Fix 5 second freeze when reading secrets via the XDG portal

buzz/pip_utils.py

@@ -33,11 +33,6 @@ def is_flatpak() -> bool:
     return "FLATPAK_ID" in os.environ
 
 
-def _in_virtualenv() -> bool:
-    """Returns True if running inside a virtualenv or uv venv."""
-    return sys.prefix != sys.base_prefix or "VIRTUAL_ENV" in os.environ
-
-
 def subprocess_hide_window_kwargs() -> dict[str, Any]:
     """Return kwargs to hide the console window on Windows."""
     if sys.platform == "win32":

buzz/settings/shortcut.py

@@ -1,5 +1,4 @@
 import enum
-import typing
 from buzz.locale import _
 
 
@@ -37,7 +36,3 @@ def __new__(cls, sequence: str, description: str):
 
     CLEAR_HISTORY = ("Ctrl+S", _("Clear History"))
     STOP_TRANSCRIPTION = ("Ctrl+X", _("Cancel Transcription"))
-
-    @staticmethod
-    def get_default_shortcuts() -> typing.Dict[str, str]:
-        return {shortcut.name: shortcut.sequence for shortcut in Shortcut}

buzz/store/keyring_store.py

@@ -28,20 +28,42 @@ def _get_secrets_file_path() -> str:
     return os.path.join(data_dir, ".secrets.json")
 
 
+# The portal secret is fixed for the lifetime of the application, so it is
+# fetched once per process. Each lookup is a D-Bus round trip that can block
+# the UI thread, and get_password() is called from many widgets.
+_PORTAL_SECRET_NOT_FETCHED = object()
+_portal_secret_cache = _PORTAL_SECRET_NOT_FETCHED
+
+# How long to wait for the portal to answer the RetrieveSecret request
+PORTAL_RESPONSE_TIMEOUT_SECONDS = 5.0
+
+
 def _get_portal_secret() -> bytes | None:
     """Get the application secret from XDG Desktop Portal.
 
     The portal provides a per-application secret that can be used
     for encrypting application-specific data. This works in sandboxed
     environments (Snap/Flatpak) via the desktop plug.
     """
+    global _portal_secret_cache
+
     if not _is_linux():
         return None
 
+    if _portal_secret_cache is _PORTAL_SECRET_NOT_FETCHED:
+        _portal_secret_cache = _retrieve_portal_secret()
+
+    return _portal_secret_cache
+
+
+def _retrieve_portal_secret() -> bytes | None:
     try:
-        from jeepney import DBusAddress, new_method_call
-        from jeepney.io.blocking import open_dbus_connection
+        from jeepney import (
+            DBusAddress, HeaderFields, MatchRule, message_bus, new_method_call
+        )
+        from jeepney.io.blocking import Proxy, open_dbus_connection
         import socket
+        import time
 
         # Open connection with file descriptor support enabled
         conn = open_dbus_connection(bus="SESSION", enable_fds=True)
@@ -52,25 +74,57 @@ def _get_portal_secret() -> bytes | None:
             interface="org.freedesktop.portal.Secret",
         )
 
+        # The portal reports the outcome with a Response signal on the
+        # returned request object. Subscribe before sending the request so
+        # the signal can not be missed.
+        response_rule = MatchRule(
+            type="signal",
+            interface="org.freedesktop.portal.Request",
+            member="Response",
+        )
+
         # Create a socket pair for receiving the secret
         sock_read, sock_write = socket.socketpair(socket.AF_UNIX, socket.SOCK_STREAM)
 
         try:
-            # Build the method call with file descriptor
-            # RetrieveSecret(fd: h, options: a{sv}) -> (handle: o)
-            # Pass the socket object directly - jeepney handles fd passing
-            msg = new_method_call(portal, "RetrieveSecret", "ha{sv}", (sock_write, {}))
+            Proxy(message_bus, conn).AddMatch(response_rule)
+
+            with conn.filter(response_rule) as responses:
+                # Build the method call with file descriptor
+                # RetrieveSecret(fd: h, options: a{sv}) -> (handle: o)
+                # Pass the socket object directly - jeepney handles fd passing
+                msg = new_method_call(portal, "RetrieveSecret", "ha{sv}", (sock_write, {}))
 
-            # Send message and get reply
-            conn.send_and_get_reply(msg, timeout=10)
+                # Send message and get reply
+                reply = conn.send_and_get_reply(msg, timeout=10)
+                request_handle = reply.body[0]
+
+                # Close the write end - portal has it now
+                sock_write.close()
+                sock_write = None
+
+                # Wait for the Response of our request. A non zero code means
+                # the portal will not send a secret, e.g. when the app is not
+                # sandboxed, so there is no point waiting on the socket.
+                deadline = time.monotonic() + PORTAL_RESPONSE_TIMEOUT_SECONDS
+                while True:
+                    signal = conn.recv_until_filtered(
+                        responses, timeout=max(deadline - time.monotonic(), 0)
+                    )
+                    if signal.header.fields.get(HeaderFields.path) == request_handle:
+                        response_code = signal.body[0]
+                        break
 
-            # Close the write end - portal has it now
-            sock_write.close()
-            sock_write = None
+            if response_code != 0:
+                logging.debug(
+                    "XDG Portal secret not available, response code %s",
+                    response_code,
+                )
+                return None
 
             # Read the secret from the read end
             # The portal writes the secret and closes its end
-            sock_read.settimeout(5.0)
+            sock_read.settimeout(PORTAL_RESPONSE_TIMEOUT_SECONDS)
             secret_data = b""
             while True:
                 try:
@@ -90,6 +144,7 @@ def _get_portal_secret() -> bytes | None:
             sock_read.close()
             if sock_write is not None:
                 sock_write.close()
+            conn.close()
 
     except Exception as exc:
         logging.debug("XDG Portal secret not available: %s", exc)

buzz/transcriber/transcriber.py

@@ -162,12 +162,6 @@ class TranscriptionOptions:
     transcription_step: float = 3.5
 
 
-def humanize_language(language: str) -> str:
-    if language == "":
-        return _("Detect Language")
-    return LANGUAGES[language].title()
-
-
 @dataclass()
 class FileTranscriptionOptions:
     file_paths: Optional[List[str]] = None
@@ -228,10 +222,6 @@ class OutputFormat(enum.Enum):
     VTT = "vtt"
 
 
-class Stopped(Exception):
-    pass
-
-
 SUPPORTED_AUDIO_EXTENSIONS = (
     ".mp3", ".wav", ".m4a", ".m4b", ".aac", ".ogg", ".opus", ".flac",
 )

buzz/transcriber/whisper_file_transcriber.py

@@ -13,13 +13,11 @@
 import platform
 
 from buzz.transcriber.cuda_device import cuda_works
-import subprocess
 from platformdirs import user_cache_dir
 from multiprocessing.connection import Connection
 from threading import Lock, Thread
 from typing import Optional, List
 
-import tqdm
 import psutil
 from PyQt6.QtCore import QObject
 
@@ -334,7 +332,6 @@ def transcribe_faster_whisper(cls, task: FileTranscriptionTask) -> List[Segment]
                 "Faster Whisper model is not available locally. "
                 "Check BUZZ_MODEL_ROOT and download the model into that cache first."
             )
-            return []
 
         model_root_dir = user_cache_dir("Buzz")
         model_root_dir = os.path.join(model_root_dir, "models")

buzz/translator.py

@@ -4,10 +4,9 @@
 import queue
 
 from typing import Optional, List, Tuple
-from openai import OpenAI, max_retries
+from openai import OpenAI
 from PyQt6.QtCore import QObject, pyqtSignal
 
-from buzz.locale import _
 from buzz.settings.settings import Settings
 from buzz.store.keyring_store import get_password, Key
 from buzz.transcriber.transcriber import TranscriptionOptions

buzz/widgets/audio_player.py

@@ -54,13 +54,16 @@ def _init_sounddevice(self, file_path: str):
             logging.warning("AudioPlayer: sounddevice init failed, using Qt fallback", exc_info=True)
 
     def _init_media_player(self, file_path: str):
-        self.audio_output = QAudioOutput()
+        # Parent the multimedia objects to this widget so Qt destroys them along
+        # with it. Left parentless, they can outlive the widget until interpreter
+        # exit, where PyQt's cleanup segfaults on them.
+        self.audio_output = QAudioOutput(self)
         self.audio_output.setVolume(100)
-        self.media_player = QMediaPlayer()
+        self.media_player = QMediaPlayer(self)
         self.media_player.setSource(QUrl.fromLocalFile(file_path))
         self.media_player.setAudioOutput(self.audio_output)
 
-        self.media_devices = QMediaDevices()
+        self.media_devices = QMediaDevices(self)
         self.media_devices.audioOutputsChanged.connect(self.on_audio_outputs_changed)
 
         if self.is_video:
@@ -258,9 +261,6 @@ def set_range(self, range_ms: Tuple[int, int]):
     def clear_range(self):
         self.range_ms = None
 
-    def _reset_looping_flag(self):
-        self.is_looping = False
-
     def on_slider_moved(self, position_ms: int):
         self.set_position(position_ms)
         if self.range_ms is not None:
@@ -299,9 +299,11 @@ def on_audio_outputs_changed(self):
             return
         was_playing = self.media_player.playbackState() == QMediaPlayer.PlaybackState.PlayingState
         position = self.media_player.position()
-        self.audio_output = QAudioOutput()
+        old_audio_output = self.audio_output
+        self.audio_output = QAudioOutput(self)
         self.audio_output.setVolume(100)
         self.media_player.setAudioOutput(self.audio_output)
+        old_audio_output.deleteLater()
         self.media_player.setPosition(position)
         if was_playing:
             self.media_player.play()

buzz/widgets/icon.py

@@ -59,16 +59,6 @@ def __init__(self, parent: QWidget):
         super().__init__(get_path("assets/pause_black_24dp.svg"), parent)
 
 
-class UndoIcon(Icon):
-    def __init__(self, parent: QWidget):
-        super().__init__(get_path("assets/undo_FILL0_wght700_GRAD0_opsz48.svg"), parent)
-
-
-class RedoIcon(Icon):
-    def __init__(self, parent: QWidget):
-        super().__init__(get_path("assets/redo_FILL0_wght700_GRAD0_opsz48.svg"), parent)
-
-
 class FileDownloadIcon(Icon):
     def __init__(self, parent: QWidget):
         super().__init__(get_path("assets/file_download_black_24dp.svg"), parent)

buzz/widgets/main_window.py

@@ -3,7 +3,7 @@
 from typing import Tuple, List, Optional, Set
 from uuid import UUID
 
-from PyQt6 import QtGui
+from PyQt6 import QtGui, sip
 from PyQt6.QtCore import (
     Qt,
     QThread,
@@ -459,6 +459,12 @@ def open_transcription_viewer(self, transcription: Transcription):
             flags=Qt.WindowType.Window,
             transcriptions_updated_signal=self.transcriptions_updated,
         )
+        # Free the viewer and its media player once closed. Otherwise every
+        # closed viewer lingers until exit, holding its decoded audio, and
+        # PyQt's cleanup at interpreter exit segfaults on its media player.
+        self.transcription_viewer_widget.setAttribute(
+            Qt.WidgetAttribute.WA_DeleteOnClose
+        )
         self.transcription_viewer_widget.show()
 
     def add_task(self, task: FileTranscriptionTask):
@@ -589,8 +595,13 @@ def closeEvent(self, event: QtGui.QCloseEvent) -> None:
                 if not self.transcriber_thread.wait(2000):
                     logging.error("Transcriber thread could not be terminated")
 
-        if self.transcription_viewer_widget is not None:
-            self.transcription_viewer_widget.close()
+        # Several viewers can be open at once, close them all. Delete them right
+        # away as the app may exit before their deferred deletion runs, and
+        # PyQt's cleanup at interpreter exit segfaults on their media players.
+        for viewer in self.findChildren(TranscriptionViewerWidget):
+            viewer.close()
+            sip.delete(viewer)
+        self.transcription_viewer_widget = None
 
         try:
             from buzz.widgets.application import Application

buzz/widgets/transcriber/hugging_face_search_line_edit.py

@@ -94,9 +94,6 @@ def fetch_models(self):
 
         return self.network_manager.get(QNetworkRequest(url))
 
-    def on_popup_selected(self):
-        self.timer.stop()
-
     def on_request_response(self, network_reply: QNetworkReply):
         if network_reply.error() != QNetworkReply.NetworkError.NoError:
             logging.debug(

buzz/widgets/transcriber/mms_language_line_edit.py

@@ -42,7 +42,3 @@ def _on_text_changed(self, text: str):
     def language(self) -> str:
         """Get the current language code."""
         return self.text().strip().lower()
-
-    def setLanguage(self, language: str):
-        """Set the language code."""
-        self.setText(language.strip().lower() if language else "eng")

buzz/widgets/transcription_viewer/speaker_identification_widget.py

@@ -137,8 +137,6 @@ def hide_cuda_from_torch(device):
         torch.cuda.is_available = original_is_available
 
 
-SENTENCE_END = re.compile(r'.*[.!?。！？]')
-
 class IdentificationWorker(QObject):
     finished = pyqtSignal(list)
     progress_update = pyqtSignal(str)
@@ -188,33 +186,6 @@ def _get_language(self) -> Optional[str]:
             return self.language
         return self.transcription.language
 
-    def get_transcript(self, audio, **kwargs) -> dict:
-        buzz_segments = self._get_segments()
-
-        segments = []
-        words = []
-        text = ""
-        for buzz_segment in buzz_segments:
-            words.append({
-                'word': buzz_segment.text + " ",
-                'start': buzz_segment.start_time / 100,
-                'end': buzz_segment.end_time / 100,
-            })
-            text += buzz_segment.text + " "
-
-            if SENTENCE_END.match(buzz_segment.text):
-                segments.append({
-                    'text': text,
-                    'words': words
-                })
-                words = []
-                text = ""
-
-        return {
-            'language': self._get_language(),
-            'segments': segments
-        }
-
     def _import_libraries(self):
         from ctc_forced_aligner.ctc_forced_aligner import (
             generate_emissions,

buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py

@@ -139,11 +139,6 @@ def focusOutEvent(self, event):
         super().focusOutEvent(event)
 
 
-class TimeStampDelegate(QStyledItemDelegate):
-    def displayText(self, value, locale):
-        return to_timestamp(value)
-
-
 class TimeStampEditorDelegate(QStyledItemDelegate):
     """Delegate for editing timestamps with overlap prevention"""