Watch

Viktor Vaczi 63ed1f3c1f e2e/CI: dual-engine suites, per-engine screenshots, SwiftShader retired, prod web suite, CI-coverage gate

Squash of experiment/ff-big-modules vs main.

Big-module routing removed: native-EH shrank kicad_editor below
SpiderMonkey's x86-64 code budget (runs 29355049705/29356152413 green on
stock Firefox), so BIG_MODULE_SPECS routing and the baseline-only-JIT
crutch are gone — kicad-firefox and kicad-chromium both run the full
suite, with the module compiled the way real users' browsers compile it.

Per-engine screenshots end to end: stableShot/shotPath write
test-results/<engine>/<name>.png; baselines move to
baseline-screenshots/{chromium,firefox}/ and the whole tools/screenshots
pipeline (compare/promote/manifest/spec-map/changelog/Discord) keys on
<engine>/<name>. Previously Firefox and Chromium renders of one spec
overwrote each other and Firefox renders were never actually gated.
Seeded from CI run 29421380806 (92 new firefox baselines, +24 chromium
web-suite shots); manifest generated from the baseline tree.

One merged playwright.config.ts (kicad/asyncify/coroutine/perf as
projects); ~25 dead npm scripts dropped. The web suite is gated in CI for
the first time ever (4 rotted specs fixed, 5 broken lib-bridge specs
triaged as fixme in docs/features/web-e2e-rot/); cheap lint step after
npm ci; last 26 blind-sleep violations fixed.

SwiftShader retired: CI Chromium renders WebGL on ANGLE → Mesa llvmpipe
(--use-gl=angle --use-angle=gl --ignore-gpu-blocklist; the blocklist flag
is mandatory — llvmpipe is blocklisted and WebGL is silently unavailable
without it) in BOTH configs. Under WORKERS=4 congestion SwiftShader
transiently failed the first post-board-load draw and the recovery
cascade ended in a silent permanent Cairo fallback — that engine flip was
the "~1.2% changedRatio both directions" occ-export baseline flake.
Validated 160/160 across two 80-repeat rigs; full analysis in
docs/features/wx-parity-bugs/occ-export-context-eviction.md. Chromium
baselines shift slightly on llvmpipe — promote once from the first green
run. Deflakes the new coverage exposed: presence baselines settle before
capture; presence fixtures declare current file formats; perf gets its
own outputDir so CI evidence survives; occ-export settles the board paint
before the export dialog; menu-item waits (waitForRenderedByLabel before
clickMenuItem) in 4 specs + the TESTING.md rule.

Web suite runs the PROD build, in parallel: webServer becomes backend
`start` + the standalone's e2e:preview (build-preview.mjs: link-wasm →
stash the public/wasm symlink aside during vite build, build-demo.mjs's
move — then vite preview as the persistent server). The wasm middleware
serves /wasm/* in preview and emits COOP/COEP/CORP itself (a pthread
worker script's own response must carry COEP or Chrome kills it with
ERR_BLOCKED_BY_RESPONSE). VITE_* flags bake at build time;
VITE_ALLOW_USER_OVERRIDE joins turbo globalEnv. fullyParallel + default
workers: 5.2m → 1.4m. Determinism fixes the parallel run exposed:
shared-page specs become serial groups; locks.spec grabs alice's exact
item via the new kicadCollabTestSelectByUuid hook (cross-tab "first
footprint" order is not a ysync invariant); quit specs poll page.url()
(quit supersedes its own navigation — NS_BINDING_ABORTED on Firefox).
Suite: 51 passed / 12 skipped / 0 failed in 1.6m.

CI-coverage gate (lint:ci-coverage): every tests/**/*.spec.ts must be
reachable from the npm scripts the workflows invoke — scraped from
.github/workflows/, resolved through package.json, coverage asked from
playwright --list itself. Rules: uncovered-spec + orphan-project (with a
documented LOCAL_ONLY_PROJECTS allowlist). Gating next to
lint:determinism; 138 spec files / 13 projects accounted for.

Product fixes kept from the investigations (reachable on real GPUs too):
wx 7799fd1be5 — paint flags clear before dispatch + Invalidate always
propagates; kicad 3dcfea5e45 — SwiftShader pass-boundary flush +
per-instance font texture + first-frame GL-error drain (GAL recovery
recovers instead of falling back to Cairo) + the user-facing eeschema
switch navigates again under __EMSCRIPTEN__ (project-sync's
FaceRegistered gate had rerouted it into the hidden sync player; caught
by the newly-gated web suite).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018eUxiPApHgGiu9NFyQfhAq

2026-07-17 12:21:54 +02:00

4.7 KiB

Raw Blame History

Testing rules

Determinism rules for the Playwright specs (tests/e2e, tests/kicad, tests/web). Enforced by npm run lint:determinism (tools/lint-determinism.ts, gating in CI). Run specs from tests/ via npm run test:e2e (the full CI project set: wx-chromium, kicad-firefox, kicad-chromium, asyncify-firefox, coroutine-firefox) or npm run test:kicad (kicad-firefox only) — not playwright directly. One spec on one engine: npx playwright test --project=kicad-firefox kicad/pcbnew.spec.ts.

Waits — never blind

No page.waitForTimeout(n). Wait on a condition: expect.poll(() => predicate), a web-first assertion (expect(locator).toBeVisible()), or waitUntil(page, fn, desc) (throws loudly on timeout).
App readiness: waitForWxApp(page) (canvas visible + element registry populated) for widget/editor harnesses; waitForCanvasApp(page) for registry-less canvas apps.
The only allowed waitForTimeout is an irreducible interaction dwell — a canvas/keyboard commit with no JS-observable signal — and it MUST carry a same-line marker: // eslint-disable-line -- documented interaction dwell: <why>.
Menu clicks: wait for the specific item, not a count. Popup items register in the element registry progressively as they paint, so a coarse gate ("N menuitems rendered") can pass before the item you're about to click exists — and clickMenuItem is single-shot. Before every clickMenuItem(page, 'X'), await waitForRenderedByLabel(page, 'X', { elementType: 'menuitem' }) (same matcher as the click). clickMenuItemByText already waits internally and needs no guard. A submenu click needs its own wait: the parent menu's still-rendered items satisfy any count gate before the submenu paints.

No defensive branches

No if (await el.count()) el.click(). Assert the element exists, then act: expect(await clickByLabel(page, 'X'), '...').toBe(true). Use clickMenuItemByText (normalizes & / ... / …) instead of try-A-else-A…-else-A fallback chains.
No swallowed .catch(() => {}). Let it throw, or assert the tolerated outcome. A genuinely best-effort op must carry a marker explaining why.

Screenshots — `stableShot`, compared offline

Capture with stableShot(page, 'name.png', { fullPage }) — it settles the render (in-page canvas-hash over animation frames) then writes a raw PNG to test-results/<engine>/ (chromium/firefox, derived from the running browser — the same spec on two engines writes two files). It does not assert. Never use Playwright's toHaveScreenshot. Raw page.screenshot/fs writers must route through shotPath(page, 'name.png') for the same engine scoping.
Comparison is offline and per-engine: npm run screenshots:check diffs test-results/<engine>/ against the committed baselines in tests/baseline-screenshots/<engine>/ (+ 3d-regression/, gal-regression/). tests/screenshot-manifest.json (regen: npm run screenshots:manifest) is the authoritative {name, engine} list — CI fails the lint step if it drifts from the baseline tree.
CI's Linux render is the source of truth; baselines are promoted from CI (npm run screenshots:promote -- --run <ci-run-id>). A local (Mac) check shows font/render noise and is not the gate.
A continuously-animating state (timer, mid-slide) can't be a stable baseline — drop the shot.

Retries

retries: 0 in both configs (playwright.config.ts — the merged wasm-suite config — and playwright-web.config.ts). A failure is real; don't mask it with a retry.

Every spec runs in CI — `lint:ci-coverage`

npm run lint:ci-coverage (tools/lint-ci-coverage.ts, gating in CI next to the determinism lint) proves every *.spec.ts under tests/ is actually executed by CI: it scrapes the npm run test:… invocations from .github/workflows/, resolves them through package.json to their playwright test --config/--project flags, and asks Playwright itself (--list) which files those runs cover. No hand-maintained lists — adding a spec in a brand-new directory is exactly what it catches.

When it fires:

uncovered-spec — your new spec matches no CI-run project. Put it in a covered testDir, adjust a project's testMatch, or add the project to a CI npm script.
orphan-project — you added a config project no CI script selects. Wire it into a CI script, or (for deliberately-local system-browser projects) add it to LOCAL_ONLY_PROJECTS in the lint with a comment saying why.

Where things are

Per-test logs (JS console + cpp): tests/logs/{wxwidgets,kicad}/<test-name>/.
Guards: npm run lint:determinism, npm run lint:ci-coverage. Screenshot gate: npm run screenshots:check.

4.7 KiB Raw Blame History