Scope Esc cancel listener to active recording

src/aman.py

@@ -77,6 +77,18 @@ class Daemon:
         self.vocabulary = VocabularyEngine(cfg.vocabulary)
         self._stt_hint_kwargs_cache: dict[str, Any] | None = None
 
+    def _arm_cancel_listener(self):
+        try:
+            self.desktop.start_cancel_listener(lambda: self.cancel_recording())
+        except Exception as exc:
+            logging.error("failed to start cancel listener: %s", exc)
+
+    def _disarm_cancel_listener(self):
+        try:
+            self.desktop.stop_cancel_listener()
+        except Exception as exc:
+            logging.debug("failed to stop cancel listener: %s", exc)
+
     def set_state(self, state: str):
         with self.lock:
             prev = self.state
@@ -84,7 +96,7 @@ class Daemon:
         if prev != state:
             logging.debug("state: %s -> %s", prev, state)
         else:
-            logging.warning("redundant state set: %s, kindly inform the dev", state)
+            logging.debug("redundant state set: %s", state)
 
     def get_state(self):
         with self.lock:
@@ -123,6 +135,7 @@ class Daemon:
         prev = self.state
         self.state = State.RECORDING
         logging.debug("state: %s -> %s", prev, self.state)
+        self._arm_cancel_listener()
         logging.info("recording started")
         if self.timer:
             self.timer.cancel()
@@ -150,6 +163,7 @@ class Daemon:
         if self.timer:
             self.timer.cancel()
         self.timer = None
+        self._disarm_cancel_listener()
         prev = self.state
         self.state = State.STT
         logging.debug("state: %s -> %s", prev, self.state)
@@ -179,7 +193,6 @@ class Daemon:
             return
 
         try:
-            self.set_state(State.STT)
             logging.info("stt started")
             text = self._transcribe(audio)
         except Exception as exc:
@@ -256,6 +269,7 @@ class Daemon:
 
     def shutdown(self, timeout: float = 5.0) -> bool:
         self.request_shutdown()
+        self._disarm_cancel_listener()
         self.stop_recording(trigger="shutdown", process_audio=False)
         return self.wait_for_idle(timeout)
 
@@ -402,7 +416,6 @@ def main():
             cfg.daemon.hotkey,
             lambda: logging.info("hotkey pressed (dry-run)") if args.dry_run else daemon.toggle(),
         )
-        desktop.start_cancel_listener(lambda: daemon.cancel_recording())
     except Exception as exc:
         logging.error("hotkey setup failed: %s", exc)
         raise SystemExit(1)

src/desktop.py

@@ -11,6 +11,9 @@ class DesktopAdapter(Protocol):
     def start_cancel_listener(self, callback: Callable[[], None]) -> None:
         raise NotImplementedError
 
+    def stop_cancel_listener(self) -> None:
+        raise NotImplementedError
+
     def inject_text(
         self,
         text: str,

src/desktop_wayland.py

@@ -10,6 +10,9 @@ class WaylandAdapter:
     def start_cancel_listener(self, _callback: Callable[[], None]) -> None:
         raise SystemExit("Wayland hotkeys are not supported yet.")
 
+    def stop_cancel_listener(self) -> None:
+        raise SystemExit("Wayland hotkeys are not supported yet.")
+
     def inject_text(
         self,
         _text: str,

src/desktop_x11.py

@@ -42,6 +42,9 @@ class X11Adapter:
         self.indicator = None
         self.status_icon = None
         self.menu = None
+        self._cancel_listener_lock = threading.Lock()
+        self._cancel_listener_stop_event: threading.Event | None = None
+        self._cancel_listener_callback: Callable[[], None] | None = None
         if AppIndicator3 is not None:
             self.indicator = AppIndicator3.Indicator.new(
                 "aman",
@@ -72,9 +75,35 @@ class X11Adapter:
 
     def start_cancel_listener(self, callback: Callable[[], None]) -> None:
         mods, keysym = self._parse_hotkey("Escape")
-        thread = threading.Thread(target=self._listen, args=(mods, keysym, callback), daemon=True)
+        with self._cancel_listener_lock:
+            if self._cancel_listener_stop_event is not None:
+                self._cancel_listener_callback = callback
+                return
+            stop_event = threading.Event()
+            self._cancel_listener_stop_event = stop_event
+            self._cancel_listener_callback = callback
+
+        thread = threading.Thread(
+            target=self._listen,
+            args=(mods, keysym, self._dispatch_cancel_listener, stop_event),
+            daemon=True,
+        )
         thread.start()
 
+    def stop_cancel_listener(self) -> None:
+        with self._cancel_listener_lock:
+            stop_event = self._cancel_listener_stop_event
+            self._cancel_listener_stop_event = None
+            self._cancel_listener_callback = None
+        if stop_event is not None:
+            stop_event.set()
+
+    def _dispatch_cancel_listener(self) -> None:
+        with self._cancel_listener_lock:
+            callback = self._cancel_listener_callback
+        if callback is not None:
+            callback()
+
     def inject_text(
         self,
         text: str,
@@ -126,7 +155,14 @@ class X11Adapter:
         finally:
             self.request_quit()
 
-    def _listen(self, mods: int, keysym: int, callback: Callable[[], None]) -> None:
+    def _listen(
+        self,
+        mods: int,
+        keysym: int,
+        callback: Callable[[], None],
+        stop_event: threading.Event | None = None,
+    ) -> None:
+        local_stop = stop_event or threading.Event()
         disp = None
         root = None
         keycode = None
@@ -134,14 +170,18 @@ class X11Adapter:
             disp = display.Display()
             root = disp.screen().root
             keycode = self._grab_hotkey(disp, root, mods, keysym)
-            while True:
+            while not local_stop.is_set():
+                if disp.pending_events() == 0:
+                    time.sleep(0.05)
+                    continue
                 ev = disp.next_event()
                 if ev.type == X.KeyPress and ev.detail == keycode:
                     state = ev.state & ~(X.LockMask | X.Mod2Mask)
                     if state == mods:
                         callback()
         except Exception as exc:
-            logging.error("hotkey listener stopped: %s", exc)
+            if not local_stop.is_set():
+                logging.error("hotkey listener stopped: %s", exc)
         finally:
             if root is not None and keycode is not None and disp is not None:
                 try:
@@ -149,6 +189,11 @@ class X11Adapter:
                     disp.sync()
                 except Exception:
                     pass
+            if disp is not None:
+                try:
+                    disp.close()
+                except Exception:
+                    pass
 
     def _parse_hotkey(self, hotkey: str):
         mods = 0