[eric] voice: focus re-arms the fn tap so apps like Wispr that register later cannot silently eat the dictation key (ENG-317)

This commit is contained in:
ciregenz
2026-08-16 00:55:31 -07:00
parent 7d9fe3dc94
commit 640c872214
3 changed files with 110 additions and 11 deletions
+41 -9
View File
@@ -5,6 +5,8 @@ import CoreGraphics
import Foundation
var fnDown = false
var tapRef: CFMachPort?
var srcRef: CFRunLoopSource?
let callback: CGEventTapCallBack = { _, type, event, _ in
if type == .flagsChanged {
@@ -24,20 +26,50 @@ let callback: CGEventTapCallBack = { _, type, event, _ in
return Unmanaged.passUnretained(event)
}
var tapRef: CFMachPort?
let mask = (CGEventMask(1) << CGEventType.flagsChanged.rawValue)
guard let tap = CGEvent.tapCreate(
tap: .cgSessionEventTap, place: .headInsertEventTap, options: .listenOnly,
eventsOfInterest: mask, callback: callback, userInfo: nil
) else {
// New tap first, old tap down after: a failed re-arm keeps the working tap instead of going deaf.
// The fnDown de-dupe above makes the brief two-tap overlap print nothing twice.
func armTap() -> Bool {
guard let tap = CGEvent.tapCreate(
tap: .cgSessionEventTap, place: .headInsertEventTap, options: .listenOnly,
eventsOfInterest: mask, callback: callback, userInfo: nil
) else { return false }
let src = CFMachPortCreateRunLoopSource(kCFAllocatorDefault, tap, 0)
CFRunLoopAddSource(CFRunLoopGetMain(), src, .commonModes)
CGEvent.tapEnable(tap: tap, enable: true)
if let old = tapRef { CGEvent.tapEnable(tap: old, enable: false); CFMachPortInvalidate(old) }
if let oldSrc = srcRef { CFRunLoopRemoveSource(CFRunLoopGetMain(), oldSrc, .commonModes) }
tapRef = tap
srcRef = src
return true
}
guard armTap() else {
print("e tap-failed")
fflush(stdout)
exit(1)
}
tapRef = tap
let src = CFMachPortCreateRunLoopSource(kCFAllocatorDefault, tap, 0)
CFRunLoopAddSource(CFRunLoopGetCurrent(), src, .commonModes)
CGEvent.tapEnable(tap: tap, enable: true)
// The parent pokes "r\n" when the app gains focus: another app's tap registered after ours sits
// AHEAD of ours (head-insert) and can eat fn before we see it, with no disable event to catch, so
// re-arming is the only way to win the key back (ENG-317). stdin EOF means the parent is gone;
// exiting then stops a crashed OpenSwarm from stranding a process holding a global keyboard tap.
DispatchQueue.global().async {
while let line = readLine(strippingNewline: true) {
if line == "r" {
CFRunLoopPerformBlock(CFRunLoopGetMain(), CFRunLoopMode.commonModes.rawValue) {
if !armTap() {
print("e rearm-failed")
fflush(stdout)
}
}
CFRunLoopWakeUp(CFRunLoopGetMain())
}
}
exit(0)
}
print("r")
fflush(stdout)
CFRunLoopRun()
+13 -2
View File
@@ -155,10 +155,21 @@ function installVoiceHotkey(getMainWindow) {
}
} catch (_) { /* a machine where ps is restricted must still arm the watcher */ }
};
// Focus re-arms the watcher's tap: a tap another app registered after ours head-inserts AHEAD of
// ours and can eat fn with no disable event delivered, so this is the only recovery (ENG-317).
let lastFnPokeMs = 0;
const pokeFnWatcher = () => {
if (!fnProc || !fnProc.stdin || !fnProc.stdin.writable) return;
const now = Date.now();
if (now - lastFnPokeMs < 1000) return;
lastFnPokeMs = now;
try { fnProc.stdin.write('r\n'); } catch (_) {}
};
const startFnWatcherWith = (bin) => {
sweepStrayFnWatchers(bin);
try {
fnProc = spawn(bin, [], { stdio: ['ignore', 'pipe', 'ignore'] });
// stdin stays open on purpose: "r\n" re-arms the tap, and EOF tells an orphaned watcher to die.
fnProc = spawn(bin, [], { stdio: ['pipe', 'pipe', 'ignore'] });
} catch (e) {
console.log('[voice] fn watcher spawn failed:', e && e.message);
fnProc = null;
@@ -279,7 +290,7 @@ function installVoiceHotkey(getMainWindow) {
tryStartNativeTap();
startFnWatcher();
registerVoiceShortcut();
app.on('browser-window-focus', unregisterFallbackShortcut);
app.on('browser-window-focus', () => { unregisterFallbackShortcut(); pokeFnWatcher(); });
app.on('browser-window-blur', registerVoiceShortcut);
// The focused-window relay matches the FALLBACK chord: special primaries (fn, Ctrl+Win) are
+56
View File
@@ -0,0 +1,56 @@
// Run: node --test electron/voiceHotkeyRearm.test.js
//
// ENG-317: an event tap another app registers AFTER ours head-inserts AHEAD of ours and can eat fn
// with no disable event delivered, so the watcher's existing timeout re-enable never fires and fn
// goes silently dead (reproduced with a real adversary tap: watcher got ZERO bytes). The fix is a
// focus-time "r\n" poke that re-arms the tap, head-inserting us back in front (proven live: starved
// under the adversary, one poke, fn flowed again). These pin the wire and the protocol.
const test = require('node:test');
const assert = require('node:assert/strict');
const fs = require('node:fs');
const path = require('node:path');
const { spawnSync, spawn } = require('node:child_process');
const swiftSrc = fs.readFileSync(path.join(__dirname, 'native', 'fn-watcher.swift'), 'utf8');
const hotkeySrc = fs.readFileSync(path.join(__dirname, 'voiceHotkey.js'), 'utf8');
test('the watcher understands the re-arm poke and dies on parent loss', () => {
assert.match(swiftSrc, /if line == "r"/, 'the stdin re-arm command is the whole ENG-317 fix');
assert.match(swiftSrc, /while let line = readLine/, 'stdin must be read line-wise');
assert.match(swiftSrc, /exit\(0\)/, 'stdin EOF must exit, or a crashed parent strands a global tap');
const armBody = swiftSrc.slice(swiftSrc.indexOf('func armTap'), swiftSrc.indexOf('guard armTap'));
assert.ok(armBody.indexOf('tapCreate') < armBody.indexOf('CFMachPortInvalidate'), 'new tap up BEFORE old tap down, or a failed re-arm goes deaf');
});
test('electron pokes on focus through a piped stdin', () => {
assert.match(hotkeySrc, /stdio: \['pipe', 'pipe', 'ignore'\]/, 'an ignored stdin is /dev/null, whose instant EOF would kill the watcher at birth');
assert.match(hotkeySrc, /fnProc\.stdin\.write\('r\\n'\)/, 'the poke must reach the watcher');
const focusLine = hotkeySrc.split('\n').find((l) => l.includes("app.on('browser-window-focus'"));
assert.ok(focusLine && focusLine.includes('pokeFnWatcher'), 'focus is the moment we can win the tap back');
});
test('live protocol: armed watcher survives pokes and exits on EOF', { skip: process.platform !== 'darwin' }, async () => {
const bin = path.join(require('node:os').tmpdir(), `fn-watcher-rearm-test-${process.pid}`);
const cc = spawnSync('swiftc', ['-O', '-o', bin, path.join(__dirname, 'native', 'fn-watcher.swift')], { timeout: 120000 });
if (cc.error || cc.status !== 0) return; // no toolchain on this runner; source checks above still hold
const p = spawn(bin, [], { stdio: ['pipe', 'pipe', 'ignore'] });
let out = ''; p.stdout.on('data', (c) => { out += String(c); });
let exitCode = null; p.on('exit', (c) => { exitCode = c; });
const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
for (let waited = 0; !out.includes('r') && exitCode === null && waited < 4000; waited += 100) await sleep(100);
try {
// No boot marker = no grant, or a Gatekeeper-wedged machine hanging fresh binaries at
// _dyld_start (both seen live); asserting against a process that never ran proves nothing.
if (exitCode !== null || !out.includes('r')) return;
p.stdin.write('r\n'); p.stdin.write('r\n');
await sleep(500);
assert.equal(exitCode, null, 'pokes must not kill the watcher');
assert.ok(!out.split('\n').some((l) => l.startsWith('e')), `re-arm errored: ${out}`);
p.stdin.end();
for (let waited = 0; exitCode === null && waited < 5000; waited += 100) await sleep(100);
assert.equal(exitCode, 0, 'EOF must exit cleanly; a survivor here is the orphan-tap leak');
} finally {
try { p.kill('SIGKILL'); } catch (_) {}
try { fs.unlinkSync(bin); } catch (_) {}
}
});