Fixes for image

MANIFEST.in

@@ -11,6 +11,7 @@ recursive-include changedetectionio/realtime *
 recursive-include changedetectionio/static *
 recursive-include changedetectionio/templates *
 recursive-include changedetectionio/tests *
+recursive-include changedetectionio/translations *
 recursive-include changedetectionio/widgets *
 prune changedetectionio/static/package-lock.json
 prune changedetectionio/static/styles/node_modules

changedetectionio/__init__.py

@@ -2,7 +2,7 @@
 
 # Read more https://github.com/dgtlmoon/changedetection.io/wiki
 # Semver means never use .01, or 00. Should be .1.
-__version__ = '0.52.2'
+__version__ = '0.52.4'
 
 from changedetectionio.strtobool import strtobool
 from json.decoder import JSONDecodeError
@@ -41,9 +41,10 @@ from loguru import logger
 #
 # IMPLEMENTATION:
 #   1. Explicit contexts everywhere (primary protection):
-#      - Watch.py: ctx = multiprocessing.get_context('spawn')
 #      - playwright.py: ctx = multiprocessing.get_context('spawn')
 #      - puppeteer.py: ctx = multiprocessing.get_context('spawn')
+#      - isolated_opencv.py: ctx = multiprocessing.get_context('spawn')
+#      - isolated_libvips.py: ctx = multiprocessing.get_context('spawn')
 #
 #   2. Global default (defense-in-depth, below):
 #      - Safety net if future code forgets explicit context

changedetectionio/async_update_worker.py

@@ -1,5 +1,4 @@
 from blinker import signal
-
 from .processors.exceptions import ProcessorException
 import changedetectionio.content_fetchers.exceptions as content_fetchers_exceptions
 from changedetectionio.processors.text_json_diff.processor import FilterNotFoundInResponse
@@ -9,7 +8,6 @@ from changedetectionio.flask_app import watch_check_update
 import asyncio
 import importlib
 import os
-import queue
 import time
 
 from loguru import logger

changedetectionio/blueprint/browser_steps/__init__.py

@@ -21,31 +21,154 @@ from changedetectionio.flask_app import login_optionally_required
 from loguru import logger
 
 browsersteps_sessions = {}
+browsersteps_watch_to_session = {}  # Maps watch_uuid -> browsersteps_session_id
 io_interface_context = None
 import json
 import hashlib
 from flask import Response
 import asyncio
 import threading
+import time
 
-def run_async_in_browser_loop(coro):
-    """Run async coroutine using the existing async worker event loop"""
-    from changedetectionio import worker_handler
-    
-    # Use the existing async worker event loop instead of creating a new one
-    if worker_handler.USE_ASYNC_WORKERS and worker_handler.async_loop and not worker_handler.async_loop.is_closed():
-        logger.debug("Browser steps using existing async worker event loop")
-        future = asyncio.run_coroutine_threadsafe(coro, worker_handler.async_loop)
-        return future.result()
-    else:
-        # Fallback: create a new event loop (for sync workers or if async loop not available)
-        logger.debug("Browser steps creating temporary event loop")
-        loop = asyncio.new_event_loop()
-        asyncio.set_event_loop(loop)
+# Dedicated event loop for ALL browser steps sessions
+_browser_steps_loop = None
+_browser_steps_thread = None
+_browser_steps_loop_lock = threading.Lock()
+
+def _start_browser_steps_loop():
+    """Start a dedicated event loop for browser steps in its own thread"""
+    global _browser_steps_loop
+
+    # Create and set the event loop for this thread
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+    _browser_steps_loop = loop
+
+    logger.debug("Browser steps event loop started")
+
+    try:
+        # Run the loop forever - handles all browsersteps sessions
+        loop.run_forever()
+    except Exception as e:
+        logger.error(f"Browser steps event loop error: {e}")
+    finally:
         try:
-            return loop.run_until_complete(coro)
+            # Cancel all remaining tasks
+            pending = asyncio.all_tasks(loop)
+            for task in pending:
+                task.cancel()
+
+            # Wait for tasks to finish cancellation
+            if pending:
+                loop.run_until_complete(asyncio.gather(*pending, return_exceptions=True))
+        except Exception as e:
+            logger.debug(f"Error during browser steps loop cleanup: {e}")
         finally:
             loop.close()
+            logger.debug("Browser steps event loop closed")
+
+def _ensure_browser_steps_loop():
+    """Ensure the browser steps event loop is running"""
+    global _browser_steps_loop, _browser_steps_thread
+
+    with _browser_steps_loop_lock:
+        if _browser_steps_thread is None or not _browser_steps_thread.is_alive():
+            logger.debug("Starting browser steps event loop thread")
+            _browser_steps_thread = threading.Thread(
+                target=_start_browser_steps_loop,
+                daemon=True,
+                name="BrowserStepsEventLoop"
+            )
+            _browser_steps_thread.start()
+
+            # Wait for the loop to be ready
+            timeout = 5.0
+            start_time = time.time()
+            while _browser_steps_loop is None:
+                if time.time() - start_time > timeout:
+                    raise RuntimeError("Browser steps event loop failed to start")
+                time.sleep(0.01)
+
+            logger.debug("Browser steps event loop thread started and ready")
+
+def run_async_in_browser_loop(coro):
+    """Run async coroutine using the dedicated browser steps event loop"""
+    _ensure_browser_steps_loop()
+
+    if _browser_steps_loop and not _browser_steps_loop.is_closed():
+        logger.debug("Browser steps using dedicated event loop")
+        future = asyncio.run_coroutine_threadsafe(coro, _browser_steps_loop)
+        return future.result()
+    else:
+        raise RuntimeError("Browser steps event loop is not available")
+
+def cleanup_expired_sessions():
+    """Remove expired browsersteps sessions and cleanup their resources"""
+    global browsersteps_sessions, browsersteps_watch_to_session
+
+    expired_session_ids = []
+
+    # Find expired sessions
+    for session_id, session_data in browsersteps_sessions.items():
+        browserstepper = session_data.get('browserstepper')
+        if browserstepper and browserstepper.has_expired:
+            expired_session_ids.append(session_id)
+
+    # Cleanup expired sessions
+    for session_id in expired_session_ids:
+        logger.debug(f"Cleaning up expired browsersteps session {session_id}")
+        session_data = browsersteps_sessions[session_id]
+
+        # Cleanup playwright resources asynchronously
+        browserstepper = session_data.get('browserstepper')
+        if browserstepper:
+            try:
+                run_async_in_browser_loop(browserstepper.cleanup())
+            except Exception as e:
+                logger.error(f"Error cleaning up session {session_id}: {e}")
+
+        # Remove from sessions dict
+        del browsersteps_sessions[session_id]
+
+        # Remove from watch mapping
+        for watch_uuid, mapped_session_id in list(browsersteps_watch_to_session.items()):
+            if mapped_session_id == session_id:
+                del browsersteps_watch_to_session[watch_uuid]
+                break
+
+    if expired_session_ids:
+        logger.info(f"Cleaned up {len(expired_session_ids)} expired browsersteps session(s)")
+
+def cleanup_session_for_watch(watch_uuid):
+    """Cleanup a specific browsersteps session for a watch UUID"""
+    global browsersteps_sessions, browsersteps_watch_to_session
+
+    session_id = browsersteps_watch_to_session.get(watch_uuid)
+    if not session_id:
+        logger.debug(f"No browsersteps session found for watch {watch_uuid}")
+        return
+
+    logger.debug(f"Cleaning up browsersteps session {session_id} for watch {watch_uuid}")
+
+    session_data = browsersteps_sessions.get(session_id)
+    if session_data:
+        browserstepper = session_data.get('browserstepper')
+        if browserstepper:
+            try:
+                run_async_in_browser_loop(browserstepper.cleanup())
+            except Exception as e:
+                logger.error(f"Error cleaning up session {session_id} for watch {watch_uuid}: {e}")
+
+        # Remove from sessions dict
+        del browsersteps_sessions[session_id]
+
+    # Remove from watch mapping
+    del browsersteps_watch_to_session[watch_uuid]
+
+    logger.debug(f"Cleaned up session for watch {watch_uuid}")
+
+    # Opportunistically cleanup any other expired sessions
+    cleanup_expired_sessions()
 
 def construct_blueprint(datastore: ChangeDetectionStore):
     browser_steps_blueprint = Blueprint('browser_steps', __name__, template_folder="templates")
@@ -123,6 +246,9 @@ def construct_blueprint(datastore: ChangeDetectionStore):
         if not watch_uuid:
             return make_response('No Watch UUID specified', 500)
 
+        # Cleanup any existing session for this watch
+        cleanup_session_for_watch(watch_uuid)
+
         logger.debug("Starting connection with playwright")
         logger.debug("browser_steps.py connecting")
 
@@ -131,6 +257,10 @@ def construct_blueprint(datastore: ChangeDetectionStore):
             browsersteps_sessions[browsersteps_session_id] = run_async_in_browser_loop(
                 start_browsersteps_session(watch_uuid)
             )
+
+            # Store the mapping of watch_uuid -> browsersteps_session_id
+            browsersteps_watch_to_session[watch_uuid] = browsersteps_session_id
+
         except Exception as e:
             if 'ECONNREFUSED' in str(e):
                 return make_response('Unable to start the Playwright Browser session, is sockpuppetbrowser running? Network configuration is OK?', 401)

changedetectionio/blueprint/tags/templates/groups-overview.html

@@ -50,7 +50,8 @@
                 <td>{{ "{:,}".format(tag_count[uuid]) if uuid in tag_count else 0 }}</td>
                 <td class="title-col inline"> <a href="{{url_for('watchlist.index', tag=uuid) }}">{{ tag.title }}</a></td>
                 <td>
-                    <a class="pure-button pure-button-primary" href="{{ url_for('tags.form_tag_edit', uuid=uuid) }}">{{ _('Edit') }}</a>&nbsp;
+                    <a class="pure-button pure-button-primary" href="{{ url_for('tags.form_tag_edit', uuid=uuid) }}">{{ _('Edit') }}</a>
+                    <a href="{{ url_for('ui.form_watch_checknow', tag=uuid) }}" class="pure-button pure-button-primary" >{{ _('Recheck') }}</a>
                     <a class="pure-button button-error"
                        href="{{ url_for('tags.delete', uuid=uuid) }}"
                        data-requires-confirm

changedetectionio/blueprint/ui/edit.py

@@ -238,6 +238,13 @@ def construct_blueprint(datastore: ChangeDetectionStore, update_q, queuedWatchMe
             datastore.data['watching'][uuid] = watch_class(datastore_path=datastore.datastore_path, default=datastore.data['watching'][uuid])
             flash(gettext("Updated watch - unpaused!") if request.args.get('unpause_on_save') else gettext("Updated watch."))
 
+            # Cleanup any browsersteps session for this watch
+            try:
+                from changedetectionio.blueprint.browser_steps import cleanup_session_for_watch
+                cleanup_session_for_watch(uuid)
+            except Exception as e:
+                logger.debug(f"Error cleaning up browsersteps session: {e}")
+
             # Re #286 - We wait for syncing new data to disk in another thread every 60 seconds
             # But in the case something is added we should save straight away
             datastore.needs_write_urgent = True
@@ -325,8 +332,6 @@ def construct_blueprint(datastore: ChangeDetectionStore, update_q, queuedWatchMe
                     'url': url_for('rss.rss_single_watch', uuid=watch['uuid'], token=app_rss_token)
                 },
                 'settings_application': datastore.data['settings']['application'],
-                'system_has_playwright_configured': os.getenv('PLAYWRIGHT_DRIVER_URL'),
-                'system_has_webdriver_configured': os.getenv('WEBDRIVER_URL'),
                 'ui_edit_stats_extras': collect_ui_edit_stats_extras(watch),
                 'visual_selector_data_ready': datastore.visualselector_data_is_ready(watch_uuid=uuid),
                 'timezone_default_config': datastore.data['settings']['application'].get('scheduler_timezone_default'),

changedetectionio/blueprint/ui/templates/edit.html

@@ -206,7 +206,7 @@ Math: {{ 1 + 1 }}") }}
 
             <div class="tab-pane-inner" id="browser-steps">
             {% if capabilities.supports_browser_steps %}
-               {% if visual_selector_data_ready %}
+               {% if true %}
                 <img class="beta-logo" src="{{url_for('static_content', group='images', filename='beta-logo.png')}}" alt="New beta functionality">
                 <fieldset>
                     <div class="pure-control-group">

changedetectionio/content_fetchers/playwright.py

@@ -1,3 +1,4 @@
+import gc
 import json
 import os
 from urllib.parse import urlparse
@@ -185,20 +186,33 @@ class fetcher(Fetcher):
         super().screenshot_step(step_n=step_n)
         screenshot = await capture_full_page_async(page=self.page, screenshot_format=self.screenshot_format)
 
+        # Request GC immediately after screenshot to free memory
+        # Screenshots can be large and browser steps take many of them
+        await self.page.request_gc()
 
         if self.browser_steps_screenshot_path is not None:
             destination = os.path.join(self.browser_steps_screenshot_path, 'step_{}.jpeg'.format(step_n))
             logger.debug(f"Saving step screenshot to {destination}")
             with open(destination, 'wb') as f:
                 f.write(screenshot)
+            # Clear local reference to allow screenshot bytes to be collected
+            del screenshot
+            gc.collect()
 
     async def save_step_html(self, step_n):
         super().save_step_html(step_n=step_n)
         content = await self.page.content()
+
+        # Request GC after getting page content
+        await self.page.request_gc()
+
         destination = os.path.join(self.browser_steps_screenshot_path, 'step_{}.html'.format(step_n))
         logger.debug(f"Saving step HTML to {destination}")
         with open(destination, 'w', encoding='utf-8') as f:
             f.write(content)
+        # Clear local reference
+        del content
+        gc.collect()
 
     async def run(self,
                   fetch_favicon=True,
@@ -305,6 +319,12 @@ class fetcher(Fetcher):
 
             if self.status_code != 200 and not ignore_status_codes:
                 screenshot = await capture_full_page_async(self.page, screenshot_format=self.screenshot_format)
+                # Cleanup before raising to prevent memory leak
+                await self.page.close()
+                await context.close()
+                await browser.close()
+                # Force garbage collection to release Playwright resources immediately
+                gc.collect()
                 raise Non200ErrorCodeReceived(url=url, status_code=self.status_code, screenshot=screenshot)
 
             if not empty_pages_are_a_change and len((await self.page.content()).strip()) == 0:
@@ -313,48 +333,52 @@ class fetcher(Fetcher):
                 await browser.close()
                 raise EmptyReply(url=url, status_code=response.status)
 
-            # Run Browser Steps here
-            if self.browser_steps_get_valid_steps():
-                await self.iterate_browser_steps(start_url=url)
-
-            await self.page.wait_for_timeout(extra_wait * 1000)
-
-            now = time.time()
-            # So we can find an element on the page where its selector was entered manually (maybe not xPath etc)
-            if current_include_filters is not None:
-                await self.page.evaluate("var include_filters={}".format(json.dumps(current_include_filters)))
-            else:
-                await self.page.evaluate("var include_filters=''")
-            await self.page.request_gc()
-
-            # request_gc before and after evaluate to free up memory
-            # @todo browsersteps etc
-            MAX_TOTAL_HEIGHT = int(os.getenv("SCREENSHOT_MAX_HEIGHT", SCREENSHOT_MAX_HEIGHT_DEFAULT))
-            self.xpath_data = await self.page.evaluate(XPATH_ELEMENT_JS, {
-                "visualselector_xpath_selectors": visualselector_xpath_selectors,
-                "max_height": MAX_TOTAL_HEIGHT
-            })
-            await self.page.request_gc()
-
-            self.instock_data = await self.page.evaluate(INSTOCK_DATA_JS)
-            await self.page.request_gc()
-
-            self.content = await self.page.content()
-            await self.page.request_gc()
-            logger.debug(f"Scrape xPath element data in browser done in {time.time() - now:.2f}s")
-
-
-            # Bug 3 in Playwright screenshot handling
-            # Some bug where it gives the wrong screenshot size, but making a request with the clip set first seems to solve it
-            # JPEG is better here because the screenshots can be very very large
-
-            # Screenshots also travel via the ws:// (websocket) meaning that the binary data is base64 encoded
-            # which will significantly increase the IO size between the server and client, it's recommended to use the lowest
-            # acceptable screenshot quality here
+            # Wrap remaining operations in try/finally to ensure cleanup
             try:
+                # Run Browser Steps here
+                if self.browser_steps_get_valid_steps():
+                    await self.iterate_browser_steps(start_url=url)
+
+                    await self.page.wait_for_timeout(extra_wait * 1000)
+
+                now = time.time()
+                # So we can find an element on the page where its selector was entered manually (maybe not xPath etc)
+                if current_include_filters is not None:
+                    await self.page.evaluate("var include_filters={}".format(json.dumps(current_include_filters)))
+                else:
+                    await self.page.evaluate("var include_filters=''")
+                await self.page.request_gc()
+
+                # request_gc before and after evaluate to free up memory
+                # @todo browsersteps etc
+                MAX_TOTAL_HEIGHT = int(os.getenv("SCREENSHOT_MAX_HEIGHT", SCREENSHOT_MAX_HEIGHT_DEFAULT))
+                self.xpath_data = await self.page.evaluate(XPATH_ELEMENT_JS, {
+                    "visualselector_xpath_selectors": visualselector_xpath_selectors,
+                    "max_height": MAX_TOTAL_HEIGHT
+                })
+                await self.page.request_gc()
+
+                self.instock_data = await self.page.evaluate(INSTOCK_DATA_JS)
+                await self.page.request_gc()
+
+                self.content = await self.page.content()
+                await self.page.request_gc()
+                logger.debug(f"Scrape xPath element data in browser done in {time.time() - now:.2f}s")
+
+
+                # Bug 3 in Playwright screenshot handling
+                # Some bug where it gives the wrong screenshot size, but making a request with the clip set first seems to solve it
+                # JPEG is better here because the screenshots can be very very large
+
+                # Screenshots also travel via the ws:// (websocket) meaning that the binary data is base64 encoded
+                # which will significantly increase the IO size between the server and client, it's recommended to use the lowest
+                # acceptable screenshot quality here
                 # The actual screenshot - this always base64 and needs decoding! horrible! huge CPU usage
                 self.screenshot = await capture_full_page_async(page=self.page, screenshot_format=self.screenshot_format)
 
+            except ScreenshotUnavailable:
+                # Re-raise screenshot unavailable exceptions
+                raise
             except Exception as e:
                 # It's likely the screenshot was too long/big and something crashed
                 raise ScreenshotUnavailable(url=url, status_code=self.status_code)
@@ -389,6 +413,10 @@ class fetcher(Fetcher):
                     pass
                 browser = None
 
+                # Force Python GC to release Playwright resources immediately
+                # Playwright objects can have circular references that delay cleanup
+                gc.collect()
+
 
 # Plugin registration for built-in fetcher
 class PlaywrightFetcherPlugin:

changedetectionio/content_fetchers/webdriver_selenium.py

@@ -15,7 +15,7 @@ class fetcher(Fetcher):
     proxy_url = None
 
     # Capability flags
-    supports_browser_steps = True
+    supports_browser_steps = False
     supports_screenshots = True
     supports_xpath_element_data = True
 

changedetectionio/flask_app.py

@@ -9,6 +9,7 @@ import threading
 import time
 import timeago
 from blinker import signal
+from pathlib import Path
 
 from changedetectionio.strtobool import strtobool
 from threading import Event
@@ -84,6 +85,10 @@ app.config['NEW_VERSION_AVAILABLE'] = False
 if os.getenv('FLASK_SERVER_NAME'):
     app.config['SERVER_NAME'] = os.getenv('FLASK_SERVER_NAME')
 
+# Babel/i18n configuration
+app.config['BABEL_TRANSLATION_DIRECTORIES'] = str(Path(__file__).parent / 'translations')
+app.config['BABEL_DEFAULT_LOCALE'] = 'en_GB'
+
 #app.config["EXPLAIN_TEMPLATE_LOADING"] = True
 
 
@@ -395,13 +400,9 @@ def changedetection_app(config=None, datastore_o=None):
     def get_locale():
         # 1. Try to get locale from session (user explicitly selected)
         if 'locale' in session:
-            locale = session['locale']
-            logger.trace(f"DEBUG: get_locale() returning from session: {locale}")
-            return locale
+            return session['locale']
         # 2. Fall back to Accept-Language header
-        locale = request.accept_languages.best_match(language_codes)
-        logger.trace(f"DEBUG: get_locale() returning from Accept-Language: {locale}")
-        return locale
+        return request.accept_languages.best_match(language_codes)
 
     # Initialize Babel with locale selector
     babel = Babel(app, locale_selector=get_locale)
@@ -518,9 +519,20 @@ def changedetection_app(config=None, datastore_o=None):
     @app.route('/set-language/<locale>')
     def set_language(locale):
         """Set the user's preferred language in the session"""
+        if not request.cookies:
+            logger.error("Cannot set language without session cookie")
+            flash("Cannot set language without session cookie", 'error')
+            return redirect(url_for('watchlist.index'))
+
         # Validate the locale against available languages
         if locale in language_codes:
             session['locale'] = locale
+
+            # CRITICAL: Flask-Babel caches the locale in the request context (ctx.babel_locale)
+            # We must refresh to clear this cache so the new locale takes effect immediately
+            # This is especially important for tests where multiple requests happen rapidly
+            from flask_babel import refresh
+            refresh()
         else:
             logger.error(f"Invalid locale {locale}, available: {language_codes}")
 

changedetectionio/languages.py

@@ -29,6 +29,9 @@ def get_timeago_locale(flask_locale):
     """
     locale_map = {
         'zh': 'zh_CN',      # Chinese Simplified
+        # timeago library just hasn't been updated to use the more modern locale naming convention, before BCP 47 / RFC 5646.
+        'zh_TW': 'zh_TW',   # Chinese Traditional (timeago uses zh_TW)
+        'zh_Hant_TW': 'zh_TW',  # Flask-Babel normalizes zh_TW to zh_Hant_TW, map back to timeago's zh_TW
         'pt': 'pt_PT',      # Portuguese (Portugal)
         'sv': 'sv_SE',      # Swedish
         'no': 'nb_NO',      # Norwegian Bokmål
@@ -53,7 +56,7 @@ LANGUAGE_DATA = {
     'it': {'flag': 'fi fi-it fis', 'name': 'Italiano'},
     'ja': {'flag': 'fi fi-jp fis', 'name': '日本語'},
     'zh': {'flag': 'fi fi-cn fis', 'name': '中文 (简体)'},
-    'zh_TW': {'flag': 'fi fi-tw fis', 'name': '繁體中文'},
+    'zh_Hant_TW': {'flag': 'fi fi-tw fis', 'name': '繁體中文'},
     'ru': {'flag': 'fi fi-ru fis', 'name': 'Русский'},
     'pl': {'flag': 'fi fi-pl fis', 'name': 'Polski'},
     'nl': {'flag': 'fi fi-nl fis', 'name': 'Nederlands'},

changedetectionio/model/Watch.py

@@ -18,21 +18,31 @@ BROTLI_COMPRESS_SIZE_THRESHOLD = int(os.getenv('SNAPSHOT_BROTLI_COMPRESSION_THRE
 minimum_seconds_recheck_time = int(os.getenv('MINIMUM_SECONDS_RECHECK_TIME', 3))
 mtable = {'seconds': 1, 'minutes': 60, 'hours': 3600, 'days': 86400, 'weeks': 86400 * 7}
 
-def _brotli_compress_worker(conn, filepath, mode=None):
+def _brotli_save(contents, filepath, mode=None, fallback_uncompressed=False):
     """
-    Worker function to compress data with brotli in a separate process.
-    This isolates memory - when process exits, OS reclaims all memory.
+    Save compressed data using native brotli.
+    Testing shows no memory leak when using gc.collect() after compression.
 
     Args:
-        conn: multiprocessing.Pipe connection to receive data
+        contents: data to compress (str or bytes)
         filepath: destination file path
         mode: brotli compression mode (e.g., brotli.MODE_TEXT)
+        fallback_uncompressed: if True, save uncompressed on failure; if False, raise exception
+
+    Returns:
+        str: actual filepath saved (may differ from input if fallback used)
+
+    Raises:
+        Exception: if compression fails and fallback_uncompressed is False
     """
     import brotli
+    import gc
+
+    # Ensure contents are bytes
+    if isinstance(contents, str):
+        contents = contents.encode('utf-8')
 
     try:
-        # Receive data from parent process via pipe (avoids pickle overhead)
-        contents = conn.recv()
         logger.debug(f"Starting brotli compression of {len(contents)} bytes.")
 
         if mode is not None:
@@ -43,111 +53,25 @@ def _brotli_compress_worker(conn, filepath, mode=None):
         with open(filepath, 'wb') as f:
             f.write(compressed_data)
 
-        # Send success status back
-        conn.send(True)
         logger.debug(f"Finished brotli compression - From {len(contents)} to {len(compressed_data)} bytes.")
-        # No need for explicit cleanup - process exit frees all memory
-    except Exception as e:
-        logger.critical(f"Brotli compression worker failed: {e}")
-        conn.send(False)
-    finally:
-        conn.close()
 
+        # Force garbage collection to prevent memory buildup
+        gc.collect()
 
-def _brotli_subprocess_save(contents, filepath, mode=None, timeout=30, fallback_uncompressed=False):
-    """
-    Save compressed data using subprocess to isolate memory.
-    Uses Pipe to avoid pickle overhead for large data.
-
-    Args:
-        contents: data to compress (str or bytes)
-        filepath: destination file path
-        mode: brotli compression mode (e.g., brotli.MODE_TEXT)
-        timeout: subprocess timeout in seconds
-        fallback_uncompressed: if True, save uncompressed on failure; if False, raise exception
-
-    Returns:
-        str: actual filepath saved (may differ from input if fallback used)
-
-    Raises:
-        Exception: if compression fails and fallback_uncompressed is False
-    """
-    import multiprocessing
-    import sys
-
-    # Ensure contents are bytes
-    if isinstance(contents, str):
-        contents = contents.encode('utf-8')
-
-    # Use explicit spawn context for thread safety (avoids fork() with multi-threaded parent)
-    # Always use spawn - consistent behavior in tests and production
-    ctx = multiprocessing.get_context('spawn')
-    parent_conn, child_conn = ctx.Pipe()
-
-    # Run compression in subprocess using spawn (not fork)
-    proc = ctx.Process(target=_brotli_compress_worker, args=(child_conn, filepath, mode))
-
-    # Windows-safe: Set daemon=False explicitly to avoid issues with process cleanup
-    proc.daemon = False
-    proc.start()
-
-    try:
-        # Send data to subprocess via pipe (avoids pickle)
-        parent_conn.send(contents)
-
-        # Wait for result with timeout
-        if parent_conn.poll(timeout):
-            success = parent_conn.recv()
-        else:
-            success = False
-            logger.warning(f"Brotli compression subprocess timed out after {timeout}s")
-            # Graceful termination with platform-aware cleanup
-            try:
-                proc.terminate()
-            except Exception as term_error:
-                logger.debug(f"Process termination issue (may be normal on Windows): {term_error}")
-
-        parent_conn.close()
-        proc.join(timeout=5)
-
-        # Force kill if still alive after graceful termination
-        if proc.is_alive():
-            try:
-                if sys.platform == 'win32':
-                    # Windows: use kill() which is more forceful
-                    proc.kill()
-                else:
-                    # Unix: terminate() already sent SIGTERM, now try SIGKILL
-                    proc.kill()
-                proc.join(timeout=2)
-            except Exception as kill_error:
-                logger.warning(f"Failed to kill brotli compression process: {kill_error}")
-
-        # Check if file was created successfully
-        if success and os.path.exists(filepath):
-            return filepath
+        return filepath
 
     except Exception as e:
         logger.error(f"Brotli compression error: {e}")
-        try:
-            parent_conn.close()
-        except:
-            pass
-        try:
-            proc.terminate()
-            proc.join(timeout=2)
-        except:
-            pass
 
-    # Compression failed
-    if fallback_uncompressed:
-        logger.warning(f"Brotli compression failed for {filepath}, saving uncompressed")
-        fallback_path = filepath.replace('.br', '')
-        with open(fallback_path, 'wb') as f:
-            f.write(contents)
-        return fallback_path
-    else:
-        raise Exception(f"Brotli compression subprocess failed for {filepath}")
+        # Compression failed
+        if fallback_uncompressed:
+            logger.warning(f"Brotli compression failed for {filepath}, saving uncompressed")
+            fallback_path = filepath.replace('.br', '')
+            with open(fallback_path, 'wb') as f:
+                f.write(contents)
+            return fallback_path
+        else:
+            raise Exception(f"Brotli compression failed for {filepath}: {e}")
 
 
 class model(watch_base):
@@ -523,7 +447,7 @@ class model(watch_base):
 
                 if not os.path.exists(dest):
                     try:
-                        actual_dest = _brotli_subprocess_save(contents, dest, mode=brotli.MODE_TEXT, fallback_uncompressed=True)
+                        actual_dest = _brotli_save(contents, dest, mode=brotli.MODE_TEXT, fallback_uncompressed=True)
                         if actual_dest != dest:
                             snapshot_fname = os.path.basename(actual_dest)
                     except Exception as e:
@@ -949,13 +873,13 @@ class model(watch_base):
     def save_last_text_fetched_before_filters(self, contents):
         import brotli
         filepath = os.path.join(self.watch_data_dir, 'last-fetched.br')
-        _brotli_subprocess_save(contents, filepath, mode=brotli.MODE_TEXT, fallback_uncompressed=False)
+        _brotli_save(contents, filepath, mode=brotli.MODE_TEXT, fallback_uncompressed=False)
 
     def save_last_fetched_html(self, timestamp, contents):
         self.ensure_data_dir_exists()
         snapshot_fname = f"{timestamp}.html.br"
         filepath = os.path.join(self.watch_data_dir, snapshot_fname)
-        _brotli_subprocess_save(contents, filepath, mode=None, fallback_uncompressed=True)
+        _brotli_save(contents, filepath, mode=None, fallback_uncompressed=True)
         self._prune_last_fetched_html_snapshots()
 
     def get_fetched_html(self, timestamp):

changedetectionio/processors/image_ssim_diff/image_handler/isolated_libvips.py

@@ -13,14 +13,9 @@ Research: https://github.com/libvips/pyvips/issues/234
 
 import multiprocessing
 
-# CRITICAL: Use 'spawn' instead of 'fork' to avoid inheriting parent's
+# CRITICAL: Use 'spawn' context instead of 'fork' to avoid inheriting parent's
 # LibVIPS threading state which can cause hangs in gaussblur operations
 # https://docs.python.org/3/library/multiprocessing.html#contexts-and-start-methods
-try:
-    multiprocessing.set_start_method('spawn', force=False)
-except RuntimeError:
-    # Already set, ignore
-    pass
 
 
 def _worker_generate_diff(conn, img_bytes_from, img_bytes_to, threshold, blur_sigma, max_width, max_height):
@@ -95,9 +90,10 @@ def generate_diff_isolated(img_bytes_from, img_bytes_to, threshold, blur_sigma,
     Returns:
         bytes: JPEG diff image or None on failure
     """
-    parent_conn, child_conn = multiprocessing.Pipe()
+    ctx = multiprocessing.get_context('spawn')
+    parent_conn, child_conn = ctx.Pipe()
 
-    p = multiprocessing.Process(
+    p = ctx.Process(
         target=_worker_generate_diff,
         args=(child_conn, img_bytes_from, img_bytes_to, threshold, blur_sigma, max_width, max_height)
     )
@@ -140,7 +136,8 @@ def calculate_change_percentage_isolated(img_bytes_from, img_bytes_to, threshold
     Returns:
         float: Change percentage
     """
-    parent_conn, child_conn = multiprocessing.Pipe()
+    ctx = multiprocessing.get_context('spawn')
+    parent_conn, child_conn = ctx.Pipe()
 
     def _worker_calculate(conn):
         try:
@@ -185,7 +182,7 @@ def calculate_change_percentage_isolated(img_bytes_from, img_bytes_to, threshold
         finally:
             conn.close()
 
-    p = multiprocessing.Process(target=_worker_calculate, args=(child_conn,))
+    p = ctx.Process(target=_worker_calculate, args=(child_conn,))
     p.start()
 
     result = 0.0
@@ -233,7 +230,8 @@ def compare_images_isolated(img_bytes_from, img_bytes_to, threshold, blur_sigma,
         tuple: (changed_detected, change_percentage)
     """
     print(f"[Parent] Starting compare_images_isolated subprocess", flush=True)
-    parent_conn, child_conn = multiprocessing.Pipe()
+    ctx = multiprocessing.get_context('spawn')
+    parent_conn, child_conn = ctx.Pipe()
 
     def _worker_compare(conn):
         try:
@@ -301,7 +299,7 @@ def compare_images_isolated(img_bytes_from, img_bytes_to, threshold, blur_sigma,
         finally:
             conn.close()
 
-    p = multiprocessing.Process(target=_worker_compare, args=(child_conn,))
+    p = ctx.Process(target=_worker_compare, args=(child_conn,))
     print(f"[Parent] Starting subprocess (pid will be assigned)", flush=True)
     p.start()
     print(f"[Parent] Subprocess started (pid={p.pid}), waiting for result (30s timeout)", flush=True)

changedetectionio/templates/base.html

@@ -1,5 +1,5 @@
 <!DOCTYPE html>
-<html lang="{{ get_locale() }}" data-darkmode="{{ get_darkmode_state() }}">
+<html lang="{{ get_locale()|replace('_', '-') }}" data-darkmode="{{ get_darkmode_state() }}">
 
   <head>
     <meta charset="utf-8" >

changedetectionio/tests/test_i18n.py

@@ -1,7 +1,71 @@
 #!/usr/bin/env python3
 
 from flask import url_for
-from .util import live_server_setup
+from .util import live_server_setup, wait_for_all_checks
+
+
+def test_zh_TW(client, live_server, measure_memory_usage, datastore_path):
+    import time
+    test_url = url_for('test_endpoint', _external=True)
+
+    # Be sure we got a session cookie
+    res = client.get(url_for("watchlist.index"), follow_redirects=True)
+
+    res = client.get(
+        url_for("set_language", locale="zh_Hant_TW"), # Traditional
+        follow_redirects=True
+    )
+    # HTML follows BCP 47 language tag rules, not underscore-based locale formats.
+    assert b'<html lang="zh-Hant-TW"' in res.data
+    assert b'Cannot set language without session cookie' not in res.data
+    assert '選擇語言'.encode() in res.data
+
+    # Check second set works
+    res = client.get(
+        url_for("set_language", locale="en_GB"),
+        follow_redirects=True
+    )
+    assert b'Cannot set language without session cookie' not in res.data
+    res = client.get(url_for("watchlist.index"), follow_redirects=True)
+    assert b"Select Language" in res.data, "Second set of language worked"
+
+    # Check arbitration between zh_Hant_TW<->zh
+    res = client.get(
+        url_for("set_language", locale="zh"), # Simplified chinese
+        follow_redirects=True
+    )
+    res = client.get(url_for("watchlist.index"), follow_redirects=True)
+    assert "选择语言".encode() in res.data, "Simplified chinese worked and it means the flask-babel cache worked"
+
+
+# timeago library just hasn't been updated to use the more modern locale naming convention, before BCP 47 / RFC 5646.
+# The Python timeago library (https://github.com/hustcc/timeago) supports 48 locales but uses different naming conventions than Flask-Babel.
+def test_zh_Hant_TW_timeago_integration():
+    """Test that zh_Hant_TW mapping works and timeago renders Traditional Chinese correctly"""
+    import timeago
+    from datetime import datetime, timedelta
+    from changedetectionio.languages import get_timeago_locale
+
+    # 1. Test the mapping
+    mapped_locale = get_timeago_locale('zh_Hant_TW')
+    assert mapped_locale == 'zh_TW', "zh_Hant_TW should map to timeago's zh_TW"
+    assert get_timeago_locale('zh_TW') == 'zh_TW', "zh_TW should also map to zh_TW"
+
+    # 2. Test timeago library renders Traditional Chinese with the mapped locale
+    now = datetime.now()
+
+    # Test various time periods with Traditional Chinese strings
+    result_15s = timeago.format(now - timedelta(seconds=15), now, mapped_locale)
+    assert '秒前' in result_15s, f"Expected '秒前' in '{result_15s}'"
+
+    result_5m = timeago.format(now - timedelta(minutes=5), now, mapped_locale)
+    assert '分鐘前' in result_5m, f"Expected '分鐘前' in '{result_5m}'"
+
+    result_2h = timeago.format(now - timedelta(hours=2), now, mapped_locale)
+    assert '小時前' in result_2h, f"Expected '小時前' in '{result_2h}'"
+
+    result_3d = timeago.format(now - timedelta(days=3), now, mapped_locale)
+    assert '天前' in result_3d, f"Expected '天前' in '{result_3d}'"
 
 
 def test_language_switching(client, live_server, measure_memory_usage, datastore_path):
@@ -13,6 +77,9 @@ def test_language_switching(client, live_server, measure_memory_usage, datastore
     3. Switch back to English and verify English text appears
     """
 
+    # Establish session cookie
+    client.get(url_for("watchlist.index"), follow_redirects=True)
+
     # Step 1: Set the language to Italian using the /set-language endpoint
     res = client.get(
         url_for("set_language", locale="it"),
@@ -61,6 +128,9 @@ def test_invalid_locale(client, live_server, measure_memory_usage, datastore_pat
     The app should ignore invalid locales and continue working.
     """
 
+    # Establish session cookie
+    client.get(url_for("watchlist.index"), follow_redirects=True)
+
     # First set to English
     res = client.get(
         url_for("set_language", locale="en"),
@@ -93,6 +163,9 @@ def test_language_persistence_in_session(client, live_server, measure_memory_usa
     within the same session.
     """
 
+    # Establish session cookie
+    client.get(url_for("watchlist.index"), follow_redirects=True)
+
     # Set language to Italian
     res = client.get(
         url_for("set_language", locale="it"),
@@ -119,6 +192,9 @@ def test_set_language_with_redirect(client, live_server, measure_memory_usage, d
     """
     from flask import url_for
 
+    # Establish session cookie
+    client.get(url_for("watchlist.index"), follow_redirects=True)
+
     # Set language with a redirect parameter (simulating language change from /settings)
     res = client.get(
         url_for("set_language", locale="de", redirect="/settings"),

changedetectionio/tests/visualselector/test_fetch_data.py

@@ -144,7 +144,6 @@ def test_basic_browserstep(client, live_server, measure_memory_usage, datastore_
 
 def test_non_200_errors_report_browsersteps(client, live_server, measure_memory_usage, datastore_path):
 
-
     four_o_four_url =  url_for('test_endpoint', status_code=404, _external=True)
     four_o_four_url = four_o_four_url.replace('localhost.localdomain', 'cdio')
     four_o_four_url = four_o_four_url.replace('localhost', 'cdio')
@@ -186,3 +185,65 @@ def test_non_200_errors_report_browsersteps(client, live_server, measure_memory_
         url_for("ui.form_delete", uuid="all"),
         follow_redirects=True
     )
+
+def test_browsersteps_edit_UI_startsession(client, live_server, measure_memory_usage, datastore_path):
+
+    assert os.getenv('PLAYWRIGHT_DRIVER_URL'), "Needs PLAYWRIGHT_DRIVER_URL set for this test"
+
+    # Add a watch first
+    test_url = url_for('test_interactive_html_endpoint', _external=True)
+    test_url = test_url.replace('localhost.localdomain', 'cdio')
+    test_url = test_url.replace('localhost', 'cdio')
+
+    uuid = client.application.config.get('DATASTORE').add_watch(url=test_url, extras={'fetch_backend': 'html_webdriver', 'paused': True})
+
+    # Test starting a browsersteps session
+    res = client.get(
+        url_for("browser_steps.browsersteps_start_session", uuid=uuid),
+        follow_redirects=True
+    )
+
+    assert res.status_code == 200
+    assert res.is_json
+    json_data = res.get_json()
+    assert 'browsersteps_session_id' in json_data
+    assert json_data['browsersteps_session_id']  # Not empty
+
+    browsersteps_session_id = json_data['browsersteps_session_id']
+
+    # Verify the session exists in browsersteps_sessions
+    from changedetectionio.blueprint.browser_steps import browsersteps_sessions, browsersteps_watch_to_session
+    assert browsersteps_session_id in browsersteps_sessions
+    assert uuid in browsersteps_watch_to_session
+    assert browsersteps_watch_to_session[uuid] == browsersteps_session_id
+
+    # Verify browsersteps UI shows up on edit page
+    res = client.get(url_for("ui.ui_edit.edit_page", uuid=uuid))
+    assert b'browsersteps-click-start' in res.data, "Browsersteps manual UI shows up"
+
+    # Session should still exist after GET (not cleaned up yet)
+    assert browsersteps_session_id in browsersteps_sessions
+    assert uuid in browsersteps_watch_to_session
+
+    # Test cleanup happens on save (POST)
+    res = client.post(
+        url_for("ui.ui_edit.edit_page", uuid=uuid),
+        data={
+            "url": test_url,
+            "tags": "",
+            'fetch_backend': "html_webdriver",
+            "time_between_check_use_default": "y",
+        },
+        follow_redirects=True
+    )
+    assert b"Updated watch" in res.data
+
+    # NOW verify the session was cleaned up after save
+    assert browsersteps_session_id not in browsersteps_sessions
+    assert uuid not in browsersteps_watch_to_session
+
+    # Cleanup
+    client.get(
+        url_for("ui.form_delete", uuid="all"),
+        follow_redirects=True
+    )

requirements.txt

@@ -91,7 +91,7 @@ jq~=1.3; python_version >= "3.8" and sys_platform == "linux"
 
 # playwright is installed at Dockerfile build time because it's not available on all platforms
 
-pyppeteer-ng==2.0.0rc10
+pyppeteer-ng==2.0.0rc11
 pyppeteerstealth>=0.0.4
 
 # Include pytest, so if theres a support issue we can ask them to run these tests on their setup
@@ -100,7 +100,7 @@ pytest-flask ~=1.3
 pytest-mock ~=3.15
 
 # Anything 4.0 and up but not 5.0
-jsonschema ~= 4.25
+jsonschema ~= 4.26
 
 # OpenAPI validation support
 openapi-core[flask] >= 0.19.0