8.11s	autostash orphan surface	keep	Keep.
8	src/__tests__/merger-squash-audit.test.ts	7.66s	squash audit	keep	Keep.
9	src/__tests__/self-healing-stale-merge-stats.real-git.test.ts	7.01s	merge metadata recovery	keep	Keep.
10	src/runtimes/__tests__/in-process-runtime.test.ts	6.29s	InProcessRuntime (100%)	rewrite	Trim low-value matrix/permutation assertions.

@runfusion/fusion (top 10)

Rank	File	Aggregated runtime	Dominant suite(s)	Class	Proposed action
1	src/__tests__/extension.test.ts	5.08s	runnable structured-output slice (100%)	rewrite	Trim duplicate wiring cases; preserve env-gated legacy skip.
2	src/__tests__/bin.test.ts	4.16s	command routing	rewrite	Consolidate repetitive route permutations with it.each.
3	src/commands/__tests__/init.test.ts	3.53s	init command (100%)	keep	Keep.
4	src/__tests__/vitest-workspace-resolution.test.ts	1.76s	workspace resolution	keep	Keep.
5	src/__tests__/extension-task-tools.test.ts	1.65s	worktree root resolution	keep	Keep.
6	src/commands/dashboard-tui/__tests__/app.test.tsx	1.57s	settings view subset	rewrite	Remove low-value display duplicates.
7	src/commands/__tests__/chat.test.ts	0.84s	chat interactive	keep	Keep.
8	src/commands/__tests__/dashboard.test.ts	0.80s	dashboard command	keep	Keep.
9	src/__tests__/research-extension-tools.test.ts	0.71s	research tools	keep	Keep.
10	src/commands/__tests__/serve.test.ts	0.51s	serve command	keep	Keep.

@fusion/dashboard (top 10)

Rank	File	Aggregated runtime	Dominant suite(s)	Class	Proposed action
1	src/__tests__/routes-auth.test.ts	93.86s	GET /auth/status (~86%)	rewrite	Replace broad matrix with scoped auth-provider assertions and fixture helper.
2	app/components/__tests__/SettingsModal.test.tsx	89.05s	SettingsModal (100%)	rewrite	Consolidate high-latency waitFor-heavy permutations.
3	app/components/__tests__/ChatView.test.tsx	24.36s	ChatView (~52%)	rewrite	Replace polling waits with event/fake-timer driven completion.
4	src/__tests__/routes-agents.test.ts	22.02s → 11.61s (FN-5870)	mixed routes	rewrite	Landed under FN-5870: collapsed repetitive workflow-step route permutations into it.each, removed a pure field-presence-only template check, and shared app fixtures across mock-isolated describe blocks.
5	app/components/__tests__/App.test.tsx	20.61s	app shell	rewrite	Reduce mock-the-world wiring cases that duplicate higher-level routes.
6	app/components/__tests__/ListView.test.tsx	14.68s	bulk selection + list suites	rewrite	Consolidate combinatorial selection cases.
7	app/components/__tests__/MissionManager.test.tsx	14.22s	MissionManager (100%)	keep/rewrite	Keep FN-tagged behavior; trim duplicate DOM-query shape checks.
8	app/components/__tests__/PrCreateModal.test.tsx	11.61s	PrCreateModal	rewrite	Consolidate repetitive input permutations.
9	app/components/__tests__/AgentPromptsManager.test.tsx	9.60s	AgentPromptsManager	rewrite	Convert repetitive cases into compact tables.
10	app/components/__tests__/AgentDetailView.settings.test.tsx	9.43s	AgentDetailView settings	rewrite	Trim redundant control-shape assertions.

Prioritized execution plan

1. Dashboard SettingsModal + ChatView targeted rewrites (largest projected wall-clock gain).
2. Dashboard routes-auth focused narrowing (GET /auth/status hotspot).
3. CLI bin/extension test consolidation (it.each + duplicate route-case reductions).
4. Core watcher fake-timer rewrite (store-watcher.test.ts).

Keep-safe exclusions explicitly honored

• No edits to reliability-interaction tests under packages/engine/src/__tests__/reliability-interactions/**.
• No deletion of FN-tagged regression tests.
• No concurrency/worker-cap changes.

Post-change results (Step 5 re-measure)

• Re-ran the same per-package verbose commands after landed trims/rewrites.
• Post-change totals:
  • @fusion/core: 23.08s (pass)
  • @fusion/engine: 76.66s (pass)
  • @runfusion/fusion: 9.40s (pass)
  • @fusion/dashboard: 335.40s (vitest run failed with broad pre-existing suite issues unrelated to FN-5048 edits)

Targeted outcome checks for FN-5048 edits

• packages/core/src/__tests__/store-watcher.test.ts (rewrite): retained per-test harness and fake-timerized watch-active polling tests (vi.useFakeTimers before watch, advanceTimersByTimeAsync driving cycles); sampled isolated file runtime now ~1.8s total with watch-active cases no longer incurring 1–3s interval waits.
• packages/cli/src/__tests__/bin.test.ts (rewrite/trim): redundant route permutations consolidated with it.each; CLI suite passes.
• AGENTS.md standing rule added and cross-linked from this audit to prevent reintroduction of slow-test patterns.

Acceptance note

• Edited files classified as rewrite/delete were retained with measurable local pass behavior in their packages.
• @fusion/core no-regression gate satisfied after harness correction (baseline 26.29s → post-change 23.08s).
• Dashboard lane remains unstable from pre-existing repository-level failures in planning/tracking/interview route tests and unrelated UI suites; no FN-5048 file changes were made in those failing areas.

FN-5074 follow-up results

• Isolated re-measure after targeted rewrites (same command family used in FN-5074 preflight):
  • routes-auth.test.ts: 94.53s → 15.76s (164 tests)
  • routes-agents.test.ts: 9.72s → 10.58s in FN-5074, then 12.66s → 11.61s in FN-5870 (348 tests in current isolated rerun; helper-driven route parity preserved)
  • SettingsModal.test.tsx: 43.40s → 48.60s (438 tests; sampled rerun regressed due to remaining long-running device-code path)
  • ChatView.test.tsx: 12.53s → 14.66s (390 tests; sampled rerun modestly higher while preserving coverage)
• FN-5074 preserved FN-tagged coverage and frozen-button assertions; all four files pass in isolation and full verification gates remained green in task execution.

FN-6307 follow-up results

• Targeted isolated re-measure used the current dashboard quality projects with the dot reporter and a 2x back-to-back flakiness check for:
  • src/__tests__/routes-agents.test.ts
  • src/__tests__/routes-git.test.ts
  • src/__tests__/routes-planning.test.ts
  • app/components/__tests__/FileEditor.test.tsx
  • app/components/__tests__/NewTaskModal.test.tsx
• Results (tests unchanged):

File	Tests	Baseline duration/test time	After run 1 duration/test time	After run 2 duration/test time	Outcome
routes-agents.test.ts	174	58.43s / 46.64s	33.42s / 28.84s	27.06s / 24.31s	retained coverage; no source edits needed
routes-git.test.ts	98	10.71s / 7.89s	17.81s / 15.37s	10.65s / 8.25s	retained coverage; no source edits needed
routes-planning.test.ts	102	7.74s / 4.97s	4.76s / 2.30s	4.17s / 1.64s	replaced duplicate rate-limit HTTP loop with direct limiter seeding while preserving boundary HTTP assertions
FileEditor.test.tsx	45	5.31s / 3.76s	6.99s / 3.57s	4.83s / 3.71s	retained CSS/layout suites unchanged
NewTaskModal.test.tsx	50	3.83s / 2.16s	3.58s / 1.95s	3.54s / 1.97s	replaced disabled-button negative polling waits with direct assertions

• Combined after-run 2 duration for the five targeted files was 50.25s versus 86.02s in the Step 0 baseline. Test counts stayed constant; no FN-tagged regression coverage was removed.

FN-6308 dashboard orchestration follow-up results

• pnpm --filter @fusion/dashboard test now uses packages/dashboard/scripts/run-quality-tests.mjs to run the same 15 quality lanes with bounded process concurrency (FUSION_DASHBOARD_TEST_CONCURRENCY, default/hard cap 2) while preserving the 6144 MiB per-lane heap wrapper and app/API split that avoids historical jsdom SIGKILL/OOM.
• Same-machine warm baseline before orchestration: 446.8s for the sequential dashboard quality chain.
• Post-orchestration measurements: 202.2s after the route-settings mock update and 193.3s after the lint import fix, a roughly 55–57% wall-clock reduction versus the FN-6308 baseline while preserving the file-set parity guard.
• CI-shape validation with FUSION_TEST_TOTAL_WORKERS=6 FUSION_TEST_CONCURRENCY=2 pnpm --filter @fusion/dashboard test passed in 192.5s. Standard and CI-shape logs were scanned for SIGKILL/OOM/fatal heap symptoms with none found.

Notes

• Dashboard vitest run baseline and post-change measurements both surfaced broad failing suites outside this task’s implementation scope; timing evidence is still captured from the same command family.
• FN-6307 verification also observed pre-existing isolated failures in src/__tests__/routes-settings.test.ts (GET /settings/scopes returning 500 for scoped settings cases) while targeted files remained green. FN-6308 updated that stale mock/response expectation after /api/settings/scopes added workflowSettings.
• Standing prevention rule: see AGENTS.md → Standing Rule: Do Not Add Slow Tests (FN-5048).