From 8e8b15afc684e65946b5f6101e74ced518d2eee6 Mon Sep 17 00:00:00 2001
From: Arlo Breault <abreault@wikimedia.org>
Date: Wed, 7 Jan 2015 12:46:59 -0800
Subject: [PATCH] Non-word characters shouldn't terminate tag names on the tidy
 side too

 * Follow up to Iceec404f46703065bf080dd2cbfed1f88c204fa5.

 * The accepted charset is changed to match the HTML5 parsing spec at:
   http://dev.w3.org/html5/spec-preview/tokenization.html#tag-open-state

 * Equivalent in parsoid at I462c336f9a00c8ccd11f3220a8738389e8ba7c7c.

Change-Id: I69cb000538fe195dd77273da5f91697fe1e7d283
---
 includes/Sanitizer.php       | 14 +++++++------
 tests/parser/parserTests.txt | 39 +++++++++++++++++++++++++++---------
 2 files changed, 38 insertions(+), 15 deletions(-)

diff --git a/includes/Sanitizer.php b/includes/Sanitizer.php
index f79e94d459..a2de0044a6 100644
--- a/includes/Sanitizer.php
+++ b/includes/Sanitizer.php
@@ -39,6 +39,12 @@ class Sanitizer {
 		 |&\#[xX]([0-9A-Fa-f]+);
 		 |(&)/x';
 
+	/**
+	 * Acceptable tag name charset from HTML5 parsing spec
+	 * http://dev.w3.org/html5/spec-preview/tokenization.html#tag-open-state
+	 */
+	const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
+
 	/**
 	 * Blacklist for evil uris like javascript:
 	 * WARNING: DO NOT use this in any place that actually requires blacklisting
@@ -444,7 +450,7 @@ class Sanitizer {
 				# $params: String between element name and >
 				# $brace: Ending '>' or '/>'
 				# $rest: Everything until the next element of $bits
-				if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
+				if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 					list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 				} else {
 					$slash = $t = $params = $brace = $rest = null;
@@ -567,11 +573,7 @@ class Sanitizer {
 		} else {
 			# this might be possible using tidy itself
 			foreach ( $bits as $x ) {
-				preg_match(
-					'/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
-					$x,
-					$regs
-				);
+				preg_match( self::ELEMENT_BITS_REGEX, $x, $regs );
 
 				wfSuppressWarnings();
 				list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
diff --git a/tests/parser/parserTests.txt b/tests/parser/parserTests.txt
index 18aeabb540..04e39ee3fb 100644
--- a/tests/parser/parserTests.txt
+++ b/tests/parser/parserTests.txt
@@ -1191,28 +1191,49 @@ Ruby markup (W3C-style)
 </p>
 !! end
 
-# There is a tidy bug here: http://sourceforge.net/p/tidy/bugs/946/
+# The next two test different paths in the sanitizer.
 !! test
 Non-word characters don't terminate tag names (bug 17663, 40670, 52022)
 !! wikitext
-<b→> doesn't work! </b→>
+<b→> doesn't terminate </b→>
 
-<bä> doesn't work! </bä>
+<bä> doesn't terminate </bä>
 
-<boo> works fine </boo>
+<boo> doesn't terminate </boo>
 
-<s.foo>s.foo</s.foo>
+<s.foo> doesn't terminate </s.foo>
 
 <sub-ID#1>
 !! html
-<p>&lt;b→&gt; doesn't work! &lt;/b→&gt;
-</p><p>&lt;bä&gt; doesn't work! &lt;/bä&gt;
-</p><p>&lt;boo&gt; works fine &lt;/boo&gt;
-</p><p>&lt;s.foo&gt;s.foo&lt;/s.foo&gt;
+<p>&lt;b→&gt; doesn't terminate &lt;/b→&gt;
+</p><p>&lt;bä&gt; doesn't terminate &lt;/bä&gt;
+</p><p>&lt;boo&gt; doesn't terminate &lt;/boo&gt;
+</p><p>&lt;s.foo&gt; doesn't terminate &lt;/s.foo&gt;
 </p><p>&lt;sub-ID#1&gt;
 </p>
 !! end
 
+# There is a tidy bug here: http://sourceforge.net/p/tidy/bugs/946/
+!! test
+Non-word characters don't terminate tag names + tidy
+!! wikitext
+<b→> doesn't terminate </b→>
+
+<bä> doesn't terminate </bä>
+
+<boo> doesn't terminate </boo>
+
+<s.foo> doesn't terminate </s.foo>
+
+<sub-ID#1>
+!! html+tidy
+<p>&lt;b→&gt; doesn't terminate &lt;/b→&gt;</p>
+<p>&lt;bä&gt; doesn't terminate &lt;/bä&gt;</p>
+<p>&lt;boo&gt; doesn't terminate &lt;/boo&gt;</p>
+<p>&lt;s.foo&gt; doesn't terminate &lt;/s.foo&gt;</p>
+<p>&lt;sub-ID#1&gt;</p>
+!! end
+
 !! test
 Isolated close tags should be treated as literal text (bug 52760)
 !! wikitext
-- 
2.20.1