A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
183 lines
9.4 KiB
TypeScript
183 lines
9.4 KiB
TypeScript
import { test, expect } from './fixtures';
|
|
import { clickByTooltip, waitForEditorReady } from '../e2e/utils/element-tracker';
|
|
import { FATAL_WASM_PATTERNS, findNativeFailure } from './utils/native-failure';
|
|
import {
|
|
loadRectifier,
|
|
openSimulator,
|
|
runSimulation,
|
|
waitForRunToolEnabled,
|
|
} from './utils/sim-harness';
|
|
|
|
/**
|
|
* eeschema simulator worker-death recovery (findings E-10/E-12/E-11): the
|
|
* promise of the out-of-process engine is that a worker death settles
|
|
* everything in flight and the next Run transparently boots a fresh worker.
|
|
* These specs kill (or corrupt) the service at exact points and assert the
|
|
* simulator UI actually recovers:
|
|
*
|
|
* - E-10: a mid-run worker death must unlatch the client's s_bgRunning
|
|
* mirror (via the service's synthetic controlled-exit) — otherwise the
|
|
* Run tool's ENABLE(!simRunning) holds "running" forever and the promised
|
|
* fresh-worker restart is unreachable for the whole session.
|
|
* - E-12: a run whose transport dies between launch acceptance and its
|
|
* RUNNING transition delivers its crash-exit completion — the wasm-only
|
|
* unowned-event drop must not swallow an owned run's only IDLE.
|
|
* - E-11: a corrupted worker's oversized get_vec length must be clamped to
|
|
* the actually-transferred arrays — not copied into the editor heap as a
|
|
* multi-gigabyte read that traps the instance.
|
|
*/
|
|
|
|
test.describe('eeschema simulator worker-death recovery', () => {
|
|
test.setTimeout(300000);
|
|
|
|
test('a mid-run worker death re-enables Run and a rerun succeeds (E-10)', async ({ page, testLogger }) => {
|
|
await page.goto('/kicad/eeschema.html');
|
|
await waitForEditorReady(page);
|
|
await loadRectifier(page);
|
|
await openSimulator(page);
|
|
await waitForRunToolEnabled(page);
|
|
const checkpoint = await page.evaluate(() => {
|
|
const hooks = (globalThis as any).__ngspiceServiceTestHooks;
|
|
return hooks.appliedGenerationCheckpoint() as number;
|
|
});
|
|
|
|
// Start a run and inject the worker death while it is live. The check
|
|
// and the retirement happen in ONE page.evaluate — frames dispatch on
|
|
// the same main thread, so no finish frame can interleave between the
|
|
// "still running" check and the kill. Event scans are scoped past any
|
|
// frame-open activity (workbook plot restoration).
|
|
const eventFloor = await page.evaluate(
|
|
() => ((window as any).__ngspiceEvents as unknown[]).length);
|
|
expect(await clickByTooltip(page, 'Run Simulation', { elementType: 'tool' }),
|
|
'Run tool').toBe(true);
|
|
|
|
// Run accepted: the worker's bg started frame arrived.
|
|
await expect.poll(
|
|
() => page.evaluate((floor: number) =>
|
|
((window as any).__ngspiceEvents as Array<{ kind: string; finished?: boolean }>)
|
|
.slice(floor)
|
|
.some((e) => e.kind === 'bg' && e.finished === false), eventFloor),
|
|
{ message: 'the run must report bg started', timeout: 60000 },
|
|
).toBe(true);
|
|
|
|
const injected = await page.evaluate((floor: number) => {
|
|
const events = ((window as any).__ngspiceEvents as Array<{
|
|
kind: string; finished?: boolean }>).slice(floor);
|
|
const finishSeen = events.some((e) => e.kind === 'bg' && e.finished === true);
|
|
const retired = (globalThis as any).__ngspiceServiceTestHooks
|
|
.forceRetire('E-10 repro: worker death mid-run');
|
|
return { finishSeen, retired };
|
|
}, eventFloor);
|
|
expect(injected.finishSeen,
|
|
'repro window: the run must still be live when the fault is injected').toBe(false);
|
|
expect(injected.retired, 'the active generation was retired').toBe(true);
|
|
|
|
// THE E-10 oracle: without the synthetic controlled-exit the
|
|
// s_bgRunning mirror stays latched true; and without the worker's
|
|
// pre-init read guard the crash-recovery finish parks on a vector
|
|
// pull into the trapped replacement engine — either way this poll
|
|
// times out with the Run tool disabled forever.
|
|
await waitForRunToolEnabled(page);
|
|
|
|
// The crashed run's completion was delivered (cursor/finish body ran).
|
|
const crashReceipt = await page.evaluate(async (after: number) => {
|
|
const hooks = (globalThis as any).__ngspiceServiceTestHooks;
|
|
return await hooks.waitForAppliedGenerationAfter(after, 60000);
|
|
}, checkpoint);
|
|
expect(crashReceipt.generation, 'the crashed run applied its completion')
|
|
.toBeGreaterThan(checkpoint);
|
|
|
|
// The synthetic exit is visible in the event record.
|
|
const exitSeen = await page.evaluate((floor: number) =>
|
|
((window as any).__ngspiceEvents as Array<{ kind: string }>)
|
|
.slice(floor)
|
|
.some((e) => e.kind === 'exit'), eventFloor);
|
|
expect(exitSeen, 'a controlled-exit event reached the client').toBe(true);
|
|
|
|
// And the promised transparent restart: a full rerun on a fresh
|
|
// worker generation succeeds end to end.
|
|
const rerunGeneration = await runSimulation(page);
|
|
expect(rerunGeneration).toBeGreaterThan(crashReceipt.generation);
|
|
const generations = await page.evaluate(() =>
|
|
(globalThis as any).__ngspiceServiceTestHooks.snapshot());
|
|
expect(generations.retiredGenerations, 'the killed generation was retired')
|
|
.toContain(1);
|
|
|
|
expect(findNativeFailure([...testLogger.consoleLogs, ...testLogger.errors]),
|
|
'no wasm abort during the recovery').toBeUndefined();
|
|
});
|
|
|
|
test('a launch that dies before RUNNING still delivers its completion (E-12)', async ({ page, testLogger }) => {
|
|
await page.goto('/kicad/eeschema.html');
|
|
await waitForEditorReady(page);
|
|
await loadRectifier(page);
|
|
await openSimulator(page);
|
|
await waitForRunToolEnabled(page);
|
|
|
|
// Arm: the transport dies on the bg_run launch itself — after the
|
|
// native side published its run generation, before any RUNNING
|
|
// transition could fire. The retirement's synthetic exit then
|
|
// delivers this run's ONLY completion. (The arm keys on bg_run
|
|
// specifically, so frame-open plot restoration cannot consume it.)
|
|
const checkpoint = await page.evaluate(() => {
|
|
const hooks = (globalThis as any).__ngspiceServiceTestHooks;
|
|
hooks.dieOnNextBgRun();
|
|
return hooks.appliedGenerationCheckpoint() as number;
|
|
});
|
|
|
|
expect(await clickByTooltip(page, 'Run Simulation', { elementType: 'tool' }),
|
|
'Run tool').toBe(true);
|
|
|
|
// THE E-12 oracle: on the unfixed build the crash-exit IDLE carries
|
|
// generation 0 (its RUNNING never fired) and is deleted — the owned
|
|
// run's completion never applies and this receipt times out.
|
|
const receipt = await page.evaluate(async (after: number) => {
|
|
const hooks = (globalThis as any).__ngspiceServiceTestHooks;
|
|
return await hooks.waitForAppliedGenerationAfter(after, 60000);
|
|
}, checkpoint);
|
|
expect(receipt.generation, 'the dead launch applied its crash completion')
|
|
.toBeGreaterThan(checkpoint);
|
|
|
|
// Recovery stays intact: a rerun on the replacement generation works.
|
|
const rerunGeneration = await runSimulation(page);
|
|
expect(rerunGeneration).toBeGreaterThan(receipt.generation);
|
|
|
|
expect(findNativeFailure([...testLogger.consoleLogs, ...testLogger.errors]),
|
|
'no wasm abort during the recovery').toBeUndefined();
|
|
});
|
|
|
|
test('a corrupted get_vec length is clamped, not copied out of bounds (E-11)', async ({ page, testLogger }) => {
|
|
await page.goto('/kicad/eeschema.html');
|
|
await waitForEditorReady(page);
|
|
await loadRectifier(page);
|
|
await openSimulator(page);
|
|
await waitForRunToolEnabled(page);
|
|
|
|
// Arm BEFORE the run: the next vector pull reports a ~5e8-element
|
|
// length while its arrays stay ~101 elements. (Frame-open plot
|
|
// restoration also pulls vectors; whichever pull the arm hits, the
|
|
// corrupted answer flows through the same client prepare.)
|
|
await page.evaluate(() => {
|
|
(globalThis as any).__ngspiceServiceTestHooks.corruptNextGetVec();
|
|
});
|
|
|
|
// THE E-11 oracle: on the unfixed build the client copies v_length
|
|
// doubles from the small buffer. Observed death shape on this build:
|
|
// the 4 GiB std::vector throws an UNHANDLED std::length_error that
|
|
// exits the editor's main loop — the scheduler shuts down and the
|
|
// whole session is dead (an OOB trap is the sibling shape). Fixed,
|
|
// the length clamps to the transferred arrays and the run completes.
|
|
await runSimulation(page);
|
|
|
|
const scheduler = await page.evaluate(() => ({
|
|
dead: (globalThis as any).__wxScheduler?.dead === true,
|
|
terminal: (globalThis as any).__wxScheduler?.terminal === true,
|
|
}));
|
|
expect(scheduler.dead,
|
|
'the corrupted vector must not exit the editor main loop').toBe(false);
|
|
expect(scheduler.terminal, 'the editor instance must not be terminal').toBe(false);
|
|
const fatal = findNativeFailure([...testLogger.consoleLogs, ...testLogger.errors]);
|
|
expect(fatal, `no wasm trap from the corrupted vector (patterns: ${
|
|
FATAL_WASM_PATTERNS.join(', ')})`).toBeUndefined();
|
|
});
|
|
});
|