Milestone v1 (v2.0.0): Mokosh — Session Capture #1

Merged
strategy155 merged 297 commits from gsd/phase-04-harden-clean-up-optional into main 2026-05-31 15:34:17 +00:00
3 changed files with 327 additions and 55 deletions
Showing only changes of commit 116432a3cd - Show all commits

View File

@@ -3425,24 +3425,62 @@ async function assertA29(): Promise<AssertionResult> {
* setupInputLogging at line 77, setupNavigationLogging at line
* 99, setupErrorLogging at line 133, setupNetworkLogging at
* line 164) all fire on synthetic browser events dispatched
* on the harness page, producing UserEvent entries with each
* of the 5 type-values (click / input / navigation /
* js_error / network_error) in logs/events.json.
* in a probe https:// tab where the production content script
* is injected, producing UserEvent entries with each of the 5
* type-values (click / input / navigation / js_error /
* network_error) in logs/events.json.
*
* Trigger strategy (all on the harness page; no new tabs opened):
* - click: programmatic .click() on #probe-submit
* - input: set #probe-email.value + dispatch Event('input', bubbles:true)
* - navigation: history.pushState (intercepted at src/content/index.ts:121)
* Implementation note — MV3 content-script reachability (deviation):
* The plan as written assumed `<all_urls>` content_scripts coverage
* includes `chrome-extension://` URLs — empirically (Task 2
* verification dump 2026-05-20T17:36:25Z), it does NOT. The Chrome
* match-pattern docs are explicit: `<all_urls>` permits the schemes
* `http`, `https`, `file`, `ftp`, `urn` — NOT `chrome-extension`.
* The SW SAVE_ARCHIVE handler logged "Could not establish connection.
* Receiving end does not exist." when targeting the harness page,
* confirming no content script is present on chrome-extension://.
*
* A30 therefore creates a fresh `https://example.com` probe tab
* (mirrors A27's pattern, including DEC-011 Amendment 1 `tabs`
* permission), uses chrome.scripting.executeScript (default
* ISOLATED world — the content script's world) to dispatch all 5
* triggers, then SAVEs while the probe tab is active so the SW
* harvests events from the page where the content script is alive.
*
* In addition: even if `chrome-extension://` HAD been covered by
* `<all_urls>`, page-world `fetch()` from `page.evaluate(...)` would
* NOT have been intercepted by `src/content/index.ts:167`
* (`window.fetch = ...`) — content-script global mutations stay
* inside the ISOLATED world. executeScript with default ISOLATED
* world targeting + the content-script's own runtime view of fetch
* solves both issues with one mechanism.
*
* Trigger strategy (all inside an injected ISOLATED-world script on
* the example.com probe tab):
* - click: dispatch a real `MouseEvent('click')` on document.body
* - input: create a synthetic <input>, set value, dispatchEvent('input')
* - navigation: window.dispatchEvent(new PopStateEvent('popstate'))
* (production `popstate` listener at src/content/index.ts:111;
* NOTE: history.pushState was the plan-spec mechanism, but
* it triggers a Puppeteer CDP execution-context teardown
* — see deviation log + the popstate path is functionally
* equivalent since the production code in
* src/content/index.ts:121 wraps pushState by also firing
* handleNavigation which routes through the same
* listener as popstate at line 111).
* - js_error: window.dispatchEvent(new ErrorEvent('error', ...))
* - network_error: fetch(404-probe-url).catch(noop) — production
* fetch interception at src/content/index.ts:167 logs response.ok===false
* - network_error: fetch(404-probe-url).catch(noop) — runs in
* ISOLATED world so the patched window.fetch
* at src/content/index.ts:167 fires.
*
* Page-side dispatches all 5 triggers + settles + SAVE. Host-side
* driveA30 JSZip-parses logs/events.json and asserts each of the 5
* UserEvent.type literal values is present.
* Page-side opens probe tab + injects triggers + settles + SAVE.
* Host-side driveA30 JSZip-parses logs/events.json + asserts each of
* the 5 UserEvent.type literal values is present.
*
* FORBIDDEN_HOOK_STRINGS impact: NONE. A30 rides production listeners
* + existing helpers. Tier-1 inventory stays at 12.
* + chrome.tabs.* (`tabs` perm) + chrome.scripting.executeScript
* (`scripting` perm — already in manifest) + existing helpers. Tier-1
* inventory stays at 12.
*/
/** SAVE_ARCHIVE dispatch timeout for A30 — matches A24/A25/A27/A29. */
@@ -3450,18 +3488,26 @@ const A30_SAVE_ARCHIVE_TIMEOUT_MS = 15_000;
/** Pre-SAVE segment-settle window (10s rotation + 1s slack). */
const A30_SEGMENT_SETTLE_MS = 11_000;
/** Settle between trigger dispatches and SAVE so event handlers complete. */
const A30_TRIGGER_SETTLE_MS = 500;
/** 404 probe URL — chrome.tabs perm grant is irrelevant; fetch happens
* from the harness page realm. example.com is RFC 2606 reserved +
* serves a 404 reliably for unknown paths under headless Chrome. */
const A30_TRIGGER_SETTLE_MS = 1_000;
/** Wait after chrome.tabs.create for the tab navigation to complete so
* the content script attaches + production listeners are set up
* (mirrors A27_TAB_NAVIGATION_WAIT_MS = 1.5s). */
const A30_TAB_NAVIGATION_WAIT_MS = 1_500;
/** Probe tab URL — example.com is RFC 2606 reserved + serves stable
* HTML under headless Chrome (Plan 02-04 A27 fixture parity). */
const A30_PROBE_TAB_URL = 'https://example.com/';
/** 404 probe URL — same origin as the probe tab so the fetch is a
* same-origin GET (no CORS preflight noise). */
const A30_404_PROBE_URL = 'https://example.com/this-path-does-not-exist-404-probe-a30';
/**
* A30 — Event-log empirical (SPEC §10 #5 / REQ-user-event-log).
*
* Dispatches 5 synthetic browser events that exercise each of the
* production listeners; runs setupFreshRecording so event-log
* cleanup hasn't dropped anything; settles a segment; SAVEs. Host-side
* Creates a fresh `https://example.com` probe tab, injects all 5
* synthetic event triggers into the content script's ISOLATED world
* via chrome.scripting.executeScript so the production listeners fire,
* settles a segment, SAVEs while the probe tab is active so the SW
* harvests the content script's userEvents[] from that tab. Host-side
* driveA30 inspects logs/events.json from the produced zip and asserts
* each of the 5 UserEvent.type literal values appears at least once.
*
@@ -3476,6 +3522,8 @@ async function assertA30(): Promise<AssertionResult> {
diagnostics: [],
};
let probeTabId: number | undefined;
try {
diag(result, 'Step 1: setupFreshRecording (A30 owns its recording — clean event-log window)');
const setupResp = await setupFreshRecording();
@@ -3484,54 +3532,108 @@ async function assertA30(): Promise<AssertionResult> {
}
diag(result, 'Step 1 OK — REC state established');
diag(result, `Step 2: settle ${A30_SEGMENT_SETTLE_MS}ms for first segment rotation`);
diag(result, `Step 2: chrome.tabs.create(${A30_PROBE_TAB_URL}, active:true) — content script ISOLATED world is alive on https://, not on chrome-extension://`);
const probeTab = await chrome.tabs.create({ url: A30_PROBE_TAB_URL, active: true });
probeTabId = probeTab.id;
diag(result, `Step 2 result: probeTab.id=${probeTabId}, probeTab.url=${probeTab.url ?? '<pending>'}`);
if (probeTabId === undefined) {
throw new Error('chrome.tabs.create returned undefined tab.id');
}
diag(result, `Step 3: wait ${A30_TAB_NAVIGATION_WAIT_MS}ms for navigation + content script attach`);
await new Promise((r) => setTimeout(r, A30_TAB_NAVIGATION_WAIT_MS));
diag(result, `Step 4: settle ${A30_SEGMENT_SETTLE_MS}ms for first segment rotation`);
await new Promise((r) => setTimeout(r, A30_SEGMENT_SETTLE_MS));
diag(result, 'Step 3: click trigger — programmatic .click() on #probe-submit');
const submitBtn = document.querySelector<HTMLButtonElement>('#probe-submit');
if (submitBtn !== null) {
submitBtn.click();
} else {
diag(result, 'Step 3 WARN — #probe-submit missing; click trigger skipped');
}
diag(result, 'Step 5: chrome.scripting.executeScript — inject 5 synthetic triggers in ISOLATED world (content-script realm; fetch wrapper at src/content/index.ts:167 sees the fetch)');
const probeUrl = A30_404_PROBE_URL;
const injectionResults = await chrome.scripting.executeScript({
target: { tabId: probeTabId },
world: 'ISOLATED',
func: async (probe404Url: string): Promise<{
click: boolean;
input: boolean;
navigation: boolean;
jsError: boolean;
networkErrorTriggered: boolean;
fetchThrew: boolean;
}> => {
// click — synthetic MouseEvent on the document body. Production
// listener at src/content/index.ts:61 captures the click via
// document.addEventListener('click', ...).
const clickEvent = new MouseEvent('click', { bubbles: true, cancelable: true });
const clickDispatched = document.body.dispatchEvent(clickEvent);
diag(result, 'Step 4: input trigger — set #probe-email.value + dispatch input event');
const emailInput = document.querySelector<HTMLInputElement>('#probe-email');
if (emailInput !== null) {
emailInput.value = 'a30@probe.local';
emailInput.dispatchEvent(new Event('input', { bubbles: true }));
} else {
diag(result, 'Step 4 WARN — #probe-email missing; input trigger skipped');
}
// input — synthetic <input>, set value, dispatchEvent('input',
// bubbles:true). Production listener at src/content/index.ts:78
// captures via document.addEventListener('input', ...). Skips
// password type (line 82) — type='text' here.
const probeInput = document.createElement('input');
probeInput.type = 'text';
probeInput.id = 'a30-probe-input';
probeInput.value = 'a30@probe.local';
document.body.appendChild(probeInput);
const inputDispatched = probeInput.dispatchEvent(
new Event('input', { bubbles: true }),
);
probeInput.remove();
diag(result, 'Step 5: navigation trigger — history.pushState (production wrapper at src/content/index.ts:121 intercepts)');
history.pushState({}, '', window.location.pathname + '#a30-probe');
// navigation — window-level popstate event. Production listener
// at src/content/index.ts:111 captures via
// window.addEventListener('popstate', ...).
const navigationDispatched = window.dispatchEvent(
new PopStateEvent('popstate', { state: {} }),
);
diag(result, 'Step 6: js_error trigger — window.dispatchEvent(ErrorEvent("error"))');
window.dispatchEvent(new ErrorEvent('error', {
// js_error — window-level ErrorEvent. Production listener at
// src/content/index.ts:134 captures via
// window.addEventListener('error', ...).
const errorDispatched = window.dispatchEvent(
new ErrorEvent('error', {
message: 'a30-probe-js-error',
filename: 'a30-probe.js',
lineno: 1,
colno: 1,
}));
}),
);
diag(result, `Step 7: network_error trigger — fetch(${A30_404_PROBE_URL}) (.catch noop)`);
// network_error — fetch into a 404 path. The content script
// patches window.fetch at src/content/index.ts:167 in its
// ISOLATED world; this fetch is in the SAME ISOLATED world
// so it routes through the wrapper. response.ok===false →
// addUserEvent({type:'network_error'}) at line 171.
let fetchThrew = false;
try {
await fetch(A30_404_PROBE_URL);
await fetch(probe404Url);
} catch (fetchErr) {
diag(result, `Step 7 fetch threw (acceptable for network_error path): ${String(fetchErr)}`);
fetchThrew = true;
}
diag(result, `Step 8: settle ${A30_TRIGGER_SETTLE_MS}ms so async handlers (fetch.then) complete`);
return {
click: clickDispatched,
input: inputDispatched,
navigation: navigationDispatched,
jsError: errorDispatched,
networkErrorTriggered: true,
fetchThrew,
};
},
args: [probeUrl],
});
const injectionSummary = injectionResults[0]?.result ?? null;
diag(result, `Step 5 result: ${JSON.stringify(injectionSummary)}`);
diag(result, `Step 6: settle ${A30_TRIGGER_SETTLE_MS}ms so async handlers (fetch.then) complete + userEvents[] populates`);
await new Promise((r) => setTimeout(r, A30_TRIGGER_SETTLE_MS));
diag(result, 'Step 9: dispatch SAVE_ARCHIVE');
diag(result, 'Step 7: dispatch SAVE_ARCHIVE (probe tab is the active tab; SW will harvest from there)');
const ack = await sendMessageWithTimeout<{ success: boolean; error?: string }>(
{ type: 'SAVE_ARCHIVE' },
A30_SAVE_ARCHIVE_TIMEOUT_MS,
'SAVE_ARCHIVE (A30)',
);
diag(result, `Step 9 result: ${JSON.stringify(ack)}`);
diag(result, `Step 7 result: ${JSON.stringify(ack)}`);
result.checks.push({
name: 'A30.1: SAVE_ARCHIVE ack received with success=true',
@@ -3544,6 +3646,15 @@ async function assertA30(): Promise<AssertionResult> {
} catch (err) {
result.error = err instanceof Error ? err.message : String(err);
diag(result, `THREW: ${result.error}`);
} finally {
// T-02-04-04 mitigation parity: cleanup probe tab with silent-ignore.
if (probeTabId !== undefined) {
try {
await chrome.tabs.remove(probeTabId);
} catch (rmErr) {
diag(result, `(probe tab cleanup ignored: ${String(rmErr)})`);
}
}
}
return result;

View File

@@ -99,6 +99,8 @@ import {
driveA28,
// Plan 03-01 — rrweb DOM verification (SPEC §10 #4 / REQ-rrweb-dom-buffer)
driveA29,
// Plan 03-02 — event-log verification (SPEC §10 #5 / REQ-user-event-log)
driveA30,
getManifestVersion,
} from './lib/harness-page-driver';
import {
@@ -267,7 +269,7 @@ async function assertA0_GrepGate(): Promise<{
*/
async function main(): Promise<number> {
process.stdout.write('\nMokosh Plan 01-13 + 01-14 + 02-04 — UAT harness orchestrator\n');
process.stdout.write('Architecture: A0 pre-flight + extension-internal page driver (A1..A14, A15..A17, A18..A22, A23, A24, A25, A26, A27, A28, A29)\n');
process.stdout.write('Architecture: A0 pre-flight + extension-internal page driver (A1..A14, A15..A17, A18..A22, A23, A24, A25, A26, A27, A28, A29, A30)\n');
process.stdout.write('='.repeat(72) + '\n');
// A0 pre-flight (no Chrome launch needed; runs against built dist/).
@@ -339,6 +341,10 @@ async function main(): Promise<number> {
// of rrweb/session.json from the just-produced zip.
const driveA29Wrapped: (page: import('puppeteer').Page) => Promise<AssertionRecord> =
(page) => driveA29(page, handles.downloadsDir);
// Plan 03-02 — driveA30 needs downloadsDir for host-side JSZip parse
// of logs/events.json from the just-produced zip.
const driveA30Wrapped: (page: import('puppeteer').Page) => Promise<AssertionRecord> =
(page) => driveA30(page, handles.downloadsDir);
const drivers: ReadonlyArray<{
readonly name: string;
@@ -441,6 +447,13 @@ async function main(): Promise<number> {
// rrweb/session.json and asserts the EventType enum surfaces
// (Meta=4, FullSnapshot=2, IncrementalSnapshot=3) are present.
{ name: 'A29', drive: driveA29Wrapped },
// Plan 03-02 A30: event-log verification (SPEC §10 #5).
// A30 owns its SAVE because event-log cleanup runs every 60s
// (src/content/index.ts CLEANUP_INTERVAL_MS=60_000) and we need a
// fresh event-log window for the 5 synthetic triggers. Host-side
// driveA30 JSZip-parses logs/events.json and asserts presence of
// each of the 5 UserEvent.type literal values.
{ name: 'A30', drive: driveA30Wrapped },
];
const buffers = { swConsole: handles.swConsole, offConsole: handles.offConsole };

View File

@@ -44,6 +44,7 @@ import type { Page } from 'puppeteer';
import type { AssertionRecord, CheckRecord } from './assertions';
import { assertArchiveShape, extractEntryToFile } from './zip';
import type { UserEvent } from '../../../src/shared/types';
/**
* Extended assertion-record shape for A5/A12/A13 which return
@@ -1998,3 +1999,150 @@ export async function driveA29(
error: pageResult.error,
};
}
/* ─── Plan 03-02 — driveA30 (event-log verification host-side) ──────── */
/** Canonical 5-tuple of UserEvent.type literal values per
* CON-event-log-schema + src/shared/types.ts:126. Driver iterates this
* list to push one presence-check per type. */
const A30_EXPECTED_TYPES: ReadonlyArray<UserEvent['type']> = [
'click',
'input',
'navigation',
'js_error',
'network_error',
];
/**
* Drive A30 (Plan 03-02 — SPEC §10 #5 / REQ-user-event-log).
*
* Page-side assertA30 dispatches 5 synthetic event triggers
* (click/input/navigation/js_error/network_error) + setupFreshRecording
* + SAVE. Host-side driveA30 JSZip-parses logs/events.json from the
* produced zip and asserts each of the 5 UserEvent.type literal values
* appears at least once.
*
* Filter-pipeline form (no `continue`) per CLAUDE.md Control Flow §.
*
* Checks (6 total — 1 page-side + 5 host-side):
* - A30.1: SAVE_ARCHIVE ack success (page-side)
* - A30.2: logs/events.json contains >=1 'click' event
* - A30.3: logs/events.json contains >=1 'input' event
* - A30.4: logs/events.json contains >=1 'navigation' event
* - A30.5: logs/events.json contains >=1 'js_error' event
* - A30.6: logs/events.json contains >=1 'network_error' event
*
* @param page - The harness page from `launchHarnessBrowser`.
* @param downloadsDir - Absolute path to the per-run downloads directory.
* @returns AssertionRecord with 6 merged checks.
*/
export async function driveA30(
page: Page,
downloadsDir: string,
): Promise<AssertionRecord> {
// Phase 1 — page-side orchestration + SAVE.
const pageResult = await page.evaluate(async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any -- evaluate runs in browser context where Window types are loose.
const harness = (window as any).__mokoshHarness;
const r: AssertionRecord = await harness.assertA30();
return r;
}) as AssertionRecord;
const mergedChecks: CheckRecord[] = pageResult.checks.slice();
const mergedDiagnostics: string[] = pageResult.diagnostics.slice();
// Phase 2 — locate the produced zip.
const zipPath = findLatestZip(downloadsDir);
if (zipPath === null) {
mergedChecks.push({
name: 'A30.0: at least one zip present in downloadsDir',
expected: '>=1 zip',
actual: 'no zip in downloadsDir',
passed: false,
});
return {
passed: false,
name: pageResult.name,
checks: mergedChecks,
diagnostics: mergedDiagnostics,
error: pageResult.error,
};
}
mergedDiagnostics.push(`A30 zipPath=${zipPath}`);
// Phase 3 — load + inspect logs/events.json.
const zipBytes = readFileSync(zipPath);
const zip = await JSZip.loadAsync(zipBytes);
const eventsFile = zip.file('logs/events.json');
mergedChecks.push({
name: 'A30.0a: logs/events.json entry exists in zip',
expected: true,
actual: eventsFile !== null,
passed: eventsFile !== null,
});
if (eventsFile === null) {
return {
passed: false,
name: pageResult.name,
checks: mergedChecks,
diagnostics: mergedDiagnostics,
error: pageResult.error,
};
}
const eventsRaw = await eventsFile.async('string');
let userEvents: UserEvent[] = [];
let parseErr: string | null = null;
try {
userEvents = JSON.parse(eventsRaw) as UserEvent[];
} catch (err) {
parseErr = err instanceof Error ? err.message : String(err);
}
if (parseErr !== null) {
mergedChecks.push({
name: 'A30.0b: logs/events.json parses as JSON',
expected: 'JSON.parse success',
actual: `<error: ${parseErr}>`,
passed: false,
});
return {
passed: false,
name: pageResult.name,
checks: mergedChecks,
diagnostics: mergedDiagnostics,
error: pageResult.error,
};
}
// Filter-pipeline form per CLAUDE.md Control Flow §.
const typeCountsMap = new Map<string, number>();
for (const expectedType of A30_EXPECTED_TYPES) {
typeCountsMap.set(expectedType, userEvents.filter((e) => e.type === expectedType).length);
}
mergedDiagnostics.push(`A30 userEvents.length=${userEvents.length}`);
const typeCountsRepr = [...typeCountsMap.entries()].map(([t, n]) => `${t}=${n}`).join(',');
mergedDiagnostics.push(`A30 type counts: ${typeCountsRepr}`);
let checkIndex = 2;
for (const expectedType of A30_EXPECTED_TYPES) {
const count = typeCountsMap.get(expectedType) ?? 0;
mergedChecks.push({
name: `A30.${checkIndex}: logs/events.json contains at least one '${expectedType}' event`,
expected: `>=1 ${expectedType}`,
actual: count,
passed: count > 0,
});
checkIndex += 1;
}
const mergedPassed = mergedChecks.every((c) => c.passed);
return {
passed: mergedPassed,
name: pageResult.name,
checks: mergedChecks,
diagnostics: mergedDiagnostics,
error: pageResult.error,
};
}