pcbjam/tests/playwright.config.ts
Gergő Törcsvári 7dbc9fb36e
design-b D1b/c: context harness + the bounded-and-measured memory gate
Doc 20 §6 D1, second slice: the dedicated test app D1 calls for, plus
the doc 20 §7 risk-1 gate.

tests/apps/standalone/sched-context/ drives the primitives with no wx
linked (a failure can only be the contexts layer) and assertions ON —
unlike the races harness, an emscripten "cannot start an async
operation" state here IS the bug, not a tolerated state. Nine
scenarios, each pinning one invariant; the load-bearing ones:
- parked_does_not_block: a parked context does not stop others running.
  This is doc 19's freeze made unrepresentable — there a parked activity
  held a global interlock and the UI died; here "parked" simply means
  "not runnable", and three workers run past it before it resumes.
- one_transition_in_flight: a context calling drain() gets a no-op, with
  a second ready context queued so a buggy nested drain would actually
  run something and be caught.
- async_wake: a real macrotask hop (setTimeout → mark_ready → drain),
  the shape every production bridge has.
- deep_park_sizing: parks 64 live frames deep to measure what a park
  actually costs.

The memory gate asserts the ceiling (peak live contexts, peak bytes),
that nothing leaked (live=0, bytes=0, created==finished), that nothing
was left mid-transition, and that refusals occurred (zero would mean the
illegal-operation scenarios stopped provoking).

Measured here: ~34 B/frame, 2200 B for a 64-frame park, 1 MB peak for 4
concurrent contexts. Recorded with the caveat that the harness's frames
carry three locals each, so this is a FLOOR — real bridges save far more
per frame (libcontext runs 512 K after a 64 K buffer silently
overflowed). The apparatus and its units are validated; the sizing
DECISION needs deep-park numbers from real bridges at D3/D4.

playwright.config: asyncify-firefox now matches every spec in ./asyncify
instead of only asyncify-races* — a new harness there is covered by
construction rather than by remembering to widen the pattern.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TEHGiiXMShNXbBr7gSJ7iz
2026-08-10 10:14:16 +02:00

291 lines
11 KiB
TypeScript

import { defineConfig, devices } from '@playwright/test';
import { execSync } from 'child_process';
import * as fs from 'fs';
import * as path from 'path';
// THE merged Playwright config for every suite that runs against the static
// `apps` server: the wx widget suite (e2e/), the KiCad editor suite (kicad/),
// the asyncify race harness (asyncify/) and the coroutine harness
// (e2e/coroutine*). One config = one invocation = one webServer, one port file
// and ONE start-of-run outputDir wipe — which is what retired the old
// per-suite pw-artifacts/{wx,asyncify,kicad} redirect dance (each sequential
// CI invocation used to wipe the previous suite's artifacts).
//
// The web-app suite (web/) stays in playwright-web.config.ts: it runs against
// the React editor + backend stack (`pnpm --dir ../web dev`), not this server.
//
// CI runs: npm run test:e2e (wx-chromium, kicad-firefox, kicad-chromium,
// asyncify-firefox, coroutine-firefox)
// npm run test:perf (perf project, non-gating, separate invocation)
// Local-only projects (system Chrome / WebKit) are listed at the bottom.
const PORT_FILE = path.join(__dirname, '.test-port');
// Resolve the static-server port for this run.
//
// This config file is re-imported by EVERY Playwright process: the main runner
// (which launches the webServer) and each worker process (which calls
// page.goto(baseURL)). They must all agree on one port. Playwright also
// *recreates* a worker mid-run after a test times out or crashes — and that new
// worker re-imports this config.
//
// The main runner is the only process whose argv carries the `test` command
// (workers are forked with an empty argv); it always picks a fresh port and
// writes the file before any worker spawns, and workers always reuse it — no
// freshness window, so a recreated worker can never rotate to a dead port
// (the old ERR_CONNECTION_REFUSED cascade).
function resolvePort(): number {
const isMainRunner = process.argv.slice(2).includes('test');
if (!isMainRunner) {
try {
const existing = parseInt(fs.readFileSync(PORT_FILE, 'utf-8').trim(), 10);
if (existing > 0 && existing < 65536) {
return existing;
}
} catch {
// No readable port file — fall through. Shouldn't happen in a worker,
// since the main runner writes the file before spawning workers.
}
}
const port = findFreePort();
fs.writeFileSync(PORT_FILE, port.toString());
return port;
}
function findFreePort(): number {
try {
const result = execSync(
'python3 -c "import socket; s=socket.socket(); s.bind((\'\',0)); print(s.getsockname()[1]); s.close()"',
// timeout: a broken python3 must fall back to the random port below, not
// hang the whole run at config-load time.
{ encoding: 'utf-8', timeout: 5000 }
);
return parseInt(result.trim());
} catch {
return 9000 + Math.floor(Math.random() * 1000);
}
}
const port = resolvePort();
const appsDir = 'apps';
// Runtime-perf specs run ONLY on the Chromium 'perf' project: they need CDP
// CPU throttling (Chromium-only). Excluded from the kicad projects so they
// don't double-run there.
const PERF_SPECS = ['**/*-perf.spec.ts'];
// Bundled Chromium on GPU-less CI: use ANGLE over desktop GL (Mesa llvmpipe
// via the Xvfb display CI already provides — the whole e2e step runs under
// `xvfb-run -a`) instead of SwiftShader. SwiftShader under multi-worker
// contention transiently fails WebGL calls ("Requested render buffer size is
// not supported", LRU context evictions at 3 live contexts, cross-context
// object errors), which stochastically drove KiCad's GAL error-recovery and
// flipped the occ-export screenshots between runs. llvmpipe — the same
// software-GL stack the Firefox projects run on — never exhibited any of it.
// Flags tested useless against the SwiftShader eviction before the switch:
// --max-active-webgl-contexts=64, --force-gpu-mem-available-mb,
// --disable-low-end-device-mode, --disable-gpu-driver-bug-workarounds.
// See docs/features/wx-parity-bugs/occ-export-context-eviction.md.
const CHROMIUM_CI_ARGS = process.env.CI
? {
launchOptions: { args: [
'--use-gl=angle',
'--use-angle=gl',
// llvmpipe is on Chromium's software-GL blocklist (the analog of the
// Firefox projects' webgl.force-enabled pref) — without this, WebGL
// is simply unavailable and the app falls back to Cairo rendering.
'--ignore-gpu-blocklist',
] },
}
: {};
// CI-only Firefox prefs: GPU-less CI VMs can't create a headless GL context
// (FEATURE_FAILURE_WEBGL_EXHAUSTED_DRIVERS) — run headed under Xvfb, where
// GLX + Mesa llvmpipe provides software WebGL, and bypass the no-GPU
// blocklist so the GAL canvas gets a context. CI invokes the suite via
// `xvfb-run`.
const FIREFOX_CI_OPTS = process.env.CI
? {
headless: false,
launchOptions: {
firefoxUserPrefs: {
'webgl.force-enabled': true,
},
},
}
: {};
export default defineConfig({
globalSetup: './global-setup.ts',
// Every suite writes its gate screenshots to test-results/<engine>/ via
// stableShot/shotPath — NOT into outputDir. On CI, outputDir points at a
// throwaway dir so Playwright's start-of-run cleanup never touches the
// accumulated screenshots (the perf run is a separate later invocation).
outputDir: process.env.CI ? 'pw-artifacts/e2e' : 'test-results',
fullyParallel: true,
forbidOnly: !!process.env.CI,
// retries:0 — the suites are deterministic (no blind sleeps or "if element
// exists" branches; screenshots go through stableShot for the offline gate,
// not asserted inline), so a failure is a real failure rather than flake to
// mask with a retry.
retries: 0,
// Parallel workers on CI too (Playwright default ≈ 50% of cores) — the
// serial CI run was the dominant wall-clock cost.
workers: undefined,
reporter: 'html',
timeout: 60000, // wx-suite default; the heavier projects override below
// No expect.toHaveScreenshot: screenshots are captured via stableShot()
// (render-settle + raw PNG to test-results/<engine>/) and compared OFFLINE
// by tools/screenshots against tests/baseline-screenshots/<engine>/ on CI's
// deterministic Linux render.
use: {
baseURL: `http://localhost:${port}`,
trace: 'retain-on-failure',
screenshot: 'only-on-failure',
},
projects: [
// ── CI projects (npm run test:e2e) ──────────────────────────────────
{
// The wx widget/test-app suite. Chromium-only by design (the wx PORT is
// exercised cross-engine by the kicad + coroutine projects).
name: 'wx-chromium',
testDir: './e2e',
use: {
...devices['Desktop Chrome'],
permissions: ['clipboard-read', 'clipboard-write', 'local-fonts'],
...CHROMIUM_CI_ARGS,
},
},
{
name: 'kicad-firefox',
testDir: './kicad',
testIgnore: PERF_SPECS,
timeout: 180000, // KiCad WASM needs more time to load
use: {
...devices['Desktop Firefox'],
viewport: { width: 1280, height: 720 },
...FIREFOX_CI_OPTS,
},
},
{
name: 'kicad-chromium',
testDir: './kicad',
testIgnore: PERF_SPECS,
timeout: 180000,
use: {
...devices['Desktop Chrome'],
viewport: { width: 1280, height: 720 },
...CHROMIUM_CI_ARGS,
},
},
{
// Asyncify-layer harnesses: the race-condition red-green battery and
// the Design-B scheduler-context gate. One heavy WASM app at a time.
// Matches EVERY spec in ./asyncify — a new harness here is covered by
// construction rather than by remembering to widen this pattern.
name: 'asyncify-firefox',
testDir: './asyncify',
testMatch: /\.spec\.ts$/,
fullyParallel: false,
timeout: 120000,
use: {
...devices['Desktop Firefox'],
viewport: { width: 1280, height: 720 },
...FIREFOX_CI_OPTS,
},
},
{
// Coroutine harness on real Firefox — the wx-chromium project already
// covers these specs on bundled Chromium; this is the cross-engine leg.
name: 'coroutine-firefox',
testDir: './e2e',
testMatch: /coroutine.*\.spec\.ts$/,
fullyParallel: false,
timeout: 120000,
use: {
...devices['Desktop Firefox'],
viewport: { width: 1280, height: 720 },
...FIREFOX_CI_OPTS,
},
},
{
// Runtime-perf specs: bundled Chromium for CDP CPU throttling. Run as a
// SEPARATE, non-gating invocation (npm run test:perf) — hence its own CI
// outputDir so its start-of-run wipe can't delete the e2e run's traces.
name: 'perf',
testDir: './kicad',
testMatch: PERF_SPECS,
timeout: 180000,
...(process.env.CI ? { outputDir: 'pw-artifacts/perf' } : {}),
use: {
...devices['Desktop Chrome'],
viewport: { width: 1280, height: 720 },
...CHROMIUM_CI_ARGS,
},
},
// ── local-only projects (system Chrome / WebKit; CI installs neither) ─
{
// System Chrome (real GPU/V8) for headed KiCad debugging:
// npx playwright test --project=kicad-chrome --headed kicad/pcbnew.spec.ts
name: 'kicad-chrome',
testDir: './kicad',
testIgnore: PERF_SPECS,
timeout: 180000,
use: {
channel: 'chrome',
viewport: { width: 1280, height: 720 },
},
},
{
name: 'asyncify-chrome',
testDir: './asyncify',
testMatch: /asyncify-races.*\.spec\.ts$/,
fullyParallel: false,
timeout: 120000,
use: {
channel: 'chrome',
viewport: { width: 1280, height: 720 },
permissions: ['clipboard-read', 'clipboard-write'],
},
},
{
// WebKit (Safari's engine) — project policy: every asyncify spec must be
// green in all three engines. Run via npm run test:asyncify:safari.
name: 'asyncify-webkit',
testDir: './asyncify',
testMatch: /asyncify-races.*\.spec\.ts$/,
fullyParallel: false,
timeout: 120000,
use: {
...devices['Desktop Safari'],
viewport: { width: 1280, height: 720 },
},
},
{
// Coroutine harness on system Chrome (real V8/GPU — where the KiCad
// coroutine crash historically manifested). Must be --headed on ARM Mac.
name: 'coroutine-chrome',
testDir: './e2e',
testMatch: /coroutine.*\.spec\.ts$/,
fullyParallel: false,
timeout: 120000,
use: {
channel: 'chrome',
viewport: { width: 1280, height: 720 },
permissions: ['clipboard-read', 'clipboard-write', 'local-fonts'],
},
},
],
webServer: {
command: `npx serve ${appsDir} -p ${port} -c ../serve.json`,
port: port,
reuseExistingServer: !process.env.CI,
},
});