Fix to #17031. See bug ticket for test code & test-results.

diff --git a/includes/Sanitizer.php b/includes/Sanitizer.php
index e8a3a38..c0b85f9 100644 (file)
--- a/includes/Sanitizer.php
+++ b/includes/Sanitizer.php
@@ -40,10 +40,11 @@ define( 'MW_CHAR_REFS_REGEX',
  * Allows some... latitude.
  * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
  */
-$attrib = '[A-Za-z0-9]';
+$attrib_first = '[:A-Z_a-z]';
+$attrib = '[:A-Z_a-z-.0-9]';
 $space = '[\x09\x0a\x0d\x20]';
 define( 'MW_ATTRIBS_REGEX',
-       "/(?:^|$space)((?:xml:|xmlns:)?$attrib+)
+    "/(?:^|$space)({$attrib_first}{$attrib}*)
          ($space*=$space*
                (?:
                 # The attribute value: quoted or alone