fix: close discarded cached response bodies

cachecontrol/controller.py

@@ -162,7 +162,12 @@ def _load_from_cache(self, request: PreparedRequest) -> HTTPResponse | None:
         else:
             body_file = None
 
-        result = self.serializer.loads(request, cache_data, body_file)
+        result = None
+        try:
+            result = self.serializer.loads(request, cache_data, body_file)
+        finally:
+            if result is None and body_file is not None:
+                body_file.close()
         if result is None:
             logger.debug("Cache entry deserialization failed, entry ignored")
         return result
@@ -204,12 +209,14 @@ def cached_request(self, request: PreparedRequest) -> HTTPResponse | Literal[Fal
                 logger.debug("Purging cached response: no date or etag")
                 self.cache.delete(cache_url)
             logger.debug("Ignoring cached response: no date")
+            resp.close()
             return False
 
         now = time.time()
         time_tuple = parsedate_tz(headers["date"])
         if time_tuple is None:
             logger.debug("Ignoring cached response: invalid date")
+            resp.close()
             return False
         date = calendar.timegm(time_tuple[:6])
         current_age = max(0, now - date)
@@ -271,20 +278,24 @@ def cached_request(self, request: PreparedRequest) -> HTTPResponse | Literal[Fal
             self.cache.delete(cache_url)
 
         # return the original handler
+        resp.close()
         return False
 
     def conditional_headers(self, request: PreparedRequest) -> dict[str, str]:
         resp = self._load_from_cache(request)
         new_headers = {}
 
         if resp:
-            headers: CaseInsensitiveDict[str] = CaseInsensitiveDict(resp.headers)
+            try:
+                headers: CaseInsensitiveDict[str] = CaseInsensitiveDict(resp.headers)
 
-            if "etag" in headers:
-                new_headers["If-None-Match"] = headers["ETag"]
+                if "etag" in headers:
+                    new_headers["If-None-Match"] = headers["ETag"]
 
-            if "last-modified" in headers:
-                new_headers["If-Modified-Since"] = headers["Last-Modified"]
+                if "last-modified" in headers:
+                    new_headers["If-Modified-Since"] = headers["Last-Modified"]
+            finally:
+                resp.close()
 
         return new_headers
 

tests/test_cache_control.py

@@ -56,6 +56,18 @@ def test_no_cache_non_20x_response(self, cc):
         assert cc.serializer.dumps.called
         assert cc.cache.set.called
 
+    def test_load_from_cache_closes_body_on_deserialization_failure(self, tmp_path):
+        cache = SeparateBodyFileCache(str(tmp_path))
+        body_file = Mock()
+        cache.get = Mock(return_value=b"metadata")
+        cache.get_body = Mock(return_value=body_file)
+        serializer = Mock()
+        serializer.loads.return_value = None
+        controller = CacheController(cache, serializer=serializer)
+
+        assert controller._load_from_cache(self.req()) is None
+        body_file.close.assert_called_once_with()
+
     def test_no_cache_with_no_date(self, cc):
         # No date header which makes our max-age pointless
         resp = self.resp({"cache-control": "max-age=3600"})
@@ -367,6 +379,17 @@ def test_cache_request_unfresh_max_age(self):
         self.c.cache = DictCache({self.url: resp})
         r = self.req({})
         assert not r
+        resp.close.assert_called_once_with()
+
+    def test_conditional_headers_closes_cached_response(self):
+        now = time.strftime(TIME_FMT, time.gmtime())
+        resp = Mock(headers={"etag": "value", "date": now}, status=200)
+        self.c.cache = DictCache({self.url: resp})
+
+        assert self.c.conditional_headers(Mock(url=self.url, headers={})) == {
+            "If-None-Match": "value"
+        }
+        resp.close.assert_called_once_with()
 
     def test_cache_request_unfresh_permanent_redirect(self):
         earlier = time.time() - 3600