Refactor mouse-mode leak E2E test to use shell printf and live pane

Eliminate the external Node.js fixture file and streamline the E2E test
by using a POSIX printf shell builtin to arm the mouse tracking modes.

Additionally, verify the leak precondition by inspecting the active pane's
live terminal state (mouseTrackingMode) rather than querying internal
daemon buffer snapshots via window.api.pty.getMainBufferSnapshot.

tests/e2e/fixtures/leaked-mouse-mode-fixture.cjs

@@ -1,10 +0,0 @@
-// Simulates a TUI that armed mouse tracking and then died uncleanly: it emits
-// the enable sequence to stdout and exits WITHOUT the matching disable, so the
-// daemon's private-mode tracker keeps the mode and buildRehydrateSequences
-// re-arms it on every reattach. The reattach reset must clear it, or a plain
-// shell echoes pointer-motion reports as literal text. Regression fixture for
-// the mouse-mode leak fixed alongside #7329.
-const ESC = '\x1b'
-
-// ?1003h any-motion tracking + ?1006h SGR encoding, with no ?1003l/?1006l.
-process.stdout.write(`${ESC}[?1003h${ESC}[?1006h`)

tests/e2e/terminal-reattach-mouse-mode-leak.spec.ts

@@ -10,10 +10,9 @@
  *   every pointer-motion report (`<35;col;rowM`) as literal text.
  *
  * What it covers (full stack, no mocks):
- *   - First launch arms the leak through the real daemon: a fixture writes the
- *     enable sequence to stdout and exits, and the daemon's buffer snapshot is
- *     asserted to re-arm ?1003h/?1006h on reattach (the leak precondition — the
- *     assertion that pins "this would leak without the fix").
+ *   - First launch arms the leak for real: a fixture writes the enable sequence
+ *     to stdout and exits, and the live pane is asserted to enter mouse-reporting
+ *     mode (the byte flow the daemon's tracker also records and re-arms).
  *   - After a warm reattach, the renderer terminal ends with mouse reporting
  *     DISARMED: mouseTrackingMode is 'none', the enable-mouse-events class is
  *     gone, and real pointer motion produces zero mouse reports.
@@ -21,11 +20,13 @@
  *     does observe reports, so the "zero reports" assertion cannot pass vacuously.
  *
  * What it does NOT cover:
+ *   - That the daemon's buildRehydrateSequences re-arms the mode on reattach —
+ *     locked by repro-7329-remote-snapshot-corruption.test.ts against the real
+ *     serializer + xterm. This suite proves the reset half end to end.
  *   - Live-agent panes keeping mouse via POST_REPLAY_LIVE_AGENT_REATTACH_RESET
  *     (unit-tested in pty-connection.test.ts).
  */
 
-import path from 'node:path'
 import { existsSync, readFileSync } from 'node:fs'
 import type { ElectronApplication } from '@stablyai/playwright-test'
 import { test, expect } from './helpers/orca-app'
@@ -34,16 +35,12 @@ import {
   discoverActivePtyId,
   execInTerminal,
   waitForActiveTerminalManager,
-  waitForPaneCount
+  waitForPaneCount,
+  waitForTerminalOutput
 } from './helpers/terminal'
 import { ensureTerminalVisible, waitForActiveWorktree, waitForSessionReady } from './helpers/store'
 import { attachRepoAndOpenTerminal, createRestartSession } from './helpers/orca-restart'
 
-const LEAKED_MOUSE_MODE_FIXTURE_PATH = path.join(
-  process.cwd(),
-  'tests/e2e/fixtures/leaked-mouse-mode-fixture.cjs'
-)
-
 // Why: quit→relaunch against the same userDataDir relies on the daemon
 // surviving the first close; serial keeps the shared profile from competing
 // with other Electron instances for the same lock.
@@ -52,6 +49,10 @@ test.describe.configure({ mode: 'serial' })
 test.describe('reattach mouse-mode leak', () => {
   test('warm reattach disarms mouse modes a killed TUI left armed', async (// oxlint-disable-next-line no-empty-pattern -- Playwright's second fixture arg is testInfo; the first must be an object destructure to opt out of the default fixture set.
   {}, testInfo) => {
+    // Why: arms the leak with a POSIX `printf` builtin; the reset logic under
+    // test is platform-agnostic, so skipping Windows here loses no coverage.
+    test.skip(process.platform === 'win32', 'Uses a POSIX printf to arm mouse reporting')
+
     const repoPath = readFileSync(TEST_REPO_PATH_FILE, 'utf-8').trim()
     if (!repoPath || !existsSync(repoPath)) {
       test.skip(true, 'Global setup did not produce a seeded test repo')
@@ -81,32 +82,47 @@ test.describe('reattach mouse-mode leak', () => {
       await waitForPaneCount(firstLaunch.page, 1, 30_000)
       const ptyId = await discoverActivePtyId(firstLaunch.page)
 
-      // The fixture emits ?1003h?1006h to stdout and exits without the disable,
-      // so the daemon's tracker keeps the mode while a plain shell foregrounds.
-      await execInTerminal(
-        firstLaunch.page,
-        ptyId,
-        `node ${JSON.stringify(LEAKED_MOUSE_MODE_FIXTURE_PATH)}`
-      )
+      // Gate on real command execution before arming: the `$((21+21))` result
+      // only appears if the shell evaluated it (the echoed command text shows
+      // the expression, not `42`), so this rules out a shell that hasn't started.
+      await execInTerminal(firstLaunch.page, ptyId, 'echo ORCA_MOUSE_READY_$((21+21))')
+      await waitForTerminalOutput(firstLaunch.page, 'ORCA_MOUSE_READY_42', 15_000)
 
-      // Precondition: the daemon buffer snapshot re-arms both modes on reattach
-      // (data = rehydrateSequences + snapshotAnsi). Without the fix this is
-      // exactly what leaks into the reattached plain shell.
+      // Arm mouse tracking exactly as a TUI would, then let the shell foreground
+      // return — the disable is never sent, so the daemon's tracker keeps the
+      // mode and re-arms it on reattach. printf is a shell builtin (no external
+      // binary), and its typed argument is literal backslashes, so only real
+      // execution emits the ESC bytes that arm xterm below.
+      await execInTerminal(firstLaunch.page, ptyId, `printf '\\033[?1003h\\033[?1006h'`)
+
+      // Precondition: the printf armed real mouse reporting in the live
+      // pane (proving the leak's byte flow reached the terminal). The daemon's
+      // tracker sees the same output stream and re-arms this mode on every
+      // reattach — the rehydrate half is locked by the repro-7329 unit test; this
+      // suite proves the reattach reset disarms it end to end.
       await expect
         .poll(
           async () =>
-            firstLaunch.page.evaluate(async (id: string) => {
-              const snap = await window.api.pty.getMainBufferSnapshot(id)
-              const data = snap?.data ?? ''
-              return data.includes('\x1b[?1003h') && data.includes('\x1b[?1006h')
-            }, ptyId),
+            firstLaunch.page.evaluate(() => {
+              const state = window.__store?.getState()
+              const worktreeId = state?.activeWorktreeId
+              const tabId =
+                state?.activeTabType === 'terminal'
+                  ? state.activeTabId
+                  : worktreeId
+                    ? (state?.activeTabIdByWorktree?.[worktreeId] ?? null)
+                    : null
+              const manager = tabId ? window.__paneManagers?.get(tabId) : null
+              const pane = manager?.getActivePane?.() ?? manager?.getPanes?.()[0] ?? null
+              return pane?.terminal.modes.mouseTrackingMode ?? null
+            }),
           {
             timeout: 15_000,
             message:
-              'Daemon snapshot never re-armed the leaked mouse modes; leak precondition not met'
+              'printf never armed mouse reporting in the live pane; leak precondition not met'
           }
         )
-        .toBe(true)
+        .toBe('any')
 
       // Why: app.close flushes beforeunload, but the daemon is a detached fork
       // so the PTY (and its armed mouse mode) survives for the warm reattach.