Merge remote-tracking branch 'upstream/release-v1.56'

synapse/rest/media/v1/preview_url_resource.py

@@ -22,7 +22,7 @@ import shutil
 import sys
 import traceback
 from typing import TYPE_CHECKING, BinaryIO, Iterable, Optional, Tuple
-from urllib import parse as urlparse
+from urllib.parse import urljoin, urlparse, urlsplit
 from urllib.request import urlopen
 
 import attr
@@ -44,11 +44,7 @@ from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.rest.media.v1._base import get_filename_from_headers
 from synapse.rest.media.v1.media_storage import MediaStorage
 from synapse.rest.media.v1.oembed import OEmbedProvider
-from synapse.rest.media.v1.preview_html import (
-    decode_body,
-    parse_html_to_open_graph,
-    rebase_url,
-)
+from synapse.rest.media.v1.preview_html import decode_body, parse_html_to_open_graph
 from synapse.types import JsonDict, UserID
 from synapse.util import json_encoder
 from synapse.util.async_helpers import ObservableDeferred
@@ -187,7 +183,7 @@ class PreviewUrlResource(DirectServeJsonResource):
             ts = self.clock.time_msec()
 
         # XXX: we could move this into _do_preview if we wanted.
-        url_tuple = urlparse.urlsplit(url)
+        url_tuple = urlsplit(url)
         for entry in self.url_preview_url_blacklist:
             match = True
             for attrib in entry:
@@ -322,7 +318,7 @@ class PreviewUrlResource(DirectServeJsonResource):
 
                 # Parse Open Graph information from the HTML in case the oEmbed
                 # response failed or is incomplete.
-                og_from_html = parse_html_to_open_graph(tree, media_info.uri)
+                og_from_html = parse_html_to_open_graph(tree)
 
                 # Compile the Open Graph response by using the scraped
                 # information from the HTML and overlaying any information
@@ -588,12 +584,17 @@ class PreviewUrlResource(DirectServeJsonResource):
         if "og:image" not in og or not og["og:image"]:
             return
 
+        # The image URL from the HTML might be relative to the previewed page,
+        # convert it to an URL which can be requested directly.
+        image_url = og["og:image"]
+        url_parts = urlparse(image_url)
+        if url_parts.scheme != "data":
+            image_url = urljoin(media_info.uri, image_url)
+
         # FIXME: it might be cleaner to use the same flow as the main /preview_url
         # request itself and benefit from the same caching etc.  But for now we
         # just rely on the caching on the master request to speed things up.
-        image_info = await self._handle_url(
-            rebase_url(og["og:image"], media_info.uri), user, allow_data_urls=True
-        )
+        image_info = await self._handle_url(image_url, user, allow_data_urls=True)
 
         if _is_media(image_info.media_type):
             # TODO: make sure we don't choke on white-on-transparent images