Daemonize addon store, timeout web requests

source/NVDAObjects/behaviors.py

@@ -350,9 +350,9 @@ def startMonitoring(self):
 			return
 		thread = self._monitorThread = threading.Thread(
 			name=f"{self.__class__.__qualname__}._monitorThread",
-			target=self._monitor
+			target=self._monitor,
+			daemon=True,
 		)
-		thread.daemon = True
 		self._keepMonitoring = True
 		self._event.clear()
 		thread.start()

source/UIAHandler/__init__.py

@@ -427,9 +427,9 @@ def __init__(self):
 		self.MTAThreadInitException=None
 		self.MTAThread = threading.Thread(
 			name=f"{self.__class__.__module__}.{self.__class__.__qualname__}.MTAThread",
-			target=self.MTAThreadFunc
+			target=self.MTAThreadFunc,
+			daemon=True,
 		)
-		self.MTAThread.daemon=True
 		self.MTAThread.start()
 		self.MTAThreadInitEvent.wait(2)
 		if self.MTAThreadInitException:

source/_addonStore/dataManager.py

@@ -56,6 +56,7 @@
 
 
 addonDataManager: Optional["_DataManager"] = None
+FETCH_TIMEOUT_S = 120  # seconds
 
 
 def initialize():
@@ -67,6 +68,16 @@ def initialize():
 	addonDataManager = _DataManager()
 
 
+def terminate():
+	global addonDataManager
+	if config.isAppX:
+		log.info("Add-ons not supported when running as a Windows Store application")
+		return
+	addonDataManager.terminate()
+	log.debug("terminating addonStore data manager")
+	addonDataManager = None
+
+
 class _DataManager:
 	_cacheLatestFilename: str = "_cachedLatestAddons.json"
 	_cacheCompatibleFilename: str = "_cachedCompatibleAddons.json"
@@ -89,15 +100,24 @@ def __init__(self):
 		self._compatibleAddonCache = self._getCachedAddonData(self._cacheCompatibleFile)
 		self._installedAddonsCache = _InstalledAddonsCache()
 		# Fetch available add-ons cache early
-		threading.Thread(
+		self._initialiseAvailableAddonsThread = threading.Thread(
 			target=self.getLatestCompatibleAddons,
 			name="initialiseAvailableAddons",
-		).start()
+			daemon=True,
+		)
+		self._initialiseAvailableAddonsThread.start()
+
+	def terminate(self):
+		if self._initialiseAvailableAddonsThread.is_alive():
+			self._initialiseAvailableAddonsThread.join(timeout=1)
+		if self._initialiseAvailableAddonsThread.is_alive():
+			log.debugWarning("initialiseAvailableAddons thread did not terminate immediately")
 
 	def _getLatestAddonsDataForVersion(self, apiVersion: str) -> Optional[bytes]:
 		url = _getAddonStoreURL(self._preferredChannel, self._lang, apiVersion)
 		try:
-			response = requests.get(url)
+			log.debug(f"Fetching add-on data from {url}")
+			response = requests.get(url, timeout=FETCH_TIMEOUT_S)
 		except requests.exceptions.RequestException as e:
 			log.debugWarning(f"Unable to fetch addon data: {e}")
 			return None
@@ -112,7 +132,8 @@ def _getLatestAddonsDataForVersion(self, apiVersion: str) -> Optional[bytes]:
 	def _getCacheHash(self) -> Optional[str]:
 		url = _getCacheHashURL()
 		try:
-			response = requests.get(url)
+			log.debug(f"Fetching add-on data from {url}")
+			response = requests.get(url, timeout=FETCH_TIMEOUT_S)
 		except requests.exceptions.RequestException as e:
 			log.debugWarning(f"Unable to get cache hash: {e}")
 			return None

source/_addonStore/network.py

@@ -178,7 +178,10 @@ def _downloadAddonToPath(
 		if not NVDAState.shouldWriteToDisk():
 			return False
 
-		with requests.get(addonData.model.URL, stream=True) as r:
+		# Some add-ons are quite large, so we need to allow for a long download time.
+		# 1GB at 0.5 MB/s takes 4.5hr to download.
+		MAX_ADDON_DOWNLOAD_TIME = 60 * 60 * 6  # 6 hours
+		with requests.get(addonData.model.URL, stream=True, timeout=MAX_ADDON_DOWNLOAD_TIME) as r:
 			with open(downloadFilePath, 'wb') as fd:
 				# Most add-ons are small. This value was chosen quite arbitrarily, but with the intention to allow
 				# interrupting the download. This is particularly important on a slow connection, to provide

source/buildVersion.py

@@ -67,7 +67,7 @@ def formatVersionForGUI(year, major, minor):
 name = "NVDA"
 version_year = 2023
 version_major = 3
-version_minor = 3
+version_minor = 4
 version_build = 0  # Should not be set manually. Set in 'sconscript' provided by 'appVeyor.yml'
 version=_formatDevVersionString()
 publisher="unknown"

source/core.py

@@ -240,6 +240,10 @@ def resetConfiguration(factoryDefaults=False):
 	hwIo.terminate()
 	log.debug("terminating addonHandler")
 	addonHandler.terminate()
+	# Addons
+	from _addonStore import dataManager
+	log.debug("terminating addon dataManager")
+	dataManager.terminate()
 	log.debug("Reloading config")
 	config.conf.reset(factoryDefaults=factoryDefaults)
 	logHandler.setLogLevelFromConfig()
@@ -250,8 +254,6 @@ def resetConfiguration(factoryDefaults=False):
 		lang = config.conf["general"]["language"]
 	log.debug("setting language to %s"%lang)
 	languageHandler.setLanguage(lang)
-	# Addons
-	from _addonStore import dataManager
 	dataManager.initialize()
 	addonHandler.initialize()
 	# Hardware background i/o
@@ -856,6 +858,7 @@ def _doPostNvdaStartupAction():
 	_terminate(bdDetect)
 	_terminate(hwIo)
 	_terminate(addonHandler)
+	_terminate(dataManager, name="addon dataManager")
 	_terminate(garbageHandler)
 	# DMP is only started if needed.
 	# Terminate manually (and let it write to the log if necessary)

source/gui/_addonStoreGui/viewModels/store.py

@@ -383,7 +383,12 @@ def refresh(self):
 			_StatusFilterKey.AVAILABLE,
 			_StatusFilterKey.UPDATE,
 		}:
-			threading.Thread(target=self._getAvailableAddonsInBG, name="getAddonData").start()
+			self._refreshAddonsThread = threading.Thread(
+				target=self._getAvailableAddonsInBG,
+				name="getAddonData",
+				daemon=True,
+			)
+			self._refreshAddonsThread.start()
 
 		elif self._filteredStatusKey in {
 			_StatusFilterKey.INSTALLED,

source/nvwave.py

@@ -721,7 +721,8 @@ def play():
 			fileWavePlayerThread.join()
 		fileWavePlayerThread = threading.Thread(
 			name=f"{__name__}.playWaveFile({os.path.basename(fileName)})",
-			target=play
+			target=play,
+			daemon=True,
 		)
 		fileWavePlayerThread.start()
 	else:

source/remotePythonConsole.py

@@ -76,9 +76,9 @@ def initialize():
 	server.daemon_threads = True
 	thread = threading.Thread(
 		name=__name__,  # remotePythonConsole
-		target=server.serve_forever
+		target=server.serve_forever,
+		daemon=True,
 	)
-	thread.daemon = True
 	thread.start()
 
 def terminate():

source/updateCheck.py

@@ -107,6 +107,9 @@ def getQualifiedDriverClassNameForStats(cls):
 	return "%s (core)"%name
 
 
+UPDATE_FETCH_TIMEOUT_S = 30  # seconds
+
+
 def checkForUpdate(auto: bool = False) -> Optional[Dict]:
 	"""Check for an updated version of NVDA.
 	This will block, so it generally shouldn't be called from the main thread.
@@ -151,14 +154,16 @@ def checkForUpdate(auto: bool = False) -> Optional[Dict]:
 		params.update(extraParams)
 	url = "%s?%s" % (CHECK_URL, urllib.parse.urlencode(params))
 	try:
-		res = urllib.request.urlopen(url)
+		log.debug(f"Fetching update data from {url}")
+		res = urllib.request.urlopen(url, timeout=UPDATE_FETCH_TIMEOUT_S)
 	except IOError as e:
 		if isinstance(e.reason, ssl.SSLCertVerificationError) and e.reason.reason == "CERTIFICATE_VERIFY_FAILED":
 			# #4803: Windows fetches trusted root certificates on demand.
 			# Python doesn't trigger this fetch (PythonIssue:20916), so try it ourselves
 			_updateWindowsRootCertificates()
 			# and then retry the update check.
-			res = urllib.request.urlopen(url)
+			log.debug(f"Fetching update data from {url}")
+			res = urllib.request.urlopen(url, timeout=UPDATE_FETCH_TIMEOUT_S)
 		else:
 			raise
 	if res.code != 200:
@@ -253,9 +258,9 @@ def check(self):
 		"""
 		t = threading.Thread(
 			name=f"{self.__class__.__module__}.{self.check.__qualname__}",
-			target=self._bg
+			target=self._bg,
+			daemon=True,
 		)
-		t.daemon = True
 		self._started()
 		t.start()
 
@@ -614,9 +619,9 @@ def start(self):
 		self._progressDialog.Raise()
 		t = threading.Thread(
 			name=f"{self.__class__.__module__}.{self.start.__qualname__}",
-			target=self._bg
+			target=self._bg,
+			daemon=True,
 		)
-		t.daemon = True
 		t.start()
 
 	def _guiExec(self, func, *args):
@@ -656,7 +661,12 @@ def _download(self, url):
 		# #2352: Some security scanners such as Eset NOD32 HTTP Scanner
 		# cause huge read delays while downloading.
 		# Therefore, set a higher timeout.
-		remote = urllib.request.urlopen(url, timeout=120)
+		# The NVDA exe is about 35 MB.
+		# The average download speed in the world is 0.5 MB/s
+		# in some developing countries with the slowest internet.
+		# This yields an expected download time of 10min on slower networks.
+		UPDATE_DOWNLOAD_TIMEOUT = 60 * 30  # 30 min
+		remote = urllib.request.urlopen(url, timeout=UPDATE_DOWNLOAD_TIMEOUT)
 		if remote.code != 200:
 			raise RuntimeError("Download failed with code %d" % remote.code)
 		size = int(remote.headers["content-length"])
@@ -851,12 +861,17 @@ class CERT_CHAIN_PARA(ctypes.Structure):
 	)
 
 def _updateWindowsRootCertificates():
+	log.debug("Updating Windows root certificates")
 	crypt = ctypes.windll.crypt32
 	# Get the server certificate.
 	sslCont = ssl._create_unverified_context()
 	# We must specify versionType so the server doesn't return a 404 error and
 	# thus cause an exception.
-	u = urllib.request.urlopen(CHECK_URL + "?versionType=stable", context=sslCont)
+	u = urllib.request.urlopen(
+		CHECK_URL + "?versionType=stable",
+		context=sslCont,
+		timeout=UPDATE_FETCH_TIMEOUT_S,
+	)
 	cert = u.fp.raw._sock.getpeercert(True)
 	u.close()
 	# Convert to a form usable by Windows.

source/virtualBuffers/__init__.py

@@ -451,8 +451,9 @@ def loadBuffer(self):
 		self._loadProgressCallLater = wx.CallLater(1000, self._loadProgress)
 		threading.Thread(
 			name=f"{self.__class__.__module__}.{self.loadBuffer.__qualname__}",
-			target=self._loadBuffer).start(
-		)
+			target=self._loadBuffer,
+			daemon=True,
+		).start()
 
 	def _loadBuffer(self):
 		try:

source/visionEnhancementProviders/NVDAHighlighter.py

@@ -418,10 +418,10 @@ def __init__(self):
 		winGDI.gdiPlusInitialize()
 		self._highlighterThread = threading.Thread(
 			name=f"{self.__class__.__module__}.{self.__class__.__qualname__}",
-			target=self._run
+			target=self._run,
+			daemon=True,
 		)
 		self._highlighterRunningEvent = threading.Event()
-		self._highlighterThread.daemon = True
 		self._highlighterThread.start()
 		# Make sure the highlighter thread doesn't exit early.
 		waitResult = self._highlighterRunningEvent.wait(0.2)

source/watchdog.py

@@ -292,7 +292,8 @@ def initialize():
 	NVDAHelper._setDllFuncPointer(NVDAHelper.localLib, "_notifySendMessageCancelled", _notifySendMessageCancelled)
 	_watcherThread = threading.Thread(
 		name=__name__,
-		target=_watcher
+		target=_watcher,
+		daemon=True,
 	)
 	alive()
 	_watcherThread.start()

source/winInputHook.py

@@ -91,7 +91,8 @@ def initialize():
 	if hookThreadRefCount==1:
 		hookThread = threading.Thread(
 			name=__name__,  # winInputHook
-			target=hookThreadFunc
+			target=hookThreadFunc,
+			daemon=True,
 		)
 		hookThread.start()
 

tests/system/libraries/SystemTestSpy/speechSpyGlobalPlugin.py

@@ -534,7 +534,11 @@ def _start(self):
 			serve=False  # we want to start this serving on another thread so as not to block.
 		)
 		log.debug("Server address: {}".format(server.server_address))
-		server_thread = threading.Thread(target=server.serve, name="RF Test Spy Thread")
+		server_thread = threading.Thread(
+			target=server.serve,
+			name="RF Test Spy Thread",
+			daemon=True,
+		)
 		server_thread.start()
 
 	def terminate(self):

tests/system/libraries/SystemTestSpy/speechSpySynthDriver.py

@@ -39,6 +39,7 @@ def __init__(self):
 		self._doSpeechThread = threading.Thread(
 			target=self._processSpeech,
 			name="speech spy synth driver",
+			daemon=True,
 		)
 		self._doSpeechThread.start()
 

user_docs/en/changes.t2t

@@ -4,6 +4,14 @@ What's New in NVDA
 %!includeconf: ../changes.t2tconf
 %!includeconf: ./locale.t2tconf
 
+= 2023.3.4 =
+This is a patch release to fix an installer issue.
+
+== Bug Fixes ==
+- Fixed bug which caused the NVDA process to fail to exit correctly.
+When running the installer, this resulted in the installation entering an unrecoverable state. (#16122, #16123)
+-
+
 = 2023.3.3 =
 This is a patch release to fix a security issue.
 Please responsibly disclose security issues following NVDA's [security policy https://github.com/nvaccess/nvda/blob/master/security.md].