Non-ASCII query string aren't decoded properly

[Aymeric Augustin]

In [fcc8de05] I enabled unicode_literals in django.core.servers.basehttp.

This turns environ['QUERY_STRING'] into a unicode string, which later on prevents correct decoding in QueryDict.

A quick'n'dirty, Python 2 only fix is:

diff --git a/django/core/servers/basehttp.py b/django/core/servers/basehttp.py
index 19b287a..af8f2a0 100644
--- a/django/core/servers/basehttp.py
+++ b/django/core/servers/basehttp.py
@@ -144,9 +144,9 @@ class WSGIRequestHandler(simple_server.WSGIRequestHandler, object):
         env['SERVER_PROTOCOL'] = self.request_version
         env['REQUEST_METHOD'] = self.command
         if '?' in self.path:
-            path,query = self.path.split('?',1)
+            path, query = self.path.split(b'?', 1)
         else:
-            path,query = self.path,''
+            path, query = self.path, b''
 
         env['PATH_INFO'] = unquote(path)
         env['QUERY_STRING'] = query

An audit of this module seems necessary. It may even be extended to django.core.servers.

[Aymeric Augustin]

This module starts with a comment that states:

Based on wsgiref.simple_server which is part of the standard library since 2.5.

If this code was copy-pasted into Django for compatibility with Python < 2.5, and we didn't alter it significantly, we should consider switching to the classes provided by wsgiref.

After discussing this with Florian on IRC, we'd like to try to compare the current code with the 2.6 and 2.7 stdlib, and switch to the stdlib wherever possible, subclassing if necessary. Thus we'll benefit from a correct WSGI implementation both under Python 2 and 3.

[Claude Paroz]

I think it's a duplicate of #19075, where i suggested another patch. If it is not correct, I think the test case can be kept.