test(stalker): bound auth e2e worker slots

apps/stalker-mock-server/README.md

@@ -86,7 +86,7 @@ actually get wrong:
 | `00:1A:79:00:00:06` | **legacy-pagination** | No `get_all_channels` support — tests the paginated `get_ordered_list` crawl fallback for the full ITV channel list |
 | `00:1A:79:00:00:07` | **marketing-demo** | 35 original poster movies with the newest 20 first — safe for screenshots and marketing |
 | `00:1A:79:00:00:08` | **login-required** | `get_profile` answers `status: 2` until the client completes `do_auth` with non-empty credentials. The app drives this end to end: enter a username and password in the Stalker import dialog and it runs `do_auth`, then retries `get_profile` with `auth_second_step=1`. Covered by `stalker.e2e.ts` (`@stalker full portal authentication`) |
-| `00:1A:79:AE:<worker>:02` | **login-required** | Worker-scoped alias range used by parallel auth E2E runs; it has the same login contract as `00:1A:79:00:00:08` but independent per-MAC state |
+| `00:1A:79:AE:<slot>:02` | **login-required** | Parallel-slot-scoped alias range used by concurrent auth E2E runs; it has the same login contract as `00:1A:79:00:00:08` but independent per-MAC state |
 | `00:1A:79:00:00:09` | **gated-stream** | `create_link` returns a local `/stream/gated/…` URL that answers 403 without the mac cookie and the MAC's current Bearer token — proves a player's media requests really carry the portal credentials |
 | `00:1A:79:00:00:0A` | **static-channel-cmd** | ITV rows carry a directly playable `cmd` with `use_http_tmp_link`/`use_load_balancing` both `'0'` — a client honouring the flags must play them without calling `create_link` |
 | `<any other MAC>` | **auto** | MAC bytes used as seed → deterministic unique dataset |
@@ -180,9 +180,10 @@ everything. The sibling specs that talk to this server (`self-hosted.e2e.ts`,
 the `sources-pwa` helpers) own a disjoint `00:1A:79:5F:*` range for the same
 reason, and `stalker.e2e.ts` runs `mode: 'serial'` because its tests
 deliberately share scenario MACs. Its stateful authentication tests derive a
-separate `00:1A:79:AE:<worker>:*` range from Playwright's worker index and add
-only that range to the scoped reset, so browser projects and repeat workers do
-not clear one another's sessions or pinned device identity. Full contract:
+separate `00:1A:79:AE:<slot>:*` range from Playwright's bounded
+`parallelIndex` and add only that range to the scoped reset, so concurrent
+browser projects do not clear one another's sessions or pinned device
+identity; restarted workers retain their slot. Full contract:
 [`docs/architecture/stalker-mock-server.md`](../../docs/architecture/stalker-mock-server.md#test-isolation).
 
 ## EPG Behavior

apps/stalker-mock-server/src/app/handlers/auth-handlers.spec.ts

@@ -126,7 +126,7 @@ describe('stalker mock authentication handlers', () => {
         ).toBe(null);
     });
 
-    it('applies the login-required scenario to worker-scoped MACs', () => {
+    it('applies the login-required scenario to parallel-slot MACs', () => {
         const token = invoke(handleHandshake, { action: 'handshake' }, {
             mac: WORKER_LOGIN_REQUIRED_MAC,
         })['token'] as string;

apps/web-e2e/src/stalker.e2e.ts

@@ -40,9 +40,10 @@ import {
  * test; serializing one file is the cheaper trade.
  *
  * - ACROSS BROWSER PROJECTS/WORKERS: state-sensitive auth tests use one MAC
- *   range per Playwright worker. Their `beforeEach` adds only that worker's
- *   MACs to the scoped reset, so concurrent projects or repeat workers cannot
- *   clear one another's token, invalidated session or pinned device id.
+ *   range per Playwright parallel slot. Their `beforeEach` adds only that
+ *   slot's MACs to the scoped reset, so concurrent projects cannot clear one
+ *   another's token, invalidated session or pinned device id. A restarted
+ *   worker retains its bounded `parallelIndex`.
  */
 
 test.describe.configure({ mode: 'serial' });
@@ -80,8 +81,8 @@ const STATIC_CMD_MAC = '00:1A:79:00:00:0A';
 /**
  * These tests assert state transitions within one portal session, so a reset
  * from a concurrent browser project or repeat worker would invalidate the
- * assertion itself. Giving every worker its own MAC range preserves browser
- * parallelism and also keeps `--repeat-each` runs isolated.
+ * assertion itself. Giving every concurrent worker slot its own MAC range
+ * preserves browser parallelism and also keeps `--repeat-each` runs isolated.
  */
 interface StatefulAuthMacs {
     authenticatedFlow: string;
@@ -91,16 +92,25 @@ interface StatefulAuthMacs {
     reauthentication: string;
 }
 
-function getStatefulAuthMacs(workerIndex: number): StatefulAuthMacs {
+function getStatefulAuthMacs({
+    parallelIndex,
+}: {
+    parallelIndex: number;
+}): StatefulAuthMacs {
     if (
-        !Number.isSafeInteger(workerIndex) ||
-        workerIndex < 0 ||
-        workerIndex > 255
+        !Number.isSafeInteger(parallelIndex) ||
+        parallelIndex < 0 ||
+        parallelIndex > 255
     ) {
-        throw new Error(`Unsupported Playwright worker index: ${workerIndex}`);
+        throw new Error(
+            `Unsupported Playwright parallel index: ${parallelIndex}`
+        );
     }
 
-    const workerOctet = workerIndex.toString(16).padStart(2, '0').toUpperCase();
+    const workerOctet = parallelIndex
+        .toString(16)
+        .padStart(2, '0')
+        .toUpperCase();
     const workerPrefix = `00:1A:79:AE:${workerOctet}`;
 
     return {
@@ -311,7 +321,7 @@ function recordPortalRequests(
 
 test.beforeEach(async ({ page, request }, testInfo) => {
     // Reset mock server state (clears in-memory favorites and cache)
-    await resetMockServer(request, getStatefulAuthMacs(testInfo.workerIndex));
+    await resetMockServer(request, getStatefulAuthMacs(testInfo));
 
     // Playwright creates a fresh browser context per test, so extra
     // IndexedDB cleanup here only risks racing with app-managed DB handles.
@@ -332,6 +342,14 @@ test('@stalker health check — mock server is running', async ({ request }) =>
     expect(body.status).toBe('ok');
 });
 
+test('@stalker auth MAC allocation survives worker process restarts', () => {
+    const restartedWorker = { parallelIndex: 7, workerIndex: 256 };
+
+    expect(getStatefulAuthMacs(restartedWorker).loginRequired).toBe(
+        '00:1A:79:AE:07:02'
+    );
+});
+
 test('@stalker add a Stalker portal and see it in the playlist list', async ({
     page,
 }) => {
@@ -980,7 +998,7 @@ test.describe('@stalker full portal authentication', () => {
     }, testInfo) => {
         const requests = recordPortalRequests(page);
         const actionsInOrder = () => requests.map((entry) => entry.action);
-        const mac = getStatefulAuthMacs(testInfo.workerIndex).authenticatedFlow;
+        const mac = getStatefulAuthMacs(testInfo).authenticatedFlow;
 
         await addFullStalkerPortal(page, { mac });
 
@@ -1038,7 +1056,7 @@ test.describe('@stalker full portal authentication', () => {
         page,
     }, testInfo) => {
         const requests = recordPortalRequests(page);
-        const mac = getStatefulAuthMacs(testInfo.workerIndex).loginRequired;
+        const mac = getStatefulAuthMacs(testInfo).loginRequired;
         // The second auth step must only be claimed on the retry AFTER
         // do_auth — a client that always sends auth_second_step=1 works
         // against the mock but violates the documented protocol.
@@ -1093,7 +1111,7 @@ test.describe('@stalker full portal authentication', () => {
     test('explains a login-required portal and recovers once credentials are entered', async ({
         page,
     }, testInfo) => {
-        const mac = getStatefulAuthMacs(testInfo.workerIndex).loginRequired;
+        const mac = getStatefulAuthMacs(testInfo).loginRequired;
         // First attempt without credentials: the import must fail with the
         // portal's actual reason, not a generic "check URL and MAC" error.
         await page.getByRole('button', { name: 'Add playlist' }).click();
@@ -1135,7 +1153,7 @@ test.describe('@stalker full portal authentication', () => {
         page,
     }, testInfo) => {
         const requests = recordPortalRequests(page);
-        const mac = getStatefulAuthMacs(testInfo.workerIndex).tokenReuse;
+        const mac = getStatefulAuthMacs(testInfo).tokenReuse;
 
         await addFullStalkerPortal(page, { mac });
 
@@ -1258,7 +1276,7 @@ test.describe('@stalker full portal authentication', () => {
         page,
         request,
     }, testInfo) => {
-        const mac = getStatefulAuthMacs(testInfo.workerIndex).deviceConflict;
+        const mac = getStatefulAuthMacs(testInfo).deviceConflict;
         // Pin a device id the way another client (StbEmu, a set-top box)
         // would have: the stock server binds the first non-empty `device_id`
         // it sees to the MAC and refuses every different one afterwards.
@@ -1315,7 +1333,7 @@ test.describe('@stalker full portal authentication', () => {
         request,
     }, testInfo) => {
         const requests = recordPortalRequests(page);
-        const mac = getStatefulAuthMacs(testInfo.workerIndex).reauthentication;
+        const mac = getStatefulAuthMacs(testInfo).reauthentication;
 
         await addFullStalkerPortal(page, { mac });
 

docs/architecture/stalker-mock-server.md

@@ -319,9 +319,9 @@ supporting scenarios, `get_all_channels` (`get-all-channels.handler.ts`,
 (adult) genres.
 
 The login-required scenario also recognizes
-`00:1A:79:AE:<worker>:02`. This test-only alias range preserves the same
-`status: 2`/`do_auth` contract while giving concurrent Playwright workers
-independent per-MAC auth state.
+`00:1A:79:AE:<slot>:02`. This test-only alias range preserves the same
+`status: 2`/`do_auth` contract while giving concurrent Playwright parallel
+slots independent per-MAC auth state.
 
 The `marketing-demo` scenario (`00:1A:79:00:00:07`) replaces faker-generated
 VOD with the 35-movie provider-neutral showcase catalog from
@@ -401,10 +401,12 @@ file running in parallel workers, so isolation is per-MAC rather than global:
   (their fixture shapes are what the assertions are written against), so the
   file uses `test.describe.configure({ mode: 'serial' })`.
 - Authentication tests whose assertions span multiple requests derive a
-  disjoint `00:1A:79:AE:<worker>:*` range from Playwright's `workerIndex`.
-  Their `beforeEach` adds only the current worker's range to the batched reset,
-  so browser projects and `--repeat-each` workers cannot clear one another's
-  token, login completion, invalidated session or pinned device identity.
+  disjoint `00:1A:79:AE:<slot>:*` range from Playwright's bounded
+  `parallelIndex`. Their `beforeEach` adds only the current parallel slot's
+  range to the batched reset, so browser projects cannot clear one another's
+  token, login completion, invalidated session or pinned device identity; a
+  restarted worker retains the same slot instead of consuming a wider MAC
+  value.
 
 A reset drops the generated content cache, favorites (`data-store.ts`), the
 auth/session record (`auth-store.ts`, including any pinned device identity) and