[1.5.x] Fixed #19519 -- Fired request_finished in the WSGI iterable's close().

Backport of acc5396.

3 files changed, 27 insertions, 0 deletions

diff --git a/docs/ref/request-response.txt b/docs/ref/request-response.txt
index 2775c974d0..954e2d057e 100644
--- a/docs/ref/request-response.txt
+++ b/docs/ref/request-response.txt
@@ -804,6 +804,8 @@ types of HTTP responses. Like ``HttpResponse``, these subclasses live in
     :class:`~django.template.response.SimpleTemplateResponse`, and the
     ``render`` method must itself return a valid response object.
 
+.. _httpresponse-streaming:
+
 StreamingHttpResponse objects
 =============================
 
diff --git a/docs/ref/signals.txt b/docs/ref/signals.txt
index c31c90f4e8..f797b0f555 100644
--- a/docs/ref/signals.txt
+++ b/docs/ref/signals.txt
@@ -448,6 +448,18 @@ request_finished
 
 Sent when Django finishes processing an HTTP request.
 
+.. note::
+
+    When a view returns a :ref:`streaming response <httpresponse-streaming>`,
+    this signal is sent only after the entire response is consumed by the
+    client (strictly speaking, by the WSGI gateway).
+
+.. versionchanged:: 1.5
+
+    Before Django 1.5, this signal was fired before sending the content to the
+    client. In order to accomodate streaming responses, it is now fired after
+    sending the content.
+
 Arguments sent with this signal:
 
 ``sender``
diff --git a/docs/releases/1.5.txt b/docs/releases/1.5.txt
index a449f4ab12..c5e8c61922 100644
--- a/docs/releases/1.5.txt
+++ b/docs/releases/1.5.txt
@@ -411,6 +411,19 @@ attribute. Developers wishing to access the raw POST data for these cases,
 should use the :attr:`request.body <django.http.HttpRequest.body>` attribute
 instead.
 
+:data:`~django.core.signals.request_finished` signal
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Django used to send the :data:`~django.core.signals.request_finished` signal
+as soon as the view function returned a response. This interacted badly with
+:ref:`streaming responses <httpresponse-streaming>` that delay content
+generation.
+
+This signal is now sent after the content is fully consumed by the WSGI
+gateway. This might be backwards incompatible if you rely on the signal being
+fired before sending the response content to the client. If you do, you should
+consider using a middleware instead.
+
 OPTIONS, PUT and DELETE requests in the test client
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~