Do not fail completely if oEmbed autodiscovery fails. (#15092)

clokep · web-flow · commit 682151a464f6 · 2023-02-23T16:08:53.000-05:00
Previously if an autodiscovered oEmbed request failed (e.g. the
oEmbed endpoint is down or does not exist) then the entire URL
preview would fail. Instead we now return everything we can, even
if this additional request fails.
diff --git a/changelog.d/15092.bugfix b/changelog.d/15092.bugfix
@@ -0,0 +1 @@
+Fix a long-standing bug where a URL preview would break if the discovered oEmbed failed to download.
diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py
@@ -163,6 +163,10 @@ class PreviewUrlResource(DirectServeJsonResource):
        7. Stores the result in the database cache.
     4. Returns the result.
 
+    If any additional requests (e.g. from oEmbed autodiscovery, step 5.3 or
+    image thumbnailing, step 5.4 or 6.4) fails then the URL preview as a whole
+    does not fail. As much information as possible is returned.
+
     The in-memory cache expires after 1 hour.
 
     Expired entries in the database cache (and their associated media files) are
@@ -364,16 +368,25 @@ async def _do_preview(self, url: str, user: UserID, ts: int) -> bytes:
                 oembed_url = self._oembed.autodiscover_from_html(tree)
                 og_from_oembed: JsonDict = {}
                 if oembed_url:
-                    oembed_info = await self._handle_url(
-                        oembed_url, user, allow_data_urls=True
-                    )
-                    (
-                        og_from_oembed,
-                        author_name,
-                        expiration_ms,
-                    ) = await self._handle_oembed_response(
-                        url, oembed_info, expiration_ms
-                    )
+                    try:
+                        oembed_info = await self._handle_url(
+                            oembed_url, user, allow_data_urls=True
+                        )
+                    except Exception as e:
+                        # Fetching the oEmbed info failed, don't block the entire URL preview.
+                        logger.warning(
+                            "oEmbed fetch failed during URL preview: %s errored with %s",
+                            oembed_url,
+                            e,
+                        )
+                    else:
+                        (
+                            og_from_oembed,
+                            author_name,
+                            expiration_ms,
+                        ) = await self._handle_oembed_response(
+                            url, oembed_info, expiration_ms
+                        )
 
                 # Parse Open Graph information from the HTML in case the oEmbed
                 # response failed or is incomplete.
diff --git a/tests/rest/media/v1/test_url_preview.py b/tests/rest/media/v1/test_url_preview.py
@@ -657,7 +657,7 @@ def test_nonexistent_image(self) -> None:
         """If the preview image doesn't exist, ensure some data is returned."""
         self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")]
 
-        end_content = (
+        result = (
             b"""<html><body><img src="http://cdn.matrix.org/foo.jpg"></body></html>"""
         )
 
@@ -678,8 +678,8 @@ def test_nonexistent_image(self) -> None:
                 b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n"
                 b'Content-Type: text/html; charset="utf8"\r\n\r\n'
             )
-            % (len(end_content),)
-            + end_content
+            % (len(result),)
+            + result
         )
 
         self.pump()
@@ -688,6 +688,44 @@ def test_nonexistent_image(self) -> None:
         # The image should not be in the result.
         self.assertNotIn("og:image", channel.json_body)
 
+    def test_oembed_failure(self) -> None:
+        """If the autodiscovered oEmbed URL fails, ensure some data is returned."""
+        self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")]
+
+        result = b"""
+        <title>oEmbed Autodiscovery Fail</title>
+        <link rel="alternate" type="application/json+oembed"
+            href="http://example.com/oembed?url=http%3A%2F%2Fmatrix.org&format=json"
+            title="matrixdotorg" />
+        """
+
+        channel = self.make_request(
+            "GET",
+            "preview_url?url=http://matrix.org",
+            shorthand=False,
+            await_result=False,
+        )
+        self.pump()
+
+        client = self.reactor.tcpClients[0][2].buildProtocol(None)
+        server = AccumulatingProtocol()
+        server.makeConnection(FakeTransport(client, self.reactor))
+        client.makeConnection(FakeTransport(server, self.reactor))
+        client.dataReceived(
+            (
+                b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n"
+                b'Content-Type: text/html; charset="utf8"\r\n\r\n'
+            )
+            % (len(result),)
+            + result
+        )
+
+        self.pump()
+        self.assertEqual(channel.code, 200)
+
+        # The image should not be in the result.
+        self.assertEqual(channel.json_body["og:title"], "oEmbed Autodiscovery Fail")
+
     def test_data_url(self) -> None:
         """
         Requesting to preview a data URL is not supported.

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1 @@`
	`1`	`+Fix a long-standing bug where a URL preview would break if the discovered oEmbed failed to download.`