Fix URL preview bugs (type error when loading cache from db, content-type including quotes) (#4157)

2 files changed, 20 insertions, 10 deletions

diff --git a/synapse/http/server.py b/synapse/http/server.py
index b4b25cab19..6a427d96a6 100644
--- a/synapse/http/server.py
+++ b/synapse/http/server.py
@@ -468,13 +468,13 @@ def set_cors_headers(request):
     Args:
         request (twisted.web.http.Request): The http request to add CORs to.
     """
-    request.setHeader("Access-Control-Allow-Origin", "*")
+    request.setHeader(b"Access-Control-Allow-Origin", b"*")
     request.setHeader(
-        "Access-Control-Allow-Methods", "GET, POST, PUT, DELETE, OPTIONS"
+        b"Access-Control-Allow-Methods", b"GET, POST, PUT, DELETE, OPTIONS"
     )
     request.setHeader(
-        "Access-Control-Allow-Headers",
-        "Origin, X-Requested-With, Content-Type, Accept, Authorization"
+        b"Access-Control-Allow-Headers",
+        b"Origin, X-Requested-With, Content-Type, Accept, Authorization"
     )
 
 
diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py
index 1a7bfd6b56..91d1dafe64 100644
--- a/synapse/rest/media/v1/preview_url_resource.py
+++ b/synapse/rest/media/v1/preview_url_resource.py
@@ -12,6 +12,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 import cgi
 import datetime
 import errno
@@ -24,6 +25,7 @@ import shutil
 import sys
 import traceback
 
+import six
 from six import string_types
 from six.moves import urllib_parse as urlparse
 
@@ -98,7 +100,7 @@ class PreviewUrlResource(Resource):
         # XXX: if get_user_by_req fails, what should we do in an async render?
         requester = yield self.auth.get_user_by_req(request)
         url = parse_string(request, "url")
-        if "ts" in request.args:
+        if b"ts" in request.args:
             ts = parse_integer(request, "ts")
         else:
             ts = self.clock.time_msec()
@@ -180,7 +182,12 @@ class PreviewUrlResource(Resource):
             cache_result["expires_ts"] > ts and
             cache_result["response_code"] / 100 == 2
         ):
-            defer.returnValue(cache_result["og"])
+            # It may be stored as text in the database, not as bytes (such as
+            # PostgreSQL). If so, encode it back before handing it on.
+            og = cache_result["og"]
+            if isinstance(og, six.text_type):
+                og = og.encode('utf8')
+            defer.returnValue(og)
             return
 
         media_info = yield self._download_url(url, user)
@@ -213,14 +220,17 @@ class PreviewUrlResource(Resource):
         elif _is_html(media_info['media_type']):
             # TODO: somehow stop a big HTML tree from exploding synapse's RAM
 
-            file = open(media_info['filename'])
-            body = file.read()
-            file.close()
+            with open(media_info['filename'], 'rb') as file:
+                body = file.read()
 
             # clobber the encoding from the content-type, or default to utf-8
             # XXX: this overrides any <meta/> or XML charset headers in the body
             # which may pose problems, but so far seems to work okay.
-            match = re.match(r'.*; *charset=(.*?)(;|$)', media_info['media_type'], re.I)
+            match = re.match(
+                r'.*; *charset="?(.*?)"?(;|$)',
+                media_info['media_type'],
+                re.I
+            )
             encoding = match.group(1) if match else "utf-8"
 
             og = decode_and_calc_og(body, media_info['uri'], encoding)