From: Aryeh Gregor <simetrical@users.mediawiki.org>
Date: Wed, 25 Aug 2010 17:56:03 +0000 (+0000)
Subject: Blacklist % in HTML id's
X-Git-Tag: 1.31.0-rc.0~35324
X-Git-Url: http://git.cyclocoop.org/?a=commitdiff_plain;h=839491a18c687f4619e8b243c359e0b8c5e1fdc7;p=lhc%2Fweb%2Fwiklou.git

Blacklist % in HTML id's

As pointed out by entlinkt at bug 24918, this could cause problems --
browsers don't interpret percent escapes in fragments consistently, so
any percent sign that might be part of an escape could be a problem.
For simplicity, just strip all of them.
---

diff --git a/includes/Sanitizer.php b/includes/Sanitizer.php
index 122a11e733..7fdd1df5f0 100644
--- a/includes/Sanitizer.php
+++ b/includes/Sanitizer.php
@@ -981,7 +981,9 @@ class Sanitizer {
 	 *
 	 * To ensure we don't have to bother escaping anything, we also strip ', ",
 	 * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
-	 * We also strip # because it upsets IE6.
+	 * We also strip # because it upsets IE, and % because it could be
+	 * ambiguous if it's part of something that looks like a percent escape
+	 * (which don't work reliably in fragments cross-browser).
 	 *
 	 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 	 *                                                          in the id and
@@ -1007,7 +1009,7 @@ class Sanitizer {
 
 		if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
 			$id = Sanitizer::decodeCharReferences( $id );
-			$id = preg_replace( '/[ \t\n\r\f_\'"&#]+/', '_', $id );
+			$id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
 			$id = trim( $id, '_' );
 			if ( $id === '' ) {
 				# Must have been all whitespace to start with.