From 4b43d063843891153be8f6caebaa2ad5aba8d04d Mon Sep 17 00:00:00 2001
From: Ali Sabil <ali.sabil@tillitech.com>
Date: Sun, 8 Apr 2012 11:51:51 +0200
Subject: Make charsets parsing more relaxed

Certain user agents send invalid Accept-Charset headers, like the
following: "ISO-8859-1;utf-8;q=0.7,*;q=0.7"

The user agent with which this behavior was observed presented itself
with the User-Agent string: "Mozilla/5.0 (Windows; U; Windows NT 5.1;
en-US; rv:1.9) Gecko/2008052906 Firefox/3.0" Although this doesn't
appear to be correct. The request might have been mangled by a
transparent proxy.
---
 src/cowboy_http.erl | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/src/cowboy_http.erl b/src/cowboy_http.erl
index 2f4f982..7bea8e0 100644
--- a/src/cowboy_http.erl
+++ b/src/cowboy_http.erl
@@ -269,7 +269,15 @@ maybe_qparam(Data, Fun) ->
 		fun (<< $;, Rest/binary >>) ->
 			whitespace(Rest,
 				fun (Rest2) ->
-					qparam(Rest2, Fun)
+					%% This is a non-strict parsing clause required by some user agents
+					%% that use the wrong delimiter putting a charset where a qparam is
+					%% expected.
+					try qparam(Rest2, Fun) of
+						Result -> Result
+					catch
+						error:function_clause ->
+							Fun(<<",", Rest2/binary>>, 1000)
+					end
 				end);
 			(Rest) ->
 				Fun(Rest, 1000)
@@ -879,6 +887,12 @@ nonempty_charset_list_test_() ->
 		{<<"iso-8859-5, unicode-1-1;q=0.8">>, [
 			{<<"iso-8859-5">>, 1000},
 			{<<"unicode-1-1">>, 800}
+		]},
+		%% Some user agents send this invalid value for the Accept-Charset header
+		{<<"ISO-8859-1;utf-8;q=0.7,*;q=0.7">>, [
+			{<<"iso-8859-1">>, 1000},
+			{<<"utf-8">>, 700},
+			{<<"*">>, 700}
 		]}
 	],
 	[{V, fun() -> R = nonempty_list(V, fun conneg/2) end} || {V, R} <- Tests].
-- 
cgit v1.2.3