Document that follow_all requests share meta and cb_kwargs values (#7962)

scrapy/http/response/__init__.py

@@ -305,6 +305,11 @@ class Response(object_ref):
         :class:`~.TextResponse` provides a :meth:`~.TextResponse.follow_all`
         method which supports selectors in addition to absolute/relative URLs
         and Link objects.
+
+        .. caution:: Every returned request gets its own *meta* and
+            *cb_kwargs* dictionaries, but the values within them are shared.
+            Mutating one of those values, e.g. appending to a list, affects
+            all the returned requests.
         """
         if not hasattr(urls, "__iter__"):
             raise TypeError("'urls' argument must be an iterable")

scrapy/http/response/text.py

@@ -276,6 +276,9 @@ class TextResponse(Response):
         using the ``css`` or ``xpath`` parameters, this method will not produce requests for
         selectors from which links cannot be obtained (for instance, anchor tags without an
         ``href`` attribute)
+
+        .. seealso:: :meth:`.Response.follow_all`, for a caution about mutable
+            *meta* and *cb_kwargs* values.
         """
         arguments = [x for x in (urls, css, xpath) if x is not None]
         if len(arguments) != 1:

scrapy/spidermiddlewares/metacopy.py

@@ -15,8 +15,7 @@ logger = logging.getLogger(__name__)
 
 class MetaCopyDetectionMiddleware(BaseSpiderMiddleware):
     """Warn when a spider yields a request with internal meta keys that should
-    not be copied from response.meta, or when two requests share the same meta
-    dict object.
+    not be copied from response.meta.
 
     Each warning is emitted at most once per crawl.
     """