refactor(client): dé-poller resolveAccountReliably — barrière au lieu de retry

Le polling est un anti-pattern NextGraph (par abonnement). La résolution de compte
retentait ×8 la lecture du shim tant qu'elle rendait 0 (lag de sync) — c'est du
polling. Remplacé par la BARRIÈRE d'abonnement, déjà le mécanisme de open-repo :

- `resolveAccountReliably` : `await ensureRepoOpen(did🆖${privateStoreId})`
  (subscribe + attendre le 1er State — le shim vit dans le graphe du private store),
  PUIS lecture UNIQUE. Après la barrière, 0 ligne = compte réellement inexistant →
  provision 1×, lignes présentes = réutilisé (garantie NO-FORK préservée). Plus de
  boucle de re-lecture.
- timed-out (barrière expirée) : throw explicite, NE provisionne PAS (un provision
  sur sync incomplète re-forkerait). Le « trop long » est un signal, pas un feu vert.
- fake ng sans doc_subscribe : ensureRepoOpen no-op → lecture immédiate (unit intact).
- `_forceOpenedSyncState` : helper test-only (underscore, non ré-exporté).

anti-fork.test.ts réécrit (5 tests : no-fork, neuf→1 provision, idempotence, fake
no-op, timed-out→throw) ; plus aucun test de comptage de retry.

gate : tsc propre ; bun test 117. e2e À RE-VALIDER quand le broker répond (dégradé
ce jour : crash Chromium post-connexion) — la barrière ensureRepoOpen est déjà
validée e2e (CONTRAT 3 + reconnexion) en broker sain. provisionRetry devient un
champ mort de StoreRegistryDeps (nettoyage ultérieur).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Sylvain Duchesne
2026-07-09 13:28:56 +02:00
parent d3a2c33b92
commit 45dbd9a33a
3 changed files with 153 additions and 146 deletions
+11
View File
@@ -110,6 +110,17 @@ export function resetOpenedRepos(): void {
boundSessionId = null; boundSessionId = null;
} }
/**
* @internal TEST-ONLY — force a nuri into the opened registry with a given sync
* state, without going through `doc_subscribe`. Used by `anti-fork.test.ts` to
* simulate a timed-out barrier without waiting the full `OPEN_TIMEOUT_MS` (8s).
* Do NOT call from production code.
*/
export function _forceOpenedSyncState(nuri: Nuri, state: SyncState): void {
opened.add(nuri);
syncState.set(nuri, state);
}
/** /**
* The bootstrap sync state of `nuri` (lib-internal accessor; NOT a reactive * The bootstrap sync state of `nuri` (lib-internal accessor; NOT a reactive
* app-facing hook — that is a later phase). Returns `"unknown"` if the repo was * app-facing hook — that is a later phase). Returns `"unknown"` if the repo was
+63 -57
View File
@@ -36,27 +36,30 @@ import { docCreate, sparqlUpdate, sparqlQuery } from "./docs";
import { getStoreRegistryDeps } from "./polyfill"; import { getStoreRegistryDeps } from "./polyfill";
import { ensureRepoOpen } from "./open-repo"; import { ensureRepoOpen } from "./open-repo";
// --- provisioning anti-fork retry budget (polyfill-era) -------------------- // --- provisioning anti-fork guard: barrier-first (polyfill-era) ------------
// //
// The shim (account→docs trust root) lives in the shared wallet's PRIVATE store. // The shim (account→docs trust root) lives in the shared wallet's PRIVATE store.
// On a fresh session over a persistent wallet, that store's repo may not have // On a fresh session over a persistent wallet, that repo may not have synced yet
// synced yet from the broker: a targeted account resolve then reads 0 rows even // from the broker: a targeted account resolve would then read 0 rows even though
// though the account WAS persisted in an earlier session. If ensureAccount took // the account WAS persisted in an earlier session. If ensureAccount took that 0
// that 0 at face value it would RE-PROVISION a second set of scope documents — // at face value it would RE-PROVISION a second set of scope documents — an account
// an account FORK: one session writes/reads one set, another (or the same after // FORK: one session writes/reads one set, another (or the same after a cache drop)
// a cache drop) the other, EMPTY set → the user "loses" their data on reconnect. // the other, EMPTY set → the user "loses" their data on reconnect.
// //
// Guard: before deciding an account is genuinely new, re-read it with a BOUNDED // Guard (barrier-first): BEFORE reading the shim, open/subscribe the private-store
// backoff (open + await the anchor repo, then retry the targeted query). Only if // repo and AWAIT its first `State` push — the deterministic sync barrier (CONTRACT 3
// EVERY attempt within the budget still reads 0 do we treat it as a first-time // in e2e/). After the barrier, presence is GUARANTEED and absence DEFINITIVE. Then
// account and provision. A pre-existing account merely lagging by sync is thus // read the shim ONCE: 0 rows = account genuinely absent → provision exactly once;
// found on a later attempt and REUSES its docs (no fork); a truly new account // rows present = account found → reuse it (NO-FORK). No retry loop, no poll.
// exhausts the budget cheaply and is provisioned exactly once. The budget is a //
// hard ceiling — never an unbounded poll — and is INJECTED (StoreRegistryDeps. // Timed-out barrier (conservative): if `ensureRepoOpen` exits via the 8-second
// provisionRetry): production defaults to ~8 attempts / ≲8.5s; a synchronous // fallback timeout (getSyncState → "timed-out") the sync is UNCONFIRMED — we do NOT
// in-memory store (unit fake `ng`) passes `attempts: 1` to skip it (a 0-row read // provision blindly (that would re-fork). Instead we surface a clear error so the
// is authoritative there — no lag to wait out). Disappears at the real // caller can retry the full flow rather than silently creating a duplicate account.
// multi-store migration (deterministic per-user store NURIs make this moot). // The timeout is rare (pathological broker) and a hard error there is far safer than
// a silent fork. Disappears at the real multi-store migration (per-user store NURIs
// make this moot — barrier becomes the native store-open which is always confirmed).
import { getSyncState } from "./open-repo";
import { escapeLiteral, escapeIri, assertNuri } from "./sparql"; import { escapeLiteral, escapeIri, assertNuri } from "./sparql";
import type { Nuri, Scope } from "./types"; import type { Nuri, Scope } from "./types";
@@ -280,48 +283,51 @@ export async function resolveAccount(id: string): Promise<AccountRecord | null>
} }
/** /**
* Resolve ONE account, retrying a BOUNDED number of times before giving up, to * Resolve ONE account after waiting for the private-store sync barrier, so a
* distinguish a genuinely-absent account from one whose shim record has simply * 0-row result is DEFINITIVE (account genuinely absent) rather than ambiguous
* not synced yet on a fresh session (the account-FORK guard — see the * (broker lag). This is the ANTI-FORK guard for the polyfill-era shim.
* PROVISION_RETRY_* budget note at the top of this module).
* *
* On the FIRST miss it opens/subscribes the shim anchor repo (the shared * Flow:
* wallet's private store, where the shim lives) so the broker pushes its state, * 1. Cache hit → return immediately (session-idempotent, no I/O).
* then re-reads with capped exponential backoff. Returns the record as soon as * 2. Open/subscribe the private-store repo (`did:ng:${privateStoreId}`) and
* any attempt finds it (populating `accountCache` via resolveAccount), or `null` * await the first `State` push — the deterministic sync barrier (CONTRACT 3).
* if the whole budget is exhausted with 0 rows — the only case in which the * After the barrier, presence is guaranteed and absence definitive.
* caller may treat the account as new and provision it. * 3. If the barrier TIMED-OUT (getSyncState → "timed-out"), the sync is
* unconfirmed — provisioning here would risk a fork. Throw a clear error
* instead so the caller can retry the full login flow. This is the
* conservative-safe choice: a hard error is recoverable; a silent fork is not.
* 4. Read the shim ONCE. 0 rows = genuinely absent → caller provisions exactly
* once. Rows present → reuse (NO-FORK preserved).
* *
* No-op fast path: a cache hit returns immediately with no retries. With the * With the unit fake `ng` (no `doc_subscribe`), `ensureRepoOpen` is a no-op
* unit fake `ng` (no doc_subscribe, synchronous store) a present account is * (getSyncState → "unknown") and the single read is immediate — synchronous
* found on attempt 0 and an absent one exhausts the (short-sleep) budget without * behaviour preserved, no lag to wait out.
* changing behaviour — the fork can only arise against a real, lagging broker.
*/ */
async function resolveAccountReliably(id: string): Promise<AccountRecord | null> { async function resolveAccountReliably(id: string): Promise<AccountRecord | null> {
// Cache hit → session-idempotent, no query, no fork risk. // Cache hit → session-idempotent, no query, no fork risk.
const cached = accountCache.get(accountKey(id)); const cached = accountCache.get(accountKey(id));
if (cached) return cached; if (cached) return cached;
const budget = getStoreRegistryDeps().provisionRetry; // The shim lives in the private store. Open/subscribe it and await the first
const attempts = Math.max(1, budget.attempts ?? 8); // `State` push (the sync barrier). NURI: `did:ng:${session.privateStoreId}`.
const baseMs = budget.baseMs ?? 300; const nuri = await anchorNuri();
const maxStepMs = budget.maxStepMs ?? 1500; await ensureRepoOpen(nuri);
for (let attempt = 0; attempt < attempts; attempt++) { // Conservative timed-out guard: if the barrier did not confirm sync,
const found = await resolveAccount(id); // provisioning blind risks creating a fork. Raise a clear error — the caller
if (found) return found; // (or the app's retry-login flow) should re-attempt when the broker is reachable.
if (attempt === attempts - 1) break; // budget exhausted → genuinely absent // "unknown" means the fake-ng no-op path: no barrier semantics → safe to read.
// The shim anchor repo may not be synced on a fresh session; open it once so const syncResult = getSyncState(nuri);
// the broker pushes its state, then back off before the next targeted read. if (syncResult === "timed-out") {
try { throw new Error(
await ensureRepoOpen(await anchorNuri()); "[storeRegistry] sync barrier timed out for the private store — " +
} catch { "shim state unconfirmed, refusing to provision to avoid account fork. " +
/* tolerant: a failed open just leaves the next read to behave as before */ "Retry when the broker is reachable.",
} );
const step = Math.min(baseMs * 2 ** attempt, maxStepMs);
await new Promise((r) => setTimeout(r, step));
} }
return null;
// Barrier reached (or fake-ng "unknown" path): single read, result is definitive.
return resolveAccount(id);
} }
/** All known accounts (from the shim). */ /** All known accounts (from the shim). */
@@ -346,13 +352,13 @@ export async function ensureAccount(id: string): Promise<AccountRecord> {
// HOT PATH: targeted O(1) lookup — does THIS account already exist? — instead // HOT PATH: targeted O(1) lookup — does THIS account already exist? — instead
// of a full-shim scan (loadShim). Off the read/write hot path entirely. // of a full-shim scan (loadShim). Off the read/write hot path entirely.
// //
// ANTI-FORK (polyfill-era): resolve with a BOUNDED retry, NOT a single read. // ANTI-FORK (polyfill-era): open the private-store repo and await its first
// A single 0-row read on a fresh, still-syncing session would misfire as // `State` push (the sync barrier) BEFORE reading the shim, so a 0-row result
// "new account" and RE-PROVISION a second set of scope docs (the fork — see // is DEFINITIVE rather than a lag artifact. Only after the barrier do we treat
// the PROVISION_RETRY_* note at the top of this module). We only fall through // 0 rows as "genuinely new" and provision. A cache hit inside
// to provisioning once the whole retry budget has confirmed 0 rows. A cache // resolveAccountReliably keeps same-session resolves free and deterministic
// hit inside resolveAccountReliably keeps same-session resolves free and // (the cached record wins over any re-provision). Throws if the barrier timed
// deterministic (the cached record wins over any re-provision). // out — safer than provisioning blind and creating a fork.
const existing = await resolveAccountReliably(id); const existing = await resolveAccountReliably(id);
if (existing) return existing; if (existing) return existing;
+79 -89
View File
@@ -2,16 +2,17 @@
* anti-fork.test.ts — behavioral tests for the ANTI-FORK guard in ensureAccount / * anti-fork.test.ts — behavioral tests for the ANTI-FORK guard in ensureAccount /
* resolveAccountReliably (src/store-registry.ts). * resolveAccountReliably (src/store-registry.ts).
* *
* The guard is: before treating a 0-row read as "genuinely new account" and * The guard is: BEFORE reading the shim, open the private-store repo and await
* provisioning a new set of scope docs, the registry retries a bounded number of * the first `State` push (the deterministic sync barrier, CONTRACT 3). After the
* times (with ensureRepoOpen + backoff between each). If the account IS found on * barrier, 0 rows is DEFINITIVE (account genuinely absent) → provision exactly
* any attempt, it is reused (no second set of docs = no fork). Only if EVERY * once; rows present → reuse (NO-FORK). No retry loop, no polling.
* attempt still reads 0 rows does the registry treat it as a first-time account
* and provision exactly once.
* *
* This file uses a FAKE `ng` that simulates broker sync lag: the sparql_query for * With the unit fake `ng` (no `doc_subscribe`), `ensureRepoOpen` is a no-op
* the account record returns 0 rows for the first K calls, then returns real rows. * (getSyncState → "unknown") and the single shim read is immediate — no lag to
* We control the retry budget via `provisionRetry.attempts`. * wait out, synchronous behaviour preserved.
*
* Timed-out barrier: if the barrier expires without a `State`, we refuse to
* provision (throwing a clear error) rather than risk a fork.
*/ */
import { describe, it, expect, mock, beforeEach, afterAll } from "bun:test"; import { describe, it, expect, mock, beforeEach, afterAll } from "bun:test";
@@ -26,7 +27,7 @@ import {
resetStoreRegistry, resetStoreRegistry,
resetConfig, resetConfig,
} from "../src/polyfill"; } from "../src/polyfill";
import { resetOpenedRepos } from "../src/open-repo"; import { resetOpenedRepos, _forceOpenedSyncState } from "../src/open-repo";
afterAll(() => { afterAll(() => {
resetConfig(); resetConfig();
@@ -38,11 +39,11 @@ afterAll(() => {
const SESSION: RegistrySession = { sessionId: "sid-af", privateStoreId: "PRIV-AF" }; const SESSION: RegistrySession = { sessionId: "sid-af", privateStoreId: "PRIV-AF" };
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
// Fake ng with sync-lag simulation // Fake ng — simple in-memory store (no doc_subscribe → fake-ng no-op path)
// //
// `lagCalls` controls how many account-record reads return 0 rows before the // Queries return data from the `quads` array immediately. No lag simulation:
// real data is visible. Each call to sparql_query that matches the account // the barrier mechanism (ensureRepoOpen) is a no-op in the fake-ng path, so
// SELECT decrements the counter; once it hits 0 the real data is returned. // the single shim read after "the barrier" is already authoritative.
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
interface Quad { g: string; s: string; p: string; o: string } interface Quad { g: string; s: string; p: string; o: string }
@@ -60,7 +61,7 @@ function unescapeLiteral(s: string): string {
return out; return out;
} }
function makeLaggedNg(lagCalls: number) { function makeFakeNg() {
const quads: Quad[] = []; const quads: Quad[] = [];
let docCounter = 0; let docCounter = 0;
let accountQueryCount = 0; let accountQueryCount = 0;
@@ -100,15 +101,9 @@ function makeLaggedNg(lagCalls: number) {
const anchor = a[3] as string | undefined; const anchor = a[3] as string | undefined;
if (query.includes("<urn:ng-eventually:shim:id>")) { if (query.includes("<urn:ng-eventually:shim:id>")) {
// This is an account SELECT (resolveAccount / loadShim) // Account SELECT (resolveAccount / loadShim)
accountQueryCount++; accountQueryCount++;
// Still in the lag window → pretend the store hasn't synced yet
if (accountQueryCount <= lagCalls) {
return { results: { bindings: [] } };
}
// Lag lifted → return real data from the quads store
const subjM = query.match(/GRAPH <[^>]+>\s*\{\s*<([^>]+)>\s+a\s+<urn:ng-eventually:shim:Account>/); const subjM = query.match(/GRAPH <[^>]+>\s*\{\s*<([^>]+)>\s+a\s+<urn:ng-eventually:shim:Account>/);
const onlySubject = subjM ? subjM[1]! : null; const onlySubject = subjM ? subjM[1]! : null;
const bySubject = new Map<string, Record<string, string>>(); const bySubject = new Map<string, Record<string, string>>();
@@ -149,12 +144,14 @@ function makeLaggedNg(lagCalls: number) {
}; };
} }
function inject(laggedNg: ReturnType<typeof makeLaggedNg>, attempts: number) { function inject(fakeNg: ReturnType<typeof makeFakeNg>) {
configure({ ng: laggedNg as any, useShape: (() => {}) as any }); // No doc_subscribe → ensureRepoOpen is a no-op (fake-ng path).
configure({ ng: fakeNg as any, useShape: (() => {}) as any });
configureStoreRegistry({ configureStoreRegistry({
getSession: async () => SESSION, getSession: async () => SESSION,
normalizeId: (u) => u.trim().toLowerCase(), normalizeId: (u) => u.trim().toLowerCase(),
provisionRetry: { attempts, baseMs: 0, maxStepMs: 0 }, // provisionRetry is now unused by the barrier mechanism; kept for API compat.
provisionRetry: { attempts: 1 },
}); });
resetRegistryCache(); resetRegistryCache();
resetOpenedRepos(); resetOpenedRepos();
@@ -164,104 +161,97 @@ function inject(laggedNg: ReturnType<typeof makeLaggedNg>, attempts: number) {
// Tests // Tests
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
describe("anti-fork: resolveAccountReliably / ensureAccount", () => { describe("anti-fork: barrier-first resolveAccountReliably / ensureAccount", () => {
beforeEach(() => { beforeEach(() => {
resetRegistryCache(); resetRegistryCache();
resetOpenedRepos(); resetOpenedRepos();
}); });
it("(a) sync lag: 0 rows for K-1 calls then real rows → reuses existing docs (no fork)", async () => { it("(a) NO-FORK: account already in shim → reused, 0 new doc_create", async () => {
// budget = 3 attempts; lag = 1 call (the first targeted read returns 0, // Provision an account in an initial "session" (quads are written).
// but the second (after retry) returns the real data written in the same const fakeNg = makeFakeNg();
// session before the cache was dropped). We simulate a scenario where the inject(fakeNg);
// account IS already provisioned (written to quads by a previous const first = await ensureAccount("LauraBarrier");
// ensureAccount call in an earlier "session"), then we reset the cache and expect(fakeNg.doc_create).toHaveBeenCalledTimes(3);
// re-call with lag to simulate broker sync delay.
const laggedNg = makeLaggedNg(0); // no lag — first provision a real account // Simulate a fresh session: clear caches but keep quads intact (same fakeNg).
inject(laggedNg, 1); // In the barrier model the single post-barrier read finds the account immediately.
const first = await ensureAccount("LauraLag"); resetRegistryCache();
expect(laggedNg.doc_create).toHaveBeenCalledTimes(3); resetOpenedRepos();
// Now simulate a fresh session with lag: cache cleared, first K reads are 0 const second = await ensureAccount("LauraBarrier");
// We reuse the SAME quads store (same laggedNg with quads intact) but
// rebuild with lag > 0 so the first read returns 0, then real data.
const laggedNg2 = makeLaggedNg(1); // 1 empty read before real data
// Copy the quads from the first ng into laggedNg2 so it "knows" the account
for (const q of laggedNg._quads) {
laggedNg2._quads.push(q);
}
inject(laggedNg2, 3); // 3 attempts, lag=1 → second attempt succeeds
const second = await ensureAccount("LauraLag"); // ANTI-FORK: same scope docs returned, no new provisioning
// ANTI-FORK: must return the SAME scope docs — no new doc_create
expect(second.docPublic).toBe(first.docPublic); expect(second.docPublic).toBe(first.docPublic);
expect(second.docProtected).toBe(first.docProtected); expect(second.docProtected).toBe(first.docProtected);
expect(second.docPrivate).toBe(first.docPrivate); expect(second.docPrivate).toBe(first.docPrivate);
// doc_create must NOT have been called (the lagged ng2 finds the account on retry) expect(fakeNg.doc_create).toHaveBeenCalledTimes(3); // still exactly 3, not 6
expect(laggedNg2.doc_create).toHaveBeenCalledTimes(0);
}); });
it("(b) always-0 rows (genuinely new account): provisions exactly ONCE (1 set of docs)", async () => { it("(b) genuinely new account (fake returns 0 rows) → provisioned exactly once (3 doc_create)", async () => {
// Lag longer than the retry budget → account is treated as new const fakeNg = makeFakeNg(); // empty quads → 0 rows on any account query
const laggedNg = makeLaggedNg(999); // always returns 0 for account queries inject(fakeNg);
inject(laggedNg, 2); // budget: 2 attempts, both see 0 → provision
const rec = await ensureAccount("NewUser"); const rec = await ensureAccount("BrandNewUser");
// Exactly 3 doc_create calls (1 set of scope docs) // Exactly 3 doc_create calls (1 set of scope docs, no fork)
expect(laggedNg.doc_create).toHaveBeenCalledTimes(3); expect(fakeNg.doc_create).toHaveBeenCalledTimes(3);
expect(rec.docPublic).toMatch(/^did:ng:o:doc/); expect(rec.docPublic).toMatch(/^did:ng:o:doc/);
expect(rec.docProtected).not.toBe(rec.docPublic); expect(rec.docProtected).not.toBe(rec.docPublic);
expect(rec.docPrivate).not.toBe(rec.docProtected); expect(rec.docPrivate).not.toBe(rec.docProtected);
// Verify budget was fully consumed: both attempts fired before provisioning // Single read — no retry loop
expect(laggedNg.getAccountQueryCount()).toBe(2); expect(fakeNg.getAccountQueryCount()).toBe(1);
}); });
it("idempotence within a session: calling ensureAccount twice never creates 2 sets", async () => { it("(c) idempotence within a session: calling ensureAccount twice never creates 2 sets", async () => {
const laggedNg = makeLaggedNg(0); // no lag const fakeNg = makeFakeNg();
inject(laggedNg, 1); inject(fakeNg);
const a = await ensureAccount("SameUser"); const a = await ensureAccount("SameUser");
const b = await ensureAccount("SameUser"); const b = await ensureAccount("SameUser");
expect(b).toEqual(a); expect(b).toEqual(a);
// Must still be exactly 3 (not 6) // Must still be exactly 3 (not 6)
expect(laggedNg.doc_create).toHaveBeenCalledTimes(3); expect(fakeNg.doc_create).toHaveBeenCalledTimes(3);
}); });
it("retry budget of 1 (sync fake): a genuinely-absent account provisions without extra retries", async () => { it("(d) fake-ng no-op barrier path: single shim read, no doc_subscribe calls", async () => {
const laggedNg = makeLaggedNg(0); // will never return lag (account not in quads) // The fake ng has no doc_subscribe → ensureRepoOpen is a no-op (getSyncState → "unknown").
inject(laggedNg, 1); // resolveAccountReliably must proceed to the single read without waiting or throwing.
const fakeNg = makeFakeNg(); // no doc_subscribe
inject(fakeNg);
await ensureAccount("BrandNew"); // Provision once, then verify a second resolve (fresh cache) finds it immediately.
const first = await ensureAccount("FakeNgUser");
resetRegistryCache();
resetOpenedRepos();
// 3 scope docs created exactly once const second = await ensureAccount("FakeNgUser");
expect(laggedNg.doc_create).toHaveBeenCalledTimes(3);
// The targeted account query was called exactly once (no retries at attempts=1) // Same docs reused (no fork), exactly 3 total doc_create across both calls
const accountQueryCalls = laggedNg.getAccountQueryCount(); expect(second.docPublic).toBe(first.docPublic);
expect(accountQueryCalls).toBe(1); expect(fakeNg.doc_create).toHaveBeenCalledTimes(3);
// The account query was called exactly twice (once per ensureAccount, no retries)
expect(fakeNg.getAccountQueryCount()).toBe(2);
}); });
it("lag of exactly (attempts-1): found on last allowed attempt → reused, not forked", async () => { it("(e) timed-out barrier: resolveAccountReliably throws rather than provisioning", async () => {
// 3 attempts, lag=2 (first 2 calls return 0, 3rd returns real data). // Simulate: the private store nuri is already in the `opened` set but with
// We pre-populate the quads by provisioning with a fresh no-lag ng first. // sync state "timed-out" (forced via _forceOpenedSyncState so the test
const seedNg = makeLaggedNg(0); // does not have to wait 8s for the real OPEN_TIMEOUT_MS to fire).
inject(seedNg, 1); const fakeNg = makeFakeNg(); // empty quads → 0 rows
const original = await ensureAccount("EdgeUser"); inject(fakeNg);
expect(seedNg.doc_create).toHaveBeenCalledTimes(3);
// Now re-inject with the same quads but with lag=2, budget=3 // Force the private-store nuri ("did:ng:PRIV-AF") into timed-out state.
const laggedNg = makeLaggedNg(2); // resolveAccountReliably calls anchorNuri() → `did:ng:${privateStoreId}`.
for (const q of seedNg._quads) laggedNg._quads.push(q); _forceOpenedSyncState("did:ng:PRIV-AF", "timed-out");
inject(laggedNg, 3);
const retried = await ensureAccount("EdgeUser"); // ensureAccount must throw (conservative anti-fork guard: do not provision blind)
await expect(ensureAccount("TimedOutUser")).rejects.toThrow(
/sync barrier timed out/,
);
// Found on the 3rd attempt (lag=2 means first 2 return 0, 3rd returns real) // Must NOT have created any scope docs (refusing to provision)
expect(retried.docPublic).toBe(original.docPublic); expect(fakeNg.doc_create).toHaveBeenCalledTimes(0);
expect(retried.docProtected).toBe(original.docProtected);
expect(laggedNg.doc_create).toHaveBeenCalledTimes(0); // no new provisioning
}); });
}); });