Accept charset in Content-Type header application/x-www-form-urlencoded.

Warn when this does not fit the default encoding.
This is my proposal from plone/buildout.coredev#844 (comment)
It fixes several problems in Plone.

CHANGES.rst

@@ -20,6 +20,9 @@ https://github.com/zopefoundation/Zope/blob/4.x/CHANGES.rst
 
   Fix encoding handling and ``:bytes`` converter.
 
+  Accept but ignore illegal charset in Content-Type header ``application/x-www-form-urlencoded``.
+  Warn when this does not fit the default encoding.
+
   See `#1094 <https://github.com/zopefoundation/Zope/pull/1094>`_.
 
 - Clean out and refactor dependency configuration files.

src/ZPublisher/HTTPRequest.py

@@ -16,6 +16,7 @@
 
 import codecs
 import html
+import logging
 import os
 import random
 import re
@@ -51,6 +52,8 @@
 from .cookie import getCookieValuePolicy
 
 
+logger = logging.getLogger('ZPublisher')
+
 # DOS attack protection -- limiting the amount of memory for forms
 # probably should become configurable
 FORM_MEMORY_LIMIT = 2 ** 20  # memory limit for forms
@@ -1425,7 +1428,25 @@ def __init__(self, fp, environ):
                     disk_limit=FORM_DISK_LIMIT,
                     memfile_limit=FORM_MEMFILE_LIMIT,
                     charset="latin-1").parts()
-            elif content_type == "application/x-www-form-urlencoded":
+            elif content_type.startswith("application/x-www-form-urlencoded"):
+                # In some cases we get a charset:
+                # "application/x-www-form-urlencoded; charset=UTF-8"
+                # This is illegal according to the specification.
+                # See https://github.com/plone/buildout.coredev/pull/844
+                # We ignore it.
+                # When the charset does not match our default encoding,
+                # we log a warning, to make the user aware.
+                ct_split = content_type.split("charset=")
+                if len(ct_split) > 1:
+                    requested_charset = ct_split[1].lower()
+                    if requested_charset != default_encoding:
+                        logger.warning(
+                            "Specifying a charset in this Content-Type header "
+                            "is not allowed by the HTTP specification. "
+                            "We ignore the charset. Header is: %r",
+                            content_type
+                        )
+
                 if qs:
                     qs += "&"
                 qs += fp.read(FORM_MEMORY_LIMIT).decode("latin-1")