Re PR #3964 refactor

.github/workflows/containers.yml

@@ -66,27 +66,27 @@ jobs:
           echo ${{ github.ref }} > changedetectionio/tag.txt
 
       - name: Set up QEMU
-        uses: docker/setup-qemu-action@v3
+        uses: docker/setup-qemu-action@v4
         with:
           image: tonistiigi/binfmt:latest
           platforms: all
 
       - name: Login to GitHub Container Registry
-        uses: docker/login-action@v3
+        uses: docker/login-action@v4
         with:
           registry: ghcr.io
           username: ${{ github.actor }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Login to Docker Hub Container Registry
-        uses: docker/login-action@v3
+        uses: docker/login-action@v4
         with:
           username: ${{ secrets.DOCKER_HUB_USERNAME }}
           password: ${{ secrets.DOCKER_HUB_ACCESS_TOKEN }}
 
       - name: Set up Docker Buildx
         id: buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@v4
         with:
           install: true
           version: latest
@@ -95,7 +95,7 @@ jobs:
       # master branch -> :dev container tag
       - name: Docker meta :dev
         if: ${{ github.ref == 'refs/heads/master' && github.event_name != 'release' }}
-        uses: docker/metadata-action@v5
+        uses: docker/metadata-action@v6
         id: meta_dev
         with:
           images: |
@@ -107,7 +107,7 @@ jobs:
       - name: Build and push :dev
         id: docker_build
         if: ${{ github.ref == 'refs/heads/master' && github.event_name != 'release' }}
-        uses: docker/build-push-action@v6
+        uses: docker/build-push-action@v7
         with:
           context: ./
           file: ./Dockerfile
@@ -131,7 +131,7 @@ jobs:
           
       - name: Docker meta :tag
         if: github.event_name == 'release' && startsWith(github.event.release.tag_name, '0.')
-        uses: docker/metadata-action@v5
+        uses: docker/metadata-action@v6
         id: meta
         with:
             images: |
@@ -146,7 +146,7 @@ jobs:
       - name: Build and push :tag
         id: docker_build_tag_release
         if: github.event_name == 'release' && startsWith(github.event.release.tag_name, '0.')
-        uses: docker/build-push-action@v6
+        uses: docker/build-push-action@v7
         with:
           context: ./
           file: ./Dockerfile

.github/workflows/test-container-build.yml

@@ -60,14 +60,14 @@ jobs:
 
         # Just test that the build works, some libraries won't compile on ARM/rPi etc
         - name: Set up QEMU
-          uses: docker/setup-qemu-action@v3
+          uses: docker/setup-qemu-action@v4
           with:
             image: tonistiigi/binfmt:latest
             platforms: all
 
         - name: Set up Docker Buildx
           id: buildx
-          uses: docker/setup-buildx-action@v3
+          uses: docker/setup-buildx-action@v4
           with:
             install: true
             version: latest
@@ -75,7 +75,7 @@ jobs:
 
         - name: Test that the docker containers can build (${{ matrix.platform }} - ${{ matrix.dockerfile }})
           id: docker_build
-          uses: docker/build-push-action@v6
+          uses: docker/build-push-action@v7
           # https://github.com/docker/build-push-action#customizing
           with:
             context: ./

.github/workflows/test-only.yml

@@ -52,4 +52,13 @@ jobs:
     uses: ./.github/workflows/test-stack-reusable-workflow.yml
     with:
       python-version: '3.13'
-      skip-pypuppeteer: true
+      skip-pypuppeteer: true
+
+
+  test-application-3-14:
+    #if: github.event_name == 'push' && github.ref == 'refs/heads/master'
+    needs: lint-code
+    uses: ./.github/workflows/test-stack-reusable-workflow.yml
+    with:
+      python-version: '3.14'
+      skip-pypuppeteer: false

.github/workflows/test-stack-reusable-workflow.yml

@@ -42,10 +42,10 @@ jobs:
         run: echo "date=$(date +'%Y-%m-%d')" >> $GITHUB_OUTPUT
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@v4
 
       - name: Build changedetection.io container for testing under Python ${{ env.PYTHON_VERSION }}
-        uses: docker/build-push-action@v6
+        uses: docker/build-push-action@v7
         with:
           context: ./
           file: ./Dockerfile
@@ -706,7 +706,19 @@ jobs:
       - name: Check upgrade works without error
         run: |
           echo "=== Testing upgrade path from 0.49.1 to ${{ github.ref_name }} (${{ github.sha }}) ==="
-
+          sudo apt-get update && sudo apt-get install -y --no-install-recommends \
+              g++ \
+              gcc \
+              libc-dev \
+              libffi-dev \
+              libjpeg-dev \
+              libssl-dev \
+              libxslt-dev \
+              make \
+              patch \
+              pkg-config \
+              zlib1g-dev
+          
           # Checkout old version and create datastore
           git checkout 0.49.1
           python3 -m venv .venv

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__ = '0.54.2'
+__version__ = '0.54.4'
 
 from changedetectionio.strtobool import strtobool
 from json.decoder import JSONDecodeError
@@ -61,8 +61,22 @@ import time
 # ==============================================================================
 
 import multiprocessing
+import os
 import sys
 
+# Limit glibc malloc arena count to prevent RSS growth from concurrent requests.
+# Default: glibc creates up to 8×CPU_cores arenas. Each concurrent thread/connection
+# can trigger a new arena, and freed memory stays mapped in those arenas as RSS forever.
+# With MALLOC_ARENA_MAX=2, at most 2 arenas are used; freed pages return to the OS faster.
+# Must be set before worker threads start; env var is read lazily by glibc on first arena creation.
+if 'MALLOC_ARENA_MAX' not in os.environ:
+    os.environ['MALLOC_ARENA_MAX'] = '2'
+    try:
+        import ctypes as _ctypes
+        _ctypes.CDLL('libc.so.6').mallopt(-8, 2)  # M_ARENA_MAX = -8
+    except Exception:
+        pass
+
 # Set spawn as global default (safety net - all our code uses explicit contexts anyway)
 # Skip in tests to avoid breaking pytest-flask's LiveServer fixture (uses unpicklable local functions)
 if 'pytest' not in sys.modules:

changedetectionio/api/Watch.py

@@ -15,6 +15,7 @@ import copy
 from . import validate_openapi_request, get_readonly_watch_fields
 from ..notification import valid_notification_formats
 from ..notification.handler import newline_re
+from ..processors.text_json_diff.difference import DIFF_PREFERENCES_CONFIG, parse_diff_preferences
 
 
 def validate_time_between_check_required(json_data):
@@ -338,22 +339,14 @@ class WatchHistoryDiff(Resource):
             word_diff = True
 
         # Get boolean diff preferences with defaults from DIFF_PREFERENCES_CONFIG
-        changes_only = strtobool(request.args.get('changesOnly', 'true'))
-        ignore_whitespace = strtobool(request.args.get('ignoreWhitespace', 'false'))
-        include_removed = strtobool(request.args.get('removed', 'true'))
-        include_added = strtobool(request.args.get('added', 'true'))
-        include_replaced = strtobool(request.args.get('replaced', 'true'))
+        prefs, render_kwargs = parse_diff_preferences(request.args)
 
         # Generate the diff with all preferences
         content = diff.render_diff(
             previous_version_file_contents=from_version_file_contents,
             newest_version_file_contents=to_version_file_contents,
-            ignore_junk=ignore_whitespace,
-            include_equal=changes_only,
-            include_removed=include_removed,
-            include_added=include_added,
-            include_replaced=include_replaced,
             word_diff=word_diff,
+            **render_kwargs,
         )
 
         # Skip formatting if no_markup is set

changedetectionio/blueprint/backups/__init__.py

@@ -40,11 +40,6 @@ def create_backup(datastore_path, watches: dict, tags: dict = None):
             zipObj.write(url_watches_json, arcname="url-watches.json")
             logger.debug("Added url-watches.json to backup")
 
-        # Add the flask app secret (if it exists)
-        secret_file = os.path.join(datastore_path, "secret.txt")
-        if os.path.isfile(secret_file):
-            zipObj.write(secret_file, arcname="secret.txt")
-
         # Add tag data directories (each tag has its own {uuid}/tag.json)
         for uuid, tag in (tags or {}).items():
             for f in Path(tag.data_dir).glob('*'):
@@ -151,19 +146,22 @@ def construct_blueprint(datastore: ChangeDetectionStore):
     def download_backup(filename):
         import re
         filename = filename.strip()
-        backup_filename_regex = BACKUP_FILENAME_FORMAT.format("\d+")
-
-        full_path = os.path.join(os.path.abspath(datastore.datastore_path), filename)
-        if not full_path.startswith(os.path.abspath(datastore.datastore_path)):
-            abort(404)
+        backup_filename_regex = BACKUP_FILENAME_FORMAT.format(r"\d+")
 
+        # Resolve 'latest' before any validation so checks run against the real filename.
         if filename == 'latest':
             backups = find_backups()
+            if not backups:
+                abort(404)
             filename = backups[0]['filename']
 
         if not re.match(r"^" + backup_filename_regex + "$", filename):
             abort(400)  # Bad Request if the filename doesn't match the pattern
 
+        full_path = os.path.join(os.path.abspath(datastore.datastore_path), filename)
+        if not full_path.startswith(os.path.abspath(datastore.datastore_path) + os.sep):
+            abort(404)
+
         logger.debug(f"Backup download request for '{full_path}'")
         return send_from_directory(os.path.abspath(datastore.datastore_path), filename, as_attachment=True)
 

changedetectionio/blueprint/backups/restore.py

@@ -1,6 +1,7 @@
 import io
 import json
 import os
+import re
 import shutil
 import tempfile
 import threading
@@ -14,6 +15,16 @@ from loguru import logger
 
 from changedetectionio.flask_app import login_optionally_required
 
+# Maximum size of the uploaded zip file. Override via env var MAX_RESTORE_UPLOAD_MB.
+_MAX_UPLOAD_BYTES = int(os.getenv("MAX_RESTORE_UPLOAD_MB", 256)) * 1024 * 1024
+# Maximum total uncompressed size of all entries (zip-bomb guard). Override via MAX_RESTORE_DECOMPRESSED_MB.
+_MAX_DECOMPRESSED_BYTES = int(os.getenv("MAX_RESTORE_DECOMPRESSED_MB", 1024)) * 1024 * 1024
+# Only top-level directories whose name is a valid UUID are treated as watch/tag entries.
+_UUID_RE = re.compile(
+    r'^[0-9a-f]{8}-[0-9a-f]{4}-4[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$',
+    re.IGNORECASE,
+)
+
 
 class RestoreForm(Form):
     zip_file = FileField(_l('Backup zip file'), validators=[
@@ -50,7 +61,18 @@ def import_from_zip(zip_stream, datastore, include_groups, include_groups_replac
     with tempfile.TemporaryDirectory() as tmpdir:
         logger.debug(f"Restore: extracting zip to {tmpdir}")
         with zipfile.ZipFile(zip_stream, 'r') as zf:
-            zf.extractall(tmpdir)
+            total_uncompressed = sum(m.file_size for m in zf.infolist())
+            if total_uncompressed > _MAX_DECOMPRESSED_BYTES:
+                raise ValueError(
+                    f"Backup archive decompressed size ({total_uncompressed // (1024 * 1024)} MB) "
+                    f"exceeds the {_MAX_DECOMPRESSED_BYTES // (1024 * 1024)} MB limit"
+                )
+            resolved_dest = os.path.realpath(tmpdir)
+            for member in zf.infolist():
+                member_dest = os.path.realpath(os.path.join(resolved_dest, member.filename))
+                if not member_dest.startswith(resolved_dest + os.sep) and member_dest != resolved_dest:
+                    raise ValueError(f"Zip Slip path traversal detected in backup archive: {member.filename!r}")
+                zf.extract(member, tmpdir)
         logger.debug("Restore: zip extracted, scanning UUID directories")
 
         for entry in os.scandir(tmpdir):
@@ -58,6 +80,9 @@ def import_from_zip(zip_stream, datastore, include_groups, include_groups_replac
                 continue
 
             uuid = entry.name
+            if not _UUID_RE.match(uuid):
+                logger.warning(f"Restore: skipping non-UUID directory {uuid!r}")
+                continue
             tag_json_path = os.path.join(entry.path, 'tag.json')
             watch_json_path = os.path.join(entry.path, 'watch.json')
 
@@ -155,7 +180,9 @@ def construct_restore_blueprint(datastore):
         form = RestoreForm()
         return render_template("backup_restore.html",
                                form=form,
-                               restore_running=any(t.is_alive() for t in restore_threads))
+                               restore_running=any(t.is_alive() for t in restore_threads),
+                               max_upload_mb=_MAX_UPLOAD_BYTES // (1024 * 1024),
+                               max_decompressed_mb=_MAX_DECOMPRESSED_BYTES // (1024 * 1024))
 
     @login_optionally_required
     @restore_blueprint.route("/restore/start", methods=['POST'])
@@ -173,10 +200,22 @@ def construct_restore_blueprint(datastore):
             flash(gettext("File must be a .zip backup file"), "error")
             return redirect(url_for('backups.restore.restore'))
 
-        # Read into memory now — the request stream is gone once we return
+        # Reject oversized uploads before reading the stream into memory.
+        content_length = request.content_length
+        if content_length and content_length > _MAX_UPLOAD_BYTES:
+            flash(gettext("Backup file is too large (max %(mb)s MB)", mb=_MAX_UPLOAD_BYTES // (1024 * 1024)), "error")
+            return redirect(url_for('backups.restore.restore'))
+
+        # Read into memory now — the request stream is gone once we return.
+        # Read one byte beyond the limit so we can detect truncated-but-still-oversized streams.
         try:
-            zip_bytes = io.BytesIO(zip_file.read())
-            zipfile.ZipFile(zip_bytes)  # quick validity check before spawning
+            raw = zip_file.read(_MAX_UPLOAD_BYTES + 1)
+            if len(raw) > _MAX_UPLOAD_BYTES:
+                flash(gettext("Backup file is too large (max %(mb)s MB)", mb=_MAX_UPLOAD_BYTES // (1024 * 1024)), "error")
+                return redirect(url_for('backups.restore.restore'))
+            zip_bytes = io.BytesIO(raw)
+            with zipfile.ZipFile(zip_bytes):  # quick validity check before spawning
+                pass
             zip_bytes.seek(0)
         except zipfile.BadZipFile:
             flash(gettext("Invalid or corrupted zip file"), "error")
@@ -201,6 +240,7 @@ def construct_restore_blueprint(datastore):
             name="BackupRestore"
         )
         restore_thread.start()
+        restore_threads[:] = [t for t in restore_threads if t.is_alive()]
         restore_threads.append(restore_thread)
         flash(gettext("Restore started in background, check back in a few minutes."))
         return redirect(url_for('backups.restore.restore'))

changedetectionio/blueprint/backups/templates/backup_restore.html

@@ -19,6 +19,10 @@
 
                 <p>{{ _('Restore a backup. Must be a .zip backup file created on/after v0.53.1 (new database layout).') }}</p>
                 <p>{{ _('Note: This does not override the main application settings, only watches and groups.') }}</p>
+                <p class="pure-form-message">
+                    {{ _('Max upload size: %(upload)s MB &nbsp;·&nbsp; Max decompressed size: %(decomp)s MB',
+                         upload=max_upload_mb, decomp=max_decompressed_mb) }}
+                </p>
 
                 <form class="pure-form pure-form-stacked settings"
                       action="{{ url_for('backups.restore.backups_restore_start') }}"

changedetectionio/blueprint/rss/tag.py

@@ -7,7 +7,7 @@ def construct_tag_routes(rss_blueprint, datastore):
         datastore: The ChangeDetectionStore instance
     """
 
-    @rss_blueprint.route("/tag/<string:tag_uuid>", methods=['GET'])
+    @rss_blueprint.route("/tag/<uuid_str:tag_uuid>", methods=['GET'])
     def rss_tag_feed(tag_uuid):
 
         from flask import make_response, request, url_for

changedetectionio/blueprint/ui/__init__.py

@@ -156,9 +156,9 @@ def construct_blueprint(datastore: ChangeDetectionStore, update_q, worker_pool,
     @login_optionally_required
     def clear_all_history():
         if request.method == 'POST':
-            confirmtext = request.form.get('confirmtext')
+            confirmtext = request.form.get('confirmtext', '')
 
-            if confirmtext == 'clear':
+            if confirmtext.strip().lower() == gettext('clear').strip().lower():
                 # Run in background thread to avoid blocking
                 def clear_history_background():
                     # Capture UUIDs first to avoid race conditions

changedetectionio/blueprint/watchlist/__init__.py

@@ -81,6 +81,7 @@ def construct_blueprint(datastore: ChangeDetectionStore, update_q, queuedWatchMe
 
         sorted_tags = sorted(datastore.data['settings']['application'].get('tags').items(), key=lambda x: x[1]['title'])
 
+        proxy_list = datastore.proxy_list
         output = render_template(
             "watch-overview.html",
             active_tag=active_tag,
@@ -92,7 +93,7 @@ def construct_blueprint(datastore: ChangeDetectionStore, update_q, queuedWatchMe
             form=form,
             generate_tag_colors=processors.generate_processor_badge_colors,
             guid=datastore.data['app_guid'],
-            has_proxies=datastore.proxy_list,
+            has_proxies=proxy_list,
             hosted_sticky=os.getenv("SALTED_PASS", False) == False,
             now_time_server=round(time.time()),
             pagination=pagination,
@@ -110,6 +111,16 @@ def construct_blueprint(datastore: ChangeDetectionStore, update_q, queuedWatchMe
             watches=sorted_watches
         )
 
+        # Return freed template-building memory to the OS immediately.
+        # render_template allocates ~20MB of intermediate strings that are freed on return,
+        # but glibc keeps those pages mapped in its arenas as RSS. malloc_trim() forces
+        # glibc to release them, preventing RSS growth from concurrent Chrome connections.
+        try:
+            import ctypes
+            ctypes.CDLL('libc.so.6').malloc_trim(0)
+        except Exception:
+            pass
+
         if session.get('share-link'):
             del (session['share-link'])
 

changedetectionio/blueprint/watchlist/templates/watch-overview.html

@@ -213,12 +213,13 @@ html[data-darkmode="true"] .watch-tag-list.tag-{{ class_name }} {
                 {%- set checking_now = is_checking_now(watch) -%}
                 {%- set history_n = watch.history_n -%}
                 {%- set favicon = watch.get_favicon_filename() -%}
+                {%- set error_texts = watch.compile_error_texts(has_proxies=has_proxies) -%}
                 {%- set system_use_url_watchlist = datastore.data['settings']['application']['ui'].get('use_page_title_in_list')  -%}
                 {#  Class settings mirrored in changedetectionio/static/js/realtime.js for the frontend #}
                 {%- set row_classes = [
                     loop.cycle('pure-table-odd', 'pure-table-even'),
                     'processor-' ~ watch['processor'],
-                    'has-error' if watch.compile_error_texts()|length > 2 else '',
+                    'has-error' if error_texts|length > 2 else '',
                     'paused' if watch.paused is defined and watch.paused != False else '',
                     'unviewed' if watch.has_unviewed else '',
                     'has-restock-info' if watch.has_restock_info else 'no-restock-info',
@@ -271,7 +272,7 @@ html[data-darkmode="true"] .watch-tag-list.tag-{{ class_name }} {
                                 {% endif %}
                                <a class="external" target="_blank" rel="noopener" href="{{ watch.link.replace('source:','') }}">&nbsp;</a>
                             </span>
-                            <div class="error-text" style="display:none;">{{ watch.compile_error_texts(has_proxies=datastore.proxy_list)|safe }}</div>
+                            <div class="error-text" style="display:none;">{{ error_texts|safe }}</div>
                             {%- if watch['processor'] == 'text_json_diff'  -%}
                                 {%- if watch['has_ldjson_price_data'] and not watch['track_ldjson_price_data']  -%}
                                 <div class="ldjson-price-track-offer">Switch to Restock & Price watch mode? <a href="{{url_for('price_data_follower.accept', uuid=watch.uuid)}}" class="pure-button button-xsmall">Yes</a> <a href="{{url_for('price_data_follower.reject', uuid=watch.uuid)}}" class="">No</a></div>
@@ -305,12 +306,20 @@ html[data-darkmode="true"] .watch-tag-list.tag-{{ class_name }} {
                         {%- endif -%}
 
                         {%- if watch.get('restock') and watch['restock'].get('price') -%}
-                                {%- if watch['restock']['price'] is number -%}
-                                <span class="restock-label price" title="{{ _('Price') }}">
-                                {{ watch['restock']['price']|format_number_locale if watch['restock'].get('price') else '' }} {{ watch['restock'].get('currency','') }}
-                                </span>
-                                {%- else -%} <!-- watch['restock']['price']' is not a number, cant output it -->
+                            {%- set restock = watch['restock'] -%}
+                            {%- set price = restock.get('price') -%}
+                            {%- set cur = restock.get('currency','') -%}
+
+                            {%- if price is not none and (price|string)|regex_search('\d') -%}
+                              <span class="restock-label price" title="{{ _('Price') }}">
+                              {# @todo: make parse_currency/parse_decimal aware of the locale of the actual web page and use that instead changedetectionio/processors/restock_diff/__init__.py #}
+                                {%- if price is number -%}{# It's a number so we can convert it to their locale' #}
+                                  {{ price|format_number_locale }} {{ cur }}<!-- as number -->
+                                {%- else -%}{# It's totally fine if it arrives as something else, the website might be something weird in this field #}
+                                  {{ price }} {{ cur }}<!-- as string -->
                                 {%- endif -%}
+                              </span>
+                            {%- endif -%}
                         {%- elif not watch.has_restock_info -%}
                             <span class="restock-label error">{{ _('No information') }}</span>
                         {%- endif -%}

changedetectionio/content_fetchers/requests.py

@@ -148,10 +148,32 @@ class fetcher(Fetcher):
                         # Default to UTF-8 for XML if no encoding found
                         r.encoding = 'utf-8'
                 else:
-                    # For other content types, use chardet
-                    encoding = chardet.detect(r.content)['encoding']
-                    if encoding:
-                        r.encoding = encoding
+                    # No charset in HTTP header - sniff encoding in priority order matching browsers
+                    # (WHATWG encoding sniffing algorithm):
+                    # 1. BOM - highest confidence, check before anything else
+                    # 2. <meta charset> in first 2kb
+                    # 3. chardet statistical detection - last resort
+                    # See: https://github.com/dgtlmoon/changedetection.io/issues/3952
+                    boms = [
+                        (b'\xef\xbb\xbf', 'utf-8-sig'),
+                        (b'\xff\xfe', 'utf-16-le'),
+                        (b'\xfe\xff', 'utf-16-be'),
+                    ]
+                    bom_encoding = next((enc for bom, enc in boms if r.content.startswith(bom)), None)
+                    if bom_encoding:
+                        logger.info(f"URL: {url} Using encoding '{bom_encoding}' detected from BOM")
+                        r.encoding = bom_encoding
+                    else:
+                        meta_charset_match = re.search(rb'<meta[^>]+charset\s*=\s*["\']?\s*([^"\'\s;>]+)', r.content[:2000], re.IGNORECASE)
+                        if meta_charset_match:
+                            encoding = meta_charset_match.group(1).decode('ascii', errors='ignore')
+                            logger.info(f"URL: {url} No content-type encoding in HTTP headers - Using encoding '{encoding}' from HTML meta charset tag")
+                            r.encoding = encoding
+                        else:
+                            encoding = chardet.detect(r.content)['encoding']
+                            logger.warning(f"URL: {url} No charset in headers or meta tag, guessed encoding as '{encoding}' via chardet")
+                            if encoding:
+                                r.encoding = encoding
 
         self.headers = r.headers
 

changedetectionio/flask_app.py

@@ -4,6 +4,7 @@ import flask_login
 import locale
 import os
 import queue
+import re
 import sys
 import threading
 import time
@@ -217,9 +218,13 @@ def _jinja2_filter_format_number_locale(value: float) -> str:
     "Formats for example 4000.10 to the local locale default of 4,000.10"
     # Format the number with two decimal places (locale format string will return 6 decimal)
     formatted_value = locale.format_string("%.2f", value, grouping=True)
-
     return formatted_value
 
+@app.template_filter('regex_search')
+def _jinja2_filter_regex_search(value, pattern):
+    import re
+    return re.search(pattern, str(value)) is not None
+
 @app.template_global('is_checking_now')
 def _watch_is_checking_now(watch_obj, format="%Y-%m-%d %H:%M:%S"):
     return worker_pool.is_watch_running(watch_obj['uuid'])
@@ -383,6 +388,8 @@ def _jinja2_filter_fetcher_status_icons(fetcher_name):
 
     return ''
 
+_RE_SANITIZE_TAG = re.compile(r'[^a-zA-Z0-9]')
+
 @app.template_filter('sanitize_tag_class')
 def _jinja2_filter_sanitize_tag_class(tag_title):
     """Sanitize a tag title to create a valid CSS class name.
@@ -394,9 +401,8 @@ def _jinja2_filter_sanitize_tag_class(tag_title):
     Returns:
         str: A sanitized string suitable for use as a CSS class name
     """
-    import re
     # Remove all non-alphanumeric characters and convert to lowercase
-    sanitized = re.sub(r'[^a-zA-Z0-9]', '', tag_title).lower()
+    sanitized = _RE_SANITIZE_TAG.sub('', tag_title).lower()
     # Ensure it starts with a letter (CSS requirement)
     if sanitized and not sanitized[0].isalpha():
         sanitized = 'tag' + sanitized
@@ -484,28 +490,21 @@ def changedetection_app(config=None, datastore_o=None):
     available_languages = get_available_languages()
     language_codes = get_language_codes()
 
-    def get_locale():
-        # Locale aliases: map browser language codes to translation directory names
-        # This handles cases where browsers send standard codes (e.g., zh-TW)
-        # but our translations use more specific codes (e.g., zh_Hant_TW)
-        locale_aliases = {
-            'zh-TW': 'zh_Hant_TW',  # Traditional Chinese: browser sends zh-TW, we use zh_Hant_TW
-            'zh_TW': 'zh_Hant_TW',  # Also handle underscore variant
-        }
+    _locale_aliases = {
+        'zh-TW': 'zh_Hant_TW',  # Traditional Chinese: browser sends zh-TW, we use zh_Hant_TW
+        'zh_TW': 'zh_Hant_TW',  # Also handle underscore variant
+    }
+    _locale_match_list = language_codes + list(_locale_aliases.keys())
 
+    def get_locale():
         # 1. Try to get locale from session (user explicitly selected)
         if 'locale' in session:
             return session['locale']
 
         # 2. Fall back to Accept-Language header
-        # Get the best match from browser's Accept-Language header
-        browser_locale = request.accept_languages.best_match(language_codes + list(locale_aliases.keys()))
-
-        # 3. Check if we need to map the browser locale to our internal locale
-        if browser_locale in locale_aliases:
-            return locale_aliases[browser_locale]
-
-        return browser_locale
+        browser_locale = request.accept_languages.best_match(_locale_match_list)
+        # 3. Map browser locale to our internal locale if needed
+        return _locale_aliases.get(browser_locale, browser_locale)
 
     # Initialize Babel with locale selector
     babel = Babel(app, locale_selector=get_locale)
@@ -1018,15 +1017,16 @@ def check_for_new_version():
     import urllib3
     urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
 
+    session = requests.Session()
+    session.verify = False
+
     while not app.config.exit.is_set():
         try:
-            r = requests.post("https://changedetection.io/check-ver.php",
+            r = session.post("https://changedetection.io/check-ver.php",
                               data={'version': __version__,
                                     'app_guid': datastore.data['app_guid'],
                                     'watch_count': len(datastore.data['watching'])
-                                    },
-
-                              verify=False)
+                                    })
         except:
             pass
 

changedetectionio/forms.py

@@ -608,13 +608,12 @@ class ValidateCSSJSONXPATHInput(object):
                     raise ValidationError("XPath not permitted in this field!")
                 from lxml import etree, html
                 import elementpath
-                # xpath 2.0-3.1
-                from elementpath.xpath3 import XPath3Parser
+                from changedetectionio.html_tools import SafeXPath3Parser
                 tree = html.fromstring("<html></html>")
                 line = line.replace('xpath:', '')
 
                 try:
-                    elementpath.select(tree, line.strip(), parser=XPath3Parser)
+                    elementpath.select(tree, line.strip(), parser=SafeXPath3Parser)
                 except elementpath.ElementPathError as e:
                     message = field.gettext('\'%s\' is not a valid XPath expression. (%s)')
                     raise ValidationError(message % (line, str(e)))

changedetectionio/html_tools.py

@@ -23,6 +23,53 @@ class JSONNotFound(ValueError):
     def __init__(self, msg):
         ValueError.__init__(self, msg)
 
+
+_DEFAULT_UNSAFE_XPATH3_FUNCTIONS = [
+    'unparsed-text',
+    'unparsed-text-lines',
+    'unparsed-text-available',
+    'doc',
+    'doc-available',
+    'environment-variable',
+    'available-environment-variables',
+]
+
+
+def _build_safe_xpath3_parser():
+    """Return an XPath3Parser subclass with filesystem/environment access functions removed.
+
+    XPath 3.0 includes functions that can read arbitrary files or environment variables:
+      - unparsed-text / unparsed-text-lines / unparsed-text-available  (file read)
+      - doc / doc-available                                             (XML fetch from URI)
+      - environment-variable / available-environment-variables         (env var leakage)
+
+    Subclassing gives us an independent symbol_table copy (not shared with the parent class),
+    so removing entries here does not affect XPath3Parser itself.
+
+    Override the blocked list via the XPATH_BLOCKED_FUNCTIONS environment variable
+    (comma-separated, e.g. "unparsed-text,doc,environment-variable").
+    """
+    import os
+    from elementpath.xpath3 import XPath3Parser
+
+    class SafeXPath3Parser(XPath3Parser):
+        pass
+
+    env_override = os.getenv('XPATH_BLOCKED_FUNCTIONS')
+    if env_override is not None:
+        blocked = [f.strip() for f in env_override.split(',') if f.strip()]
+    else:
+        blocked = _DEFAULT_UNSAFE_XPATH3_FUNCTIONS
+
+    for _fn in blocked:
+        SafeXPath3Parser.symbol_table.pop(_fn, None)
+
+    return SafeXPath3Parser
+
+
+# Module-level singleton — built once, reused everywhere.
+SafeXPath3Parser = _build_safe_xpath3_parser()
+
 # Doesn't look like python supports forward slash auto enclosure in re.findall
 # So convert it to inline flag "(?i)foobar" type configuration
 @lru_cache(maxsize=100)
@@ -183,8 +230,6 @@ def xpath_filter(xpath_filter, html_content, append_pretty_line_formatting=False
     """
     from lxml import etree, html
     import elementpath
-    # xpath 2.0-3.1
-    from elementpath.xpath3 import XPath3Parser
 
     parser = etree.HTMLParser()
     tree = None
@@ -210,7 +255,7 @@ def xpath_filter(xpath_filter, html_content, append_pretty_line_formatting=False
             # This allows //title to match elements in the default namespace
             namespaces[''] = tree.nsmap[None]
 
-        r = elementpath.select(tree, xpath_filter.strip(), namespaces=namespaces, parser=XPath3Parser)
+        r = elementpath.select(tree, xpath_filter.strip(), namespaces=namespaces, parser=SafeXPath3Parser)
         #@note: //title/text() now works with default namespaces (fixed by registering '' prefix)
         #@note: //title/text() wont work where <title>CDATA.. (use cdata_in_document_to_text first)
 
@@ -235,6 +280,9 @@ def xpath_filter(xpath_filter, html_content, append_pretty_line_formatting=False
             else:
                 html_block += elementpath_tostring(element)
 
+        # Drop element references before the finally block so tree.clear() can release
+        # the libxml2 document immediately (elements pin the C-level doc via refcount).
+        del r
         return html_block
     finally:
         # Explicitly clear the tree to free memory

changedetectionio/model/Watch.py

@@ -43,6 +43,11 @@ from ..html_tools import TRANSLATE_WHITESPACE_TABLE
 FAVICON_RESAVE_THRESHOLD_SECONDS=86400
 BROTLI_COMPRESS_SIZE_THRESHOLD = int(os.getenv('SNAPSHOT_BROTLI_COMPRESSION_THRESHOLD', 1024*20))
 
+# Module-level favicon filename cache: data_dir → basename (or None)
+# Keyed by data_dir so it survives Watch object recreation, deepcopy, and concurrent requests.
+# Invalidated explicitly in bump_favicon() when a new favicon is saved.
+_FAVICON_FILENAME_CACHE: dict = {}
+
 minimum_seconds_recheck_time = int(os.getenv('MINIMUM_SECONDS_RECHECK_TIME', 3))
 mtable = {'seconds': 1, 'minutes': 60, 'hours': 3600, 'days': 86400, 'weeks': 86400 * 7}
 
@@ -806,9 +811,8 @@ class model(EntityPersistenceMixin, watch_base):
                     with open(fname, 'wb') as f:
                         f.write(decoded)
 
-                    # Invalidate favicon filename cache
-                    if hasattr(self, '_favicon_filename_cache'):
-                        delattr(self, '_favicon_filename_cache')
+                    # Invalidate module-level favicon filename cache for this watch
+                    _FAVICON_FILENAME_CACHE.pop(self.data_dir, None)
 
                     # A signal that could trigger the socket server to update the browser also
                     watch_check_update = signal('watch_favicon_bump')
@@ -823,35 +827,23 @@ class model(EntityPersistenceMixin, watch_base):
 
     def get_favicon_filename(self) -> str | None:
         """
-        Find any favicon.* file in the current working directory
-        and return the contents of the newest one.
+        Find any favicon.* file in the watch data directory.
 
-        MEMORY LEAK FIX: Cache the result to avoid repeated glob.glob() operations.
-        glob.glob() causes millions of fnmatch allocations when called for every watch on page load.
+        Uses a module-level cache keyed by data_dir to survive Watch object recreation,
+        deepcopy (which drops instance attrs), and concurrent request races.
+        Invalidated by bump_favicon() when a new favicon is saved.
 
         Returns:
-            str: Basename of the newest favicon file, or None if not found.
+            str: Basename of the favicon file, or None if not found.
         """
-        # Check cache first (prevents 26M+ allocations from repeated glob operations)
-        cache_key = '_favicon_filename_cache'
-        if hasattr(self, cache_key):
-            return getattr(self, cache_key)
+        if self.data_dir in _FAVICON_FILENAME_CACHE:
+            return _FAVICON_FILENAME_CACHE[self.data_dir]
 
         import glob
-
-        # Search for all favicon.* files
         files = glob.glob(os.path.join(self.data_dir, "favicon.*"))
-
-        if not files:
-            result = None
-        else:
-            # Find the newest by modification time
-            newest_file = max(files, key=os.path.getmtime)
-            result = os.path.basename(newest_file)
-
-        # Cache the result
-        setattr(self, cache_key, result)
-        return result
+        fname = os.path.basename(files[0]) if files else None
+        _FAVICON_FILENAME_CACHE[self.data_dir] = fname
+        return fname
 
     def get_screenshot_as_thumbnail(self, max_age=3200):
         """Return path to a square thumbnail of the most recent screenshot.
@@ -1182,18 +1174,13 @@ class model(EntityPersistenceMixin, watch_base):
     def compile_error_texts(self, has_proxies=None):
         """Compile error texts for this watch.
         Accepts has_proxies parameter to ensure it works even outside app context"""
-        from flask import url_for
+        from flask import url_for, has_request_context
         from markupsafe import Markup
 
         output = []  # Initialize as list since we're using append
         last_error = self.get('last_error','')
 
-        try:
-            url_for('settings.settings_page')
-        except Exception as e:
-            has_app_context = False
-        else:
-            has_app_context = True
+        has_app_context = has_request_context()
 
         # has app+request context, we can use url_for()
         if has_app_context:

changedetectionio/processors/base.py

@@ -260,6 +260,16 @@ class difference_detection_processor():
         # @todo .quit here could go on close object, so we can run JS if change-detected
         await self.fetcher.quit(watch=self.watch)
 
+        # Sanitize lone surrogates - these can appear when servers return malformed/mixed-encoding
+        # content that gets decoded into surrogate characters (e.g. \udcad). Without this,
+        # encode('utf-8') raises UnicodeEncodeError downstream in checksums, diffs, file writes, etc.
+        # Covers all fetchers (requests, playwright, puppeteer, selenium) in one place.
+        # Also note: By this point we SHOULD know the original encoding so it can safely convert to utf-8 for the rest of the app.
+        # See: https://github.com/dgtlmoon/changedetection.io/issues/3952
+
+        if self.fetcher.content and isinstance(self.fetcher.content, str):
+            self.fetcher.content = self.fetcher.content.encode('utf-8', errors='replace').decode('utf-8')
+
         # After init, call run_changedetection() which will do the actual change-detection
 
     def get_extra_watch_config(self, filename):

changedetectionio/processors/restock_diff/__init__.py

@@ -31,6 +31,7 @@ class Restock(dict):
 
         if standardized_value:
             # Convert to float
+            # @todo locale needs to be the locale of the webpage
             return float(parse_decimal(standardized_value, locale='en'))
 
         return None

changedetectionio/processors/restock_diff/processor.py

@@ -437,17 +437,18 @@ class perform_site_check(difference_detection_processor):
 
         # Only try to process restock information (like scraping for keywords) if the page was actually rendered correctly.
         # Otherwise it will assume "in stock" because nothing suggesting the opposite was found
-        from ...html_tools import html_to_text
-        text = html_to_text(self.fetcher.content)
-        logger.debug(f"Length of text after conversion: {len(text)}")
-        if not len(text):
-            from ...content_fetchers.exceptions import ReplyWithContentButNoText
-            raise ReplyWithContentButNoText(url=watch.link,
-                                            status_code=self.fetcher.get_last_status_code(),
-                                            screenshot=self.fetcher.screenshot,
-                                            html_content=self.fetcher.content,
-                                            xpath_data=self.fetcher.xpath_data
-                                            )
+#useless
+#        from ...html_tools import html_to_text
+#        text = html_to_text(self.fetcher.content)
+#        logger.debug(f"Length of text after conversion: {len(text)}")
+#        if not len(text):
+#            from ...content_fetchers.exceptions import ReplyWithContentButNoText
+#            raise ReplyWithContentButNoText(url=watch.link,
+#                                            status_code=self.fetcher.get_last_status_code(),
+#                                            screenshot=self.fetcher.screenshot,
+#                                            html_content=self.fetcher.content,
+#                                            xpath_data=self.fetcher.xpath_data
+#                                            )
 
         # Which restock settings to compare against?
         # Settings are stored in restock_diff.json (migrated from watch.json by update_30).

changedetectionio/processors/restock_diff/pure_python_extractor.py

@@ -283,4 +283,7 @@ def query_price_availability(extracted_data):
     if not result.get('availability') and 'availability' in microdata:
         result['availability'] = microdata['availability']
 
+    # result['price'] could be float or str here, depending on the website, for example it might contain "1,00" commas, etc.
+    # using something like babel you need to know the locale of the website and even then it can be problematic
+    # we dont really do anything with the price data so far.. so just accept it the way it comes.
     return result

changedetectionio/processors/text_json_diff/difference.py

@@ -99,6 +99,26 @@ DIFF_PREFERENCES_CONFIG = {
     'type': {'default': 'diffLines', 'type': 'value'},
 }
 
+def parse_diff_preferences(args, checkbox_mode=False):
+    parsed = {}
+    render_kwargs = {}
+
+    for query_name, config in DIFF_PREFERENCES_CONFIG.items():
+        value = args.get(query_name, config['default'])
+
+        if config['type'] == 'bool':
+            if checkbox_mode:
+                value = strtobool(args.get(query_name, 'off'))
+            else:
+                value = strtobool(value)
+
+        parsed[query_name] = value
+
+        if 'render_arg' in config:
+            render_kwargs[config['render_arg']] = value
+
+    return parsed, render_kwargs
+
 def render(watch, datastore, request, url_for, render_template, flash, redirect, extract_form=None):
     """
     Render the history/diff view for text/JSON/HTML changes.
@@ -166,30 +186,12 @@ def render(watch, datastore, request, url_for, render_template, flash, redirect,
 
     datastore.set_last_viewed(uuid, time.time())
 
-    # Parse diff preferences from request using config as single source of truth
-    # Check if this is a user submission (any diff pref param exists in query string)
-    user_submitted = any(key in request.args for key in DIFF_PREFERENCES_CONFIG.keys())
-
-    diff_prefs = {}
-    for key, config in DIFF_PREFERENCES_CONFIG.items():
-        if user_submitted:
-            # User submitted form - missing checkboxes are explicitly OFF
-            if config['type'] == 'bool':
-                diff_prefs[key] = strtobool(request.args.get(key, 'off'))
-            else:
-                diff_prefs[key] = request.args.get(key, config['default'])
-        else:
-            # Initial load - use defaults from config
-            diff_prefs[key] = config['default']
+    diff_prefs, render_kwargs = parse_diff_preferences(request.args)
 
     content = diff.render_diff(previous_version_file_contents=from_version_file_contents,
                                newest_version_file_contents=to_version_file_contents,
-                               include_replaced=diff_prefs['replaced'],
-                               include_added=diff_prefs['added'],
-                               include_removed=diff_prefs['removed'],
-                               include_equal=diff_prefs['changesOnly'],
-                               ignore_junk=diff_prefs['ignoreWhitespace'],
                                word_diff=diff_prefs['type'] == 'diffWords',
+                               **render_kwargs,
                                )
 
     # Build cell grid visualizer before applying HTML color (so we can detect placemarkers)

changedetectionio/tests/conftest.py

@@ -2,6 +2,7 @@
 import psutil
 import time
 from threading import Thread
+import multiprocessing
 
 import pytest
 import arrow
@@ -191,6 +192,34 @@ def cleanup(datastore_path):
             if os.path.isfile(f):
                 os.unlink(f)
 
+def pytest_configure(config):
+    """Configure pytest environment before tests run.
+
+    CRITICAL: Set multiprocessing start method to 'fork' for Python 3.14+ compatibility.
+
+    Python 3.14 changed the default start method from 'fork' to 'forkserver' on Linux.
+    The forkserver method requires all objects to be picklable, but pytest-flask's
+    LiveServer uses nested functions that can't be pickled.
+
+    Setting 'fork' explicitly:
+    - Maintains compatibility with Python 3.10-3.13 (where 'fork' was already default)
+    - Fixes Python 3.14 pickling errors
+    - Only affects Unix-like systems (Windows uses 'spawn' regardless)
+
+    See: https://github.com/python/cpython/issues/126831
+    See: https://docs.python.org/3/whatsnew/3.14.html
+    """
+    # Only set if not already set (respects existing configuration)
+    if multiprocessing.get_start_method(allow_none=True) is None:
+        try:
+            # 'fork' is available on Unix-like systems (Linux, macOS)
+            # On Windows, this will have no effect as 'spawn' is the only option
+            multiprocessing.set_start_method('fork', force=False)
+            logger.debug("Set multiprocessing start method to 'fork' for Python 3.14+ compatibility")
+        except (ValueError, RuntimeError):
+            # Already set, not available on this platform, or context already created
+            pass
+
 def pytest_addoption(parser):
     """Add custom command-line options for pytest.
 

changedetectionio/tests/test_backup.py

@@ -3,7 +3,7 @@
 from .util import set_original_response, live_server_setup, wait_for_all_checks
 from flask import url_for
 import io
-from zipfile import ZipFile
+from zipfile import ZipFile, ZIP_DEFLATED
 import re
 import time
 from changedetectionio.model import Watch, Tag
@@ -68,6 +68,9 @@ def test_backup(client, live_server, measure_memory_usage, datastore_path):
     # Check for changedetection.json (settings file)
     assert 'changedetection.json' in l, "changedetection.json should be in backup"
 
+    # secret.txt must never be included — it contains the Flask session key
+    assert 'secret.txt' not in l, "secret.txt (Flask session key) must not be included in backup"
+
     # Get the latest one
     res = client.get(
         url_for("backups.remove_backups"),
@@ -196,4 +199,63 @@ def test_backup_restore(client, live_server, measure_memory_usage, datastore_pat
     assert restored_tag2 is not None, f"Tag {tag_uuid2} not found after restore"
     assert restored_tag2['title'] == "Tasty backup tag number two", "Restored tag 2 title does not match"
     assert isinstance(restored_tag2, Tag.model), \
-        f"Tag 2 not properly rehydrated, got {type(restored_tag2)}"
+        f"Tag 2 not properly rehydrated, got {type(restored_tag2)}"
+
+
+def test_backup_restore_zip_slip_rejected(client, live_server, measure_memory_usage, datastore_path):
+    """Zip Slip path traversal entries in a restore zip must be rejected."""
+    import pytest
+    from changedetectionio.blueprint.backups.restore import import_from_zip
+
+    # Build a zip with a path traversal entry that would escape the extraction dir
+    malicious_zip = io.BytesIO()
+    with ZipFile(malicious_zip, 'w') as zf:
+        zf.writestr("../escaped.txt", "ATTACKER-CONTROLLED")
+    malicious_zip.seek(0)
+
+    datastore = live_server.app.config['DATASTORE']
+
+    with pytest.raises(ValueError, match="Zip Slip"):
+        import_from_zip(
+            zip_stream=malicious_zip,
+            datastore=datastore,
+            include_groups=True,
+            include_groups_replace=True,
+            include_watches=True,
+            include_watches_replace=True,
+        )
+
+
+def test_backup_restore_zip_bomb_rejected(client, live_server, measure_memory_usage, datastore_path):
+    """A zip whose total uncompressed size exceeds the limit must be rejected.
+
+    The guard reads file_size from the zip central-directory metadata — no
+    actual decompression happens, so this test is fast and uses minimal RAM.
+    100 KB of zeros compresses to ~100 bytes; monkeypatching the limit to
+    50 KB is enough to trigger the check without creating any large files.
+    """
+    import pytest
+    import changedetectionio.blueprint.backups.restore as restore_mod
+    from changedetectionio.blueprint.backups.restore import import_from_zip
+
+    # ~100 KB of zeros → deflate compresses to ~100 bytes, but file_size metadata = 100 KB
+    bomb_zip = io.BytesIO()
+    with ZipFile(bomb_zip, 'w', compression=ZIP_DEFLATED) as zf:
+        zf.writestr("data.txt", b"\x00" * (100 * 1024))
+    bomb_zip.seek(0)
+
+    datastore = live_server.app.config['DATASTORE']
+    original_limit = restore_mod._MAX_DECOMPRESSED_BYTES
+    try:
+        restore_mod._MAX_DECOMPRESSED_BYTES = 50 * 1024  # 50 KB limit for this test
+        with pytest.raises(ValueError, match="decompressed size"):
+            import_from_zip(
+                zip_stream=bomb_zip,
+                datastore=datastore,
+                include_groups=True,
+                include_groups_replace=True,
+                include_watches=True,
+                include_watches_replace=True,
+            )
+    finally:
+        restore_mod._MAX_DECOMPRESSED_BYTES = original_limit

changedetectionio/tests/test_encoding.py

@@ -1,6 +1,7 @@
 #!/usr/bin/env python3
 # coding=utf-8
 
+import hashlib
 import time
 from flask import url_for
 from .util import live_server_setup, wait_for_all_checks, extract_UUID_from_client
@@ -11,6 +12,69 @@ import os
 
 
 
+def test_surrogate_characters_in_content_are_sanitized():
+    """Lone surrogates can appear in requests' r.text when a server returns malformed/mixed-encoding
+    content. Without sanitization, encoding to UTF-8 raises UnicodeEncodeError.
+    See: https://github.com/dgtlmoon/changedetection.io/issues/3952
+    """
+    content_with_surrogate = '<html><body>Hello \udcad World</body></html>'
+
+    # Confirm the raw problem exists
+    with pytest.raises(UnicodeEncodeError):
+        content_with_surrogate.encode('utf-8')
+
+    # Our fix: sanitize after fetcher.run() in processors/base.py call_browser()
+    sanitized = content_with_surrogate.encode('utf-8', errors='replace').decode('utf-8')
+    assert 'Hello' in sanitized
+    assert 'World' in sanitized
+    assert '\udcad' not in sanitized
+
+    # Checksum computation (processors/base.py get_raw_document_checksum) must not crash
+    hashlib.md5(sanitized.encode('utf-8')).hexdigest()
+
+
+def test_utf8_content_without_charset_header(client, live_server, datastore_path):
+    """Server returns UTF-8 content but no charset in Content-Type header.
+    chardet can misdetect such pages as UTF-7 (Python 3.14 then produces surrogates).
+    Our fix tries UTF-8 first before falling back to chardet.
+    See: https://github.com/dgtlmoon/changedetection.io/issues/3952
+    """
+    from .util import write_test_file_and_sync
+    # UTF-8 encoded content with non-ASCII chars - no charset will be in the header
+    html = '<html><body><p>Español</p><p>Français</p><p>日本語</p></body></html>'
+    write_test_file_and_sync(os.path.join(datastore_path, "endpoint-content.txt"), html.encode('utf-8'), mode='wb')
+
+    test_url = url_for('test_endpoint', content_type="text/html", _external=True)
+    client.application.config.get('DATASTORE').add_watch(url=test_url)
+    client.get(url_for("ui.form_watch_checknow"), follow_redirects=True)
+    wait_for_all_checks(client)
+
+    res = client.get(url_for("ui.ui_preview.preview_page", uuid="first"), follow_redirects=True)
+    # Should decode correctly as UTF-8, not produce mojibake (Español) or replacement chars
+    assert 'Español'.encode('utf-8') in res.data
+    assert 'Français'.encode('utf-8') in res.data
+    assert '日本語'.encode('utf-8') in res.data
+
+
+def test_shiftjis_with_meta_charset(client, live_server, datastore_path):
+    """Server returns Shift-JIS content with no charset in HTTP header, but the HTML
+    declares <meta charset="Shift-JIS">. We should use the meta tag, not chardet.
+    Real-world case: https://github.com/dgtlmoon/changedetection.io/issues/3952
+    """
+    from .util import write_test_file_and_sync
+    japanese_text = '日本語のページ'
+    html = f'<html><head><meta http-equiv="Content-Type" content="text/html;charset=Shift-JIS"></head><body><p>{japanese_text}</p></body></html>'
+    write_test_file_and_sync(os.path.join(datastore_path, "endpoint-content.txt"), html.encode('shift_jis'), mode='wb')
+
+    test_url = url_for('test_endpoint', content_type="text/html", _external=True)
+    client.application.config.get('DATASTORE').add_watch(url=test_url)
+    client.get(url_for("ui.form_watch_checknow"), follow_redirects=True)
+    wait_for_all_checks(client)
+
+    res = client.get(url_for("ui.ui_preview.preview_page", uuid="first"), follow_redirects=True)
+    assert japanese_text.encode('utf-8') in res.data
+
+
 def set_html_response(datastore_path):
     test_return_data = """
 <html><body><span class="nav_second_img_text">

changedetectionio/tests/test_i18n.py

@@ -624,3 +624,76 @@ def test_session_locale_overrides_accept_language(client, live_server, measure_m
     assert "분".encode() in res.data, "Expected Korean '분' for Minutes"
     assert "小時".encode() not in res.data, "Should not have Traditional Chinese '小時' when Korean is set"
     assert "分鐘".encode() not in res.data, "Should not have Traditional Chinese '分鐘' when Korean is set"
+
+
+def test_clear_history_translated_confirmation(client, live_server, measure_memory_usage, datastore_path):
+    """
+    Test that clearing snapshot history works with translated confirmation text.
+
+    Issue #3865: When the app language is set to German, the clear history
+    confirmation dialog shows the translated word (e.g. 'loschen') but the
+    backend only accepted the English word 'clear', making it impossible
+    to clear snapshots in non-English languages.
+    """
+    from flask import url_for
+
+    test_url = url_for('test_endpoint', _external=True)
+
+    # Add a watch so there is history to clear
+    res = client.post(
+        url_for("imports.import_page"),
+        data={"urls": test_url},
+        follow_redirects=True
+    )
+    assert b"1 Imported" in res.data
+    wait_for_all_checks(client)
+
+    # Set language to German
+    res = client.get(
+        url_for("set_language", locale="de"),
+        follow_redirects=True
+    )
+    assert res.status_code == 200
+
+    # Verify the clear history page shows the German confirmation word
+    res = client.get(
+        url_for("ui.clear_all_history"),
+        follow_redirects=True
+    )
+    assert res.status_code == 200
+    assert "löschen".encode() in res.data, "Expected German word 'loschen' on clear history page"
+
+    # Submit the form with the German translated word
+    res = client.post(
+        url_for("ui.clear_all_history"),
+        data={"confirmtext": "löschen"},
+        follow_redirects=True
+    )
+    assert res.status_code == 200
+    # Should NOT show error message
+    assert b"Incorrect confirmation text" not in res.data, \
+        "German confirmation word 'loschen' should be accepted (issue #3865)"
+
+    # Switch back to English and verify English word still works
+    res = client.get(
+        url_for("set_language", locale="en_US"),
+        follow_redirects=True
+    )
+
+    res = client.post(
+        url_for("ui.clear_all_history"),
+        data={"confirmtext": "clear"},
+        follow_redirects=True
+    )
+    assert res.status_code == 200
+    assert b"Incorrect confirmation text" not in res.data, \
+        "English confirmation word 'clear' should still be accepted"
+
+    # Verify that missing/empty confirmtext does not crash the server
+    res = client.post(
+        url_for("ui.clear_all_history"),
+        data={},
+        follow_redirects=True
+    )
+    assert res.status_code == 200, \
+        "Missing confirmtext should not crash the server"

changedetectionio/tests/test_restock_itemprop.py

@@ -467,3 +467,38 @@ def test_special_prop_examples(client, live_server, measure_memory_usage, datast
             assert b'155.55' in res.data
 
     delete_all_watches(client)
+
+
+def test_itemprop_as_str(client, live_server, measure_memory_usage, datastore_path):
+
+    test_return_data = f"""<html>
+       <body>
+     Some initial text<br>
+     <p>Which is across multiple lines</p>
+<span itemprop="offers" itemscope itemtype="http://schema.org/Offer">
+<meta content="767.55" itemprop="price"/>
+<meta content="EUR" itemprop="priceCurrency"/>
+<meta content="InStock" itemprop="availability"/>
+<meta content="https://www.123-test.dk" itemprop="url"/>
+</span>
+     </body>
+     </html>
+    """
+
+    with open(os.path.join(datastore_path, "endpoint-content.txt"), "w") as f:
+        f.write(test_return_data)
+
+
+    test_url = url_for('test_endpoint', _external=True)
+
+    client.post(
+        url_for("ui.ui_views.form_quick_watch_add"),
+        data={"url": test_url, "tags": 'restock tests', 'processor': 'restock_diff'},
+        follow_redirects=True
+    )
+
+    client.get(url_for("ui.form_watch_checknow"))
+    wait_for_all_checks(client)
+
+    res = client.get(url_for("watchlist.index"))
+    assert b'767.55' in res.data

changedetectionio/tests/test_xpath_selector.py

@@ -592,3 +592,74 @@ def test_rss_xpath(client, live_server, measure_memory_usage, datastore_path):
         set_rss_atom_feed_response(header=feed_header, datastore_path=datastore_path)
         for content_type in RSS_XML_CONTENT_TYPES:
             _subtest_xpath_rss(client, content_type=content_type, datastore_path=datastore_path)
+
+
+# GHSA-6fmw-82m7-jq6p — XPath arbitrary file read via unparsed-text() and friends
+# Unit-level: verify xpath_filter() and SafeXPath3Parser block all dangerous functions.
+def test_xpath_blocked_functions_unit():
+    """Dangerous XPath 3.0 functions must be rejected at the parser level (no live server needed)."""
+    import elementpath
+    from changedetectionio.html_tools import xpath_filter, SafeXPath3Parser
+    from lxml import html
+
+    html_content = '<html><body><p>safe content</p></body></html>'
+
+    dangerous_expressions = [
+        "unparsed-text('file:///etc/passwd')",
+        "unparsed-text-lines('file:///etc/passwd')",
+        "unparsed-text-available('file:///etc/passwd')",
+        "doc('file:///etc/passwd')",
+        "doc-available('file:///etc/passwd')",
+        "environment-variable('PATH')",
+        "available-environment-variables()",
+    ]
+
+    for expr in dangerous_expressions:
+        # xpath_filter() must raise, not silently return file contents
+        try:
+            result = xpath_filter(expr, html_content)
+            assert False, f"xpath_filter should have raised for: {expr!r}, got: {result!r}"
+        except elementpath.ElementPathError:
+            pass  # expected
+
+        # SafeXPath3Parser must reject the expression at parse time
+        tree = html.fromstring(html_content)
+        try:
+            elementpath.select(tree, expr, parser=SafeXPath3Parser)
+            assert False, f"SafeXPath3Parser should have raised for: {expr!r}"
+        except elementpath.ElementPathError:
+            pass  # expected
+
+    # Sanity check: normal XPath still works
+    result = xpath_filter('//p/text()', html_content)
+    assert result == 'safe content'
+
+
+# GHSA-6fmw-82m7-jq6p — form validation must also reject dangerous XPath expressions.
+def test_xpath_blocked_functions_form_validation(client, live_server, measure_memory_usage, datastore_path):
+    """Edit-form validation must reject dangerous XPath 3.0 functions before they are stored."""
+    from flask import url_for
+
+    set_original_response(datastore_path=datastore_path)
+    test_url = url_for('test_endpoint', _external=True)
+    client.application.config.get('DATASTORE').add_watch(url=test_url)
+    client.get(url_for("ui.form_watch_checknow"), follow_redirects=True)
+    wait_for_all_checks(client)
+
+    dangerous_expressions = [
+        "xpath:unparsed-text('file:///etc/passwd')",
+        "xpath:environment-variable('PATH')",
+        "xpath:doc('file:///etc/passwd')",
+    ]
+
+    for expr in dangerous_expressions:
+        res = client.post(
+            url_for("ui.ui_edit.edit_page", uuid="first"),
+            data={"include_filters": expr, "url": test_url, "tags": "", "headers": "",
+                  'fetch_backend': "html_requests", "time_between_check_use_default": "y"},
+            follow_redirects=True
+        )
+        assert b"is not a valid XPath expression" in res.data, \
+            f"Form should reject dangerous expression: {expr!r}"
+
+    delete_all_watches(client)

changedetectionio/validate_url.py

@@ -100,6 +100,19 @@ def is_safe_valid_url(test_url):
         logger.warning('URL validation failed: URL is empty or whitespace only')
         return False
 
+    # Per-request cache: same URL is often validated 2-3x per watchlist render (sort + display).
+    # Flask's g is scoped to one request and auto-cleared on teardown, so dynamic Jinja2 URLs
+    # like {{microtime()}} are always re-evaluated on the next request.
+    # Falls back gracefully when called outside a request context (e.g. background workers).
+    _cache_key = test_url
+    try:
+        from flask import g
+        _cache = g.setdefault('_url_validation_cache', {})
+        if _cache_key in _cache:
+            return _cache[_cache_key]
+    except RuntimeError:
+        _cache = None  # No app context
+
     allow_file_access = strtobool(os.getenv('ALLOW_FILE_URI', 'false'))
     safe_protocol_regex = '^(http|https|ftp|file):' if allow_file_access else '^(http|https|ftp):'
 
@@ -112,11 +125,14 @@ def is_safe_valid_url(test_url):
     test_url = r.sub('', test_url)
 
     # Check the actual rendered URL in case of any Jinja markup
-    try:
-        test_url = jinja_render(test_url)
-    except Exception as e:
-        logger.error(f'URL "{test_url}" is not correct Jinja2? {str(e)}')
-        return False
+    # Only run jinja_render when the URL actually contains Jinja2 syntax - creating a new
+    # ImmutableSandboxedEnvironment is expensive and is called once per watch per page load
+    if '{%' in test_url or '{{' in test_url:
+        try:
+            test_url = jinja_render(test_url)
+        except Exception as e:
+            logger.error(f'URL "{test_url}" is not correct Jinja2? {str(e)}')
+            return False
 
     # Check query parameters and fragment
     if re.search(r'[<>]', test_url):
@@ -142,4 +158,6 @@ def is_safe_valid_url(test_url):
         logger.warning(f'URL f"{test_url}" failed validation, aborting.')
         return False
 
+    if _cache is not None:
+        _cache[_cache_key] = True
     return True

docs/api-spec.yaml

@@ -113,8 +113,156 @@ tags:
 
   - name: Plugin API Extensions
     description: |
-      Retrieve the live OpenAPI specification for this instance. Unlike the static spec, this endpoint
-      returns the fully merged spec including schemas for any processor plugins installed on this instance.
+      ## How Processor Plugins Extend the API
+
+      changedetection.io uses a **processor plugin** system to handle different types of change detection.
+      Each processor lives in `changedetectionio/processors/<name>/` and may include an `api.yaml` file
+      that extends the core Watch schema with processor-specific configuration fields.
+
+      ### How it works
+
+      At startup, changedetection.io scans all installed processors for an `api.yaml` file. Any schemas
+      and code samples defined there are deep-merged into the live API specification, making the
+      processor's configuration fields valid on all watch create and update requests.
+
+      The live, fully-merged spec is always available at `/api/v1/full-spec` — use that URL with
+      Swagger UI or Redoc to see the complete schema for your specific installation.
+
+      ---
+
+      ### Writing a processor `api.yaml`
+
+      Place an `api.yaml` in the processor plugin's own directory, alongside its `__init__.py`
+      (e.g. `changedetectionio/processors/my_processor/api.yaml`). The schema name **must** follow the
+      convention `processor_config_<processor_name>` (e.g. `processor_config_restock_diff`). That same
+      key is used as the JSON field name when creating or updating a watch.
+
+      A minimal `api.yaml` for a hypothetical `my_processor`:
+
+      ```yaml
+      components:
+        schemas:
+          processor_config_my_processor:
+            type: object
+            description: Configuration for my_processor
+            properties:
+              some_option:
+                type: boolean
+                default: true
+                description: Enable some behaviour
+
+      paths:
+        /watch:
+          post:
+            x-code-samples:
+              - lang: curl
+                label: my_processor example
+                source: |
+                  curl -X POST "http://localhost:5000/api/v1/watch" \
+                    -H "x-api-key: YOUR_API_KEY" \
+                    -H "Content-Type: application/json" \
+                    -d '{
+                      "url": "https://example.com",
+                      "processor": "my_processor",
+                      "processor_config_my_processor": { "some_option": true }
+                    }'
+      ```
+
+      The `paths` section in `api.yaml` is used only for injecting additional `x-code-samples` into
+      existing endpoints — you cannot define new routes via plugin.
+
+      ---
+
+      ### Built-in plugin: `restock_diff`
+
+      The `restock_diff` processor is always shipped with changedetection.io. It monitors product
+      availability and price changes using structured data (JSON-LD / schema.org microdata) and
+      text heuristics. It is activated by setting `"processor": "restock_diff"` on a watch.
+
+      It adds the `processor_config_restock_diff` block to the Watch schema with these fields:
+
+      | Field | Type | Default | Description |
+      |---|---|---|---|
+      | `in_stock_processing` | string | `in_stock_only` | `in_stock_only` — only alert Out-of-Stock→In-Stock · `all_changes` — alert any availability change · `off` — disable stock tracking |
+      | `follow_price_changes` | boolean | `true` | Monitor and alert on price changes |
+      | `price_change_min` | number\|null | — | Alert when price drops **below** this value |
+      | `price_change_max` | number\|null | — | Alert when price rises **above** this value |
+      | `price_change_threshold_percent` | number\|null | — | Minimum % change since the original price to trigger an alert |
+
+      #### CREATE — Add a restock/price monitor
+
+      ```bash
+      curl -X POST "http://localhost:5000/api/v1/watch" \
+        -H "x-api-key: YOUR_API_KEY" \
+        -H "Content-Type: application/json" \
+        -d '{
+          "url": "https://example.com/product/widget",
+          "processor": "restock_diff",
+          "processor_config_restock_diff": {
+            "in_stock_processing": "in_stock_only",
+            "follow_price_changes": true,
+            "price_change_threshold_percent": 5
+          }
+        }'
+      ```
+
+      #### READ — Retrieve the monitor
+
+      The response JSON includes `processor_config_restock_diff` alongside all standard watch fields:
+
+      ```bash
+      curl -X GET "http://localhost:5000/api/v1/watch/cc0cfffa-f449-477b-83ea-0caafd1dc091" \
+        -H "x-api-key: YOUR_API_KEY"
+      ```
+
+      ```json
+      {
+        "uuid": "cc0cfffa-f449-477b-83ea-0caafd1dc091",
+        "url": "https://example.com/product/widget",
+        "processor": "restock_diff",
+        "processor_config_restock_diff": {
+          "in_stock_processing": "in_stock_only",
+          "follow_price_changes": true,
+          "price_change_threshold_percent": 5,
+          "price_change_min": null,
+          "price_change_max": null
+        }
+      }
+      ```
+
+      #### UPDATE — Change thresholds without recreating the monitor
+
+      Only fields included in the request body are updated; omitted fields are left unchanged.
+
+      ```bash
+      curl -X PUT "http://localhost:5000/api/v1/watch/cc0cfffa-f449-477b-83ea-0caafd1dc091" \
+        -H "x-api-key: YOUR_API_KEY" \
+        -H "Content-Type: application/json" \
+        -d '{
+          "processor_config_restock_diff": {
+            "in_stock_processing": "all_changes",
+            "follow_price_changes": true,
+            "price_change_min": 10.00,
+            "price_change_max": 500.00
+          }
+        }'
+      ```
+
+      #### DELETE — Remove the monitor
+
+      ```bash
+      curl -X DELETE "http://localhost:5000/api/v1/watch/cc0cfffa-f449-477b-83ea-0caafd1dc091" \
+        -H "x-api-key: YOUR_API_KEY"
+      ```
+
+      ---
+
+      For the complete schema-validated documentation including all processor fields, fetch the live spec
+      and load it into Swagger UI or Redoc:
+
+      ```
+      GET /api/v1/full-spec
+      ```
 
 components:
   securitySchemes:
@@ -1921,15 +2069,35 @@ paths:
 
         Unlike the static `api-spec.yaml` shipped with the application, this endpoint returns the
         spec dynamically merged with any `api.yaml` schemas provided by installed processor plugins.
-        Use this URL with Swagger UI or Redoc to get accurate documentation for your specific install.
+
+        **Use this URL** with Swagger UI or Redoc to get schema-accurate documentation for your
+        specific install — it includes every `processor_config_<name>` schema block contributed by
+        installed processors (e.g. `processor_config_restock_diff` from the built-in restock plugin).
+
+        This endpoint requires no authentication and returns YAML.
+
+        To load it directly in Swagger UI, paste the URL into the "Explore" box:
+        ```
+        http://localhost:5000/api/v1/full-spec
+        ```
       security: []
       x-code-samples:
         - lang: 'curl'
           source: |
+            # Fetch the live merged spec (no API key needed)
             curl -X GET "http://localhost:5000/api/v1/full-spec"
+        - lang: 'Python'
+          source: |
+            import requests
+
+            # No authentication required
+            response = requests.get('http://localhost:5000/api/v1/full-spec')
+            print(response.text)  # Returns YAML
       responses:
         '200':
-          description: Merged OpenAPI specification in YAML format
+          description: |
+            Merged OpenAPI specification in YAML format. Includes all processor plugin schemas
+            (e.g. `processor_config_restock_diff`) not present in the static `api-spec.yaml`.
           content:
             application/yaml:
               schema: