Files

gsxdsm 845fad2aa4 FN-7666: record weekly test-velocity regression finding (gate 7.7s -> 36.3s)

Narrative: update the 2026-W28 test-velocity baseline report with the newly measured merge-gate/changed-test wall-time regression and document root-cause attribution.
- Bump baseline cycle to 2026-W28 (captured 2026-07-08T09:30:50.606Z)
- Update headline metrics table: gate 7.7s -> 36.3s (+28.6s), boot smoke 17.4s -> 25.3s (+7.9s), pnpm test 9.3s -> 43.4s (+34.0s), quarantine count unchanged at 1
- Add FN-7666 finding note attributing the regression to ~4,305 new re-exported source lines across 14 modules pulled through the @fusion/core / @fusion/engine barrels, inflating per-fork import/transform cost across the 18 curated engine-core gate files; confirms it is a genuine regression (not cache/env noise) via cold-vs-warm reruns; files follow-up FN-7667 for the barrel-decoupling fix
- Update weekly history table (previous/latest/delta rows) and the #leads post text block
- Append new row capturing this week's measurements to scripts/test-velocity-history.json

Files changed:
 docs/test-velocity-baseline.md     |  23 ++++----
 scripts/test-velocity-history.json | 112 +++++++++++++++++++++++++++++++++++++
 2 files changed, 123 insertions(+), 12 deletions(-)

Fusion-Task-Id: FN-7666

Fusion-Task-Lineage: 73ec627c-2fc6-4289-8e6e-3e839ed5c675

Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>

2026-07-08 02:45:47 -07:00

6.3 KiB

Raw Blame History

Test velocity baseline

Weekly FN-6612 signal-per-second baseline. Measure and report feedback-loop velocity; do not add slow tests or wire this report into blocking PR checks. The merge gate remains the existing thin Lint, Typecheck, Build, and Gate path.

Latest baseline

Cycle: 2026-W28
Captured at: 2026-07-08T09:30:50.606Z
Timing snapshot: scripts/test-timings.json captured at 2026-06-27T05:41:42.568Z
Quarantine ledger: scripts/lib/test-quarantine.json

Metrics

Metric	Current	Delta vs previous
Merge gate wall-time (`pnpm test:gate`)	36.3s	+28.6s
Boot smoke wall-time (`pnpm smoke:boot`)	25.3s	+7.9s
Changed-only test wall-time (`pnpm test`)	43.4s	+34.0s
Quarantine / flake count	1	0
Deletion-due quarantines	0	n/a

Measurement failures

None recorded.

Timing snapshot notes

No stale or missing timing metadata detected in the rendered slowest-file rows.
FN-7666 finding (2026-07-08): the 2026-W28 merge-gate/changed-test wall-time jump (gate 7.7s → 36.3s, pnpm test 9.3s → 43.4s) is a genuine transform/import regression, not environmental/cache noise — confirmed via cold-vs-warm engine-core reruns (cold real 22.0s vs warm real 20.6s, only a ~6% delta, on an 88.7%-idle 28-core box). Bisection against the 2026-W27 baseline commit (2d8f087a9, gate=7.7s) found no vitest.config.ts pool/alias/include changes in engine-core; instead ~30 legitimate feature commits landed between 2026-07-02 and 2026-07-08 (planner-overseer/oversight, GitLab tracking+analytics, git-revert/AI-undo, provider registration/auth-storage split, bundled-plugin auto-install, Coding-Ideas workflow) that newly re-export ~4,305 source lines across 14 modules (largest: packages/engine/src/task-revert.ts, 1,652 lines) through the packages/core/src/index.ts / packages/engine/src/index.ts barrels. Because the engine-core vitest project resolves @fusion/core/@fusion/engine to those full barrels and runs each of its 18 curated gate files in its own pool:"forks" OS process, every fork independently re-imports and re-transforms the enlarged barrel even though none of the 18 gate tests exercise the new modules. None of the new modules pull in unusually heavy new third-party dependencies individually — the cost is the aggregate ~4.3K-line import surface × 18 forked processes. The proper fix (decoupling the 18 gate test files' imports from the full barrel, or splitting a lean gate-only entrypoint) is a broad cross-package test-infra rework outside FN-7666's Review-Level-1 scope; filed as follow-up FN-7667 (depends on FN-7666). No vitest.config.ts/barrel edits were made for this finding — the next weekly capture should re-confirm this attribution and FN-7667's fix should return the gate to the ~7-10s band.

Rank	File	Package	Duration
1	`packages/dashboard/src/__tests__/insights-routes.test.ts`	@fusion/dashboard	26.5s
2	`packages/engine/src/runtimes/__tests__/in-process-runtime.test.ts`	@fusion/engine	24.7s
3	`packages/dashboard/src/__tests__/workflow-routes.test.ts`	@fusion/dashboard	22.0s
4	`packages/core/src/__tests__/db.test.ts`	@fusion/core	21.2s
5	`packages/dashboard/app/components/__tests__/GitManagerModal.test.tsx`	@fusion/dashboard	16.9s
6	`packages/core/src/__tests__/mission-store.test.ts`	@fusion/core	16.0s
7	`packages/cli/src/__tests__/extension.test.ts`	@runfusion/fusion	15.7s
8	`packages/dashboard/app/components/__tests__/AgentPromptsManager.test.tsx`	@fusion/dashboard	14.8s
9	`packages/dashboard/app/components/__tests__/App.test.tsx`	@fusion/dashboard	14.6s
10	`packages/dashboard/app/components/__tests__/TaskDetailModal.inline-editing-and-integrations.test.tsx`	@fusion/dashboard	14.1s
11	`packages/dashboard/app/components/__tests__/TaskDetailModal.rendering.test.tsx`	@fusion/dashboard	13.7s
12	`packages/dashboard/src/__tests__/routes-auth.test.ts`	@fusion/dashboard	13.6s
13	`packages/core/src/__tests__/agent-store.test.ts`	@fusion/core	13.4s
14	`packages/engine/src/__tests__/workspace-merger-idempotency.test.ts`	@fusion/engine	12.7s
15	`packages/engine/src/__tests__/self-healing-workspace.test.ts`	@fusion/engine	11.8s
16	`packages/engine/src/__tests__/pr-response-run.test.ts`	@fusion/engine	11.6s
17	`packages/dashboard/app/components/__tests__/ListView.test.tsx`	@fusion/dashboard	11.3s
18	`plugins/fusion-plugin-compound-engineering/src/__tests__/sync.test.ts`	@fusion-plugin-examples/compound-engineering	11.0s
19	`packages/dashboard/app/components/__tests__/AgentDetailView.settings.test.tsx`	@fusion/dashboard	10.7s
20	`packages/dashboard/app/components/__tests__/SecretsView.test.tsx`	@fusion/dashboard	10.7s

Quarantine age buckets

Age bucket	Count
0-6 days	1
7-13 days	0
deletion due (>=14 days)	0
unknown/future	0

Deletion-due entries

File	Quarantined at	Age (days)
—	—	—

Before / after trend

Row	Captured at	Gate	Boot smoke	`pnpm test`	Quarantine count
Previous	2026-07-02T08:47:17.721Z	7.7s	17.4s	9.3s	1
Latest	2026-07-08T09:30:50.606Z	36.3s	25.3s	43.4s	1
Delta	—	+28.6s	+7.9s	+34.0s	0

Future weekly rows append to scripts/test-velocity-history.json; compare the latest row against the previous row before posting to #leads.

Post to #leads

FN-6612 weekly test velocity: gate 36.3s (+28.6s), boot smoke 25.3s (+7.9s), pnpm test 43.4s (+34.0s), quarantine ledger 1 (0). Slowest file: packages/dashboard/src/__tests__/insights-routes.test.ts at 26.5s. Deletion-due quarantines: 0.

How to refresh

pnpm test:velocity -- --measure --write-report

In measure mode, the script runs a non-measured pnpm build preflight before timing pnpm test:gate, pnpm smoke:boot, or pnpm test. The preflight time is setup only and is excluded from lane metrics; if it fails, the Measurement failures section records Build preflight (pnpm build) as the reason. Use --skip-build-preflight only when the workspace is already built by CI.

Report-only regeneration is cheap and does not run any suite:

pnpm test:velocity

6.3 KiB Raw Blame History Unescape Escape