test: cover native X11 Hangul-plus-digit PTY bytes in CI (#19013)

* test: run the native Hangul terminating-digit regression in CI

* test: distinguish X11 byte coverage from the manual Wayland repro

* test: require native IME engagement proof for the digit case

config/scripts/pr-e2e-gate-contract.test.mjs

@@ -636,13 +636,8 @@ describe('PR E2E gate contract', () => {
       .filter((spec) => nativeGateExpression.test(readFileSync(join(projectDir, spec), 'utf8')))
     expect(nativeGatedSpecs.length).toBeGreaterThan(0)
 
-    // Why exempt: the digit repro needs a nested gnome-shell, which no hosted runner provides
-    // (headless mutter never answers RemoteDesktop.CreateSession); the macOS spec needs a real
-    // macOS input source, and no macOS runner exists on any PR or scheduled lane.
-    const unreachableSpecs = new Set([
-      'tests/e2e/terminal-hangul-terminating-digit-native.spec.ts',
-      'tests/e2e/terminal-macos-2set-korean-native.spec.ts'
-    ])
+    // The macOS spec needs a native input source; PR and scheduled IME lanes use Linux.
+    const unreachableSpecs = new Set(['tests/e2e/terminal-macos-2set-korean-native.spec.ts'])
     const unclaimed = nativeGatedSpecs.filter(
       (spec) => !unreachableSpecs.has(spec) && !nativeImeRunner.includes(spec)
     )
@@ -682,8 +677,13 @@ describe('PR E2E gate contract', () => {
 
     // Why pin the titles: the runner requires one receipt per name, so a rename that nobody
     // mirrored here would fail the lane loudly instead of quietly halving it.
+    const nativeDigitSpec = readFileSync(
+      join(projectDir, 'tests/e2e/terminal-hangul-terminating-digit-native.spec.ts'),
+      'utf8'
+    )
+    expect(nativeDigitSpec).toContain('appendImeEngagementReceipt(testInfo.title, trace)')
     for (const title of EXPECTED_NATIVE_IME_TESTS) {
-      expect(nativeImeSpec, title).toContain(title)
+      expect(nativeImeSpec + nativeDigitSpec, title).toContain(title)
     }
   })
 

config/scripts/run-terminal-ibus-hangul-e2e.mjs

@@ -199,7 +199,8 @@ async function runInsideSession(evidenceDir) {
         'test:e2e:headful',
         '--workers=1',
         '--',
-        'tests/e2e/terminal-ibus-hangul-native.spec.ts'
+        'tests/e2e/terminal-ibus-hangul-native.spec.ts',
+        'tests/e2e/terminal-hangul-terminating-digit-native.spec.ts'
       ],
       {
         cwd: projectDir,

config/scripts/terminal-ime-engagement-receipt.mjs

@@ -13,7 +13,8 @@ export const IME_ENGAGEMENT_RECEIPT_ENV = 'ORCA_E2E_IME_ENGAGEMENT_RECEIPT'
 /** The tests that must each leave a receipt. Pinned so deleting one cannot quietly shrink the lane. */
 export const EXPECTED_NATIVE_IME_TESTS = [
   'forwards the issue exact-byte sequence without loss or duplication',
-  'forwards the issue sentence stress sequence without leaked ASCII'
+  'forwards the issue sentence stress sequence without leaked ASCII',
+  'a digit typed right after a Hangul syllable reaches the pty'
 ]
 
 function parseReceipts(text) {

config/scripts/terminal-ime-engagement-receipt.test.mjs

@@ -4,7 +4,7 @@ import {
   verifyImeEngagementReceipts
 } from './terminal-ime-engagement-receipt.mjs'
 
-const [firstTest, secondTest] = EXPECTED_NATIVE_IME_TESTS
+const [firstTest, secondTest, thirdTest] = EXPECTED_NATIVE_IME_TESTS
 
 function receipt(test, overrides = {}) {
   return JSON.stringify({
@@ -18,9 +18,11 @@ function receipt(test, overrides = {}) {
 
 describe('verifyImeEngagementReceipts', () => {
   it('accepts a run where every expected test observed real composition', () => {
-    expect(verifyImeEngagementReceipts(`${receipt(firstTest)}\n${receipt(secondTest)}\n`)).toEqual(
-      []
-    )
+    expect(
+      verifyImeEngagementReceipts(
+        `${receipt(firstTest)}\n${receipt(secondTest)}\n${receipt(thirdTest)}\n`
+      )
+    ).toEqual([])
   })
 
   // The failure this whole mechanism exists for: Playwright reports a skipped test as a pass, so
@@ -35,13 +37,20 @@ describe('verifyImeEngagementReceipts', () => {
 
   it('rejects a partial run where only one test reached the engine', () => {
     expect(verifyImeEngagementReceipts(`${receipt(firstTest)}\n`)).toEqual([
-      `no engagement receipt for "${secondTest}" — it was skipped, filtered out, or renamed`
+      `no engagement receipt for "${secondTest}" — it was skipped, filtered out, or renamed`,
+      `no engagement receipt for "${thirdTest}" — it was skipped, filtered out, or renamed`
+    ])
+  })
+
+  it('requires the digit receipt even when both original native tests passed', () => {
+    expect(verifyImeEngagementReceipts(`${receipt(firstTest)}\n${receipt(secondTest)}\n`)).toEqual([
+      `no engagement receipt for "${thirdTest}" — it was skipped, filtered out, or renamed`
     ])
   })
 
   it('rejects a run that typed keys but never opened a composition', () => {
     const problems = verifyImeEngagementReceipts(
-      `${receipt(firstTest, { compositionStart: 0 })}\n${receipt(secondTest)}\n`
+      `${receipt(firstTest, { compositionStart: 0 })}\n${receipt(secondTest)}\n${receipt(thirdTest)}\n`
     )
     expect(problems).toEqual([
       `"${firstTest}" recorded no compositionstart — the IME never engaged`
@@ -50,7 +59,7 @@ describe('verifyImeEngagementReceipts', () => {
 
   it('rejects a composition that produced no Hangul, which a latin passthrough would satisfy', () => {
     const problems = verifyImeEngagementReceipts(
-      `${receipt(firstTest, { hangulComposition: 0 })}\n${receipt(secondTest)}\n`
+      `${receipt(firstTest, { hangulComposition: 0 })}\n${receipt(secondTest)}\n${receipt(thirdTest)}\n`
     )
     expect(problems).toEqual([
       `"${firstTest}" recorded no Hangul composition data — the engine produced no syllables`
@@ -59,7 +68,7 @@ describe('verifyImeEngagementReceipts', () => {
 
   it('rejects a renamed test rather than counting it toward coverage', () => {
     const problems = verifyImeEngagementReceipts(
-      `${receipt(firstTest)}\n${receipt(secondTest)}\n${receipt('some new scenario')}\n`
+      `${receipt(firstTest)}\n${receipt(secondTest)}\n${receipt(thirdTest)}\n${receipt('some new scenario')}\n`
     )
     expect(problems).toEqual([
       'unexpected engagement receipt for "some new scenario" — update EXPECTED_NATIVE_IME_TESTS'
@@ -68,7 +77,7 @@ describe('verifyImeEngagementReceipts', () => {
 
   it('reports a truncated receipt rather than parsing around it', () => {
     const problems = verifyImeEngagementReceipts(
-      `${receipt(firstTest)}\n{"test":"trunc\n${receipt(secondTest)}\n`
+      `${receipt(firstTest)}\n{"test":"trunc\n${receipt(secondTest)}\n${receipt(thirdTest)}\n`
     )
     expect(problems).toEqual(['malformed receipt line: {"test":"trunc'])
   })

tests/e2e/terminal-hangul-terminating-digit-native.spec.ts

@@ -3,20 +3,18 @@
  * the pty. Written to reproduce #15299, where a digit typed straight after a Hangul syllable was
  * dropped under Wayland but not under X11.
  *
- * THIS DOES NOT RUN IN CI. It is gated on ORCA_E2E_NATIVE_IBUS_HANGUL=1 and needs a compositor
- * session that CI does not have, so it is a manual reproduction harness rather than coverage.
- * That is stated plainly because this repo already carries native IME specs that are skipped
- * everywhere and were mistaken for coverage they never provided.
+ * CI runs the default xdotool injector under X11, checking exact Hangul-plus-digit PTY bytes.
+ * That path passed even before the Wayland fix; it does not prove #15299 is fixed.
+ * Reproducing #15299 still requires the nested Wayland session below.
  *
- * To run it, on a machine with gnome-shell and ibus-hangul:
+ * To run the Wayland reproduction on a machine with gnome-shell and ibus-hangul:
  *
  *   Xvfb :65 -extension GLX &
  *   DISPLAY=:65 gnome-shell --nested --wayland     # nested, NOT --headless
  *   ORCA_E2E_NATIVE_IBUS_HANGUL=1 ORCA_E2E_IME_INJECTOR=nested npx playwright test \
  *     tests/e2e/terminal-hangul-terminating-digit-native.spec.ts
  *
- * Eight things that decide whether a run is real or a silent false negative, each of which cost a
- * failed attempt:
+ * Nested Wayland prerequisites:
  *
  *  - Nested, not headless. A headless mutter never answers RemoteDesktop.CreateSession, so there
  *    is no way to inject input; nested makes the whole compositor an X window that xdotool can
@@ -45,6 +43,7 @@ import { mkdirSync, writeFileSync } from 'node:fs'
 import path from 'node:path'
 import type { Page, TestInfo } from '@stablyai/playwright-test'
 import { test, expect } from './helpers/orca-app'
+import { appendImeEngagementReceipt } from './terminal-ime-engagement-receipt'
 import { ensureTerminalVisible, waitForActiveWorktree, waitForSessionReady } from './helpers/store'
 import {
   focusActiveTerminalInput,
@@ -232,6 +231,11 @@ test.describe('Hangul terminating digit @headful', () => {
       }
 
       receivedBytes = await waitForTerminalImeBytes(page, reader, 20_000)
+      expect(receivedBytes.map((hex) => Buffer.from(hex, 'hex').toString('utf8'))).toEqual(
+        Array.from({ length: REPETITIONS }, () => `${EXPECTED_LINE}\n`)
+      )
+      const trace = await readTerminalImeBoundaryTrace(page)
+      appendImeEngagementReceipt(testInfo.title, trace)
     } finally {
       await writeEvidence(page, testInfo, 'hangul-terminating-digit', {
         expectedHex,
@@ -243,8 +247,5 @@ test.describe('Hangul terminating digit @headful', () => {
       await sendToTerminal(page, ptyId, '\x03').catch(() => undefined)
       removeTerminalImeByteReader(reader)
     }
-    expect(receivedBytes.map((hex) => Buffer.from(hex, 'hex').toString('utf8'))).toEqual(
-      Array.from({ length: REPETITIONS }, () => `${EXPECTED_LINE}\n`)
-    )
   })
 })