includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for MediaWiki
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * http://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * XHTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  42         /**
  43          * Blacklist for evil uris like javascript:
  44          * WARNING: DO NOT use this in any place that actually requires blacklisting
  45          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  46          * only way to be secure from javascript: uri based xss vectors is to whitelist
  47          * things that you know are safe and deny everything else.
  48          * [1]: http://ha.ckers.org/xss.html
  49          */
  50         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  51         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  52
  53         /**
  54          * List of all named character entities defined in HTML 4.01
  55          * http://www.w3.org/TR/html4/sgml/entities.html
  56          * As well as &apos; which is only defined starting in XHTML1.
  57          * @private
  58          */
  59         static $htmlEntities = array(
  60                 'Aacute'   => 193,
  61                 'aacute'   => 225,
  62                 'Acirc'    => 194,
  63                 'acirc'    => 226,
  64                 'acute'    => 180,
  65                 'AElig'    => 198,
  66                 'aelig'    => 230,
  67                 'Agrave'   => 192,
  68                 'agrave'   => 224,
  69                 'alefsym'  => 8501,
  70                 'Alpha'    => 913,
  71                 'alpha'    => 945,
  72                 'amp'      => 38,
  73                 'and'      => 8743,
  74                 'ang'      => 8736,
  75                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  76                 'Aring'    => 197,
  77                 'aring'    => 229,
  78                 'asymp'    => 8776,
  79                 'Atilde'   => 195,
  80                 'atilde'   => 227,
  81                 'Auml'     => 196,
  82                 'auml'     => 228,
  83                 'bdquo'    => 8222,
  84                 'Beta'     => 914,
  85                 'beta'     => 946,
  86                 'brvbar'   => 166,
  87                 'bull'     => 8226,
  88                 'cap'      => 8745,
  89                 'Ccedil'   => 199,
  90                 'ccedil'   => 231,
  91                 'cedil'    => 184,
  92                 'cent'     => 162,
  93                 'Chi'      => 935,
  94                 'chi'      => 967,
  95                 'circ'     => 710,
  96                 'clubs'    => 9827,
  97                 'cong'     => 8773,
  98                 'copy'     => 169,
  99                 'crarr'    => 8629,
 100                 'cup'      => 8746,
 101                 'curren'   => 164,
 102                 'dagger'   => 8224,
 103                 'Dagger'   => 8225,
 104                 'darr'     => 8595,
 105                 'dArr'     => 8659,
 106                 'deg'      => 176,
 107                 'Delta'    => 916,
 108                 'delta'    => 948,
 109                 'diams'    => 9830,
 110                 'divide'   => 247,
 111                 'Eacute'   => 201,
 112                 'eacute'   => 233,
 113                 'Ecirc'    => 202,
 114                 'ecirc'    => 234,
 115                 'Egrave'   => 200,
 116                 'egrave'   => 232,
 117                 'empty'    => 8709,
 118                 'emsp'     => 8195,
 119                 'ensp'     => 8194,
 120                 'Epsilon'  => 917,
 121                 'epsilon'  => 949,
 122                 'equiv'    => 8801,
 123                 'Eta'      => 919,
 124                 'eta'      => 951,
 125                 'ETH'      => 208,
 126                 'eth'      => 240,
 127                 'Euml'     => 203,
 128                 'euml'     => 235,
 129                 'euro'     => 8364,
 130                 'exist'    => 8707,
 131                 'fnof'     => 402,
 132                 'forall'   => 8704,
 133                 'frac12'   => 189,
 134                 'frac14'   => 188,
 135                 'frac34'   => 190,
 136                 'frasl'    => 8260,
 137                 'Gamma'    => 915,
 138                 'gamma'    => 947,
 139                 'ge'       => 8805,
 140                 'gt'       => 62,
 141                 'harr'     => 8596,
 142                 'hArr'     => 8660,
 143                 'hearts'   => 9829,
 144                 'hellip'   => 8230,
 145                 'Iacute'   => 205,
 146                 'iacute'   => 237,
 147                 'Icirc'    => 206,
 148                 'icirc'    => 238,
 149                 'iexcl'    => 161,
 150                 'Igrave'   => 204,
 151                 'igrave'   => 236,
 152                 'image'    => 8465,
 153                 'infin'    => 8734,
 154                 'int'      => 8747,
 155                 'Iota'     => 921,
 156                 'iota'     => 953,
 157                 'iquest'   => 191,
 158                 'isin'     => 8712,
 159                 'Iuml'     => 207,
 160                 'iuml'     => 239,
 161                 'Kappa'    => 922,
 162                 'kappa'    => 954,
 163                 'Lambda'   => 923,
 164                 'lambda'   => 955,
 165                 'lang'     => 9001,
 166                 'laquo'    => 171,
 167                 'larr'     => 8592,
 168                 'lArr'     => 8656,
 169                 'lceil'    => 8968,
 170                 'ldquo'    => 8220,
 171                 'le'       => 8804,
 172                 'lfloor'   => 8970,
 173                 'lowast'   => 8727,
 174                 'loz'      => 9674,
 175                 'lrm'      => 8206,
 176                 'lsaquo'   => 8249,
 177                 'lsquo'    => 8216,
 178                 'lt'       => 60,
 179                 'macr'     => 175,
 180                 'mdash'    => 8212,
 181                 'micro'    => 181,
 182                 'middot'   => 183,
 183                 'minus'    => 8722,
 184                 'Mu'       => 924,
 185                 'mu'       => 956,
 186                 'nabla'    => 8711,
 187                 'nbsp'     => 160,
 188                 'ndash'    => 8211,
 189                 'ne'       => 8800,
 190                 'ni'       => 8715,
 191                 'not'      => 172,
 192                 'notin'    => 8713,
 193                 'nsub'     => 8836,
 194                 'Ntilde'   => 209,
 195                 'ntilde'   => 241,
 196                 'Nu'       => 925,
 197                 'nu'       => 957,
 198                 'Oacute'   => 211,
 199                 'oacute'   => 243,
 200                 'Ocirc'    => 212,
 201                 'ocirc'    => 244,
 202                 'OElig'    => 338,
 203                 'oelig'    => 339,
 204                 'Ograve'   => 210,
 205                 'ograve'   => 242,
 206                 'oline'    => 8254,
 207                 'Omega'    => 937,
 208                 'omega'    => 969,
 209                 'Omicron'  => 927,
 210                 'omicron'  => 959,
 211                 'oplus'    => 8853,
 212                 'or'       => 8744,
 213                 'ordf'     => 170,
 214                 'ordm'     => 186,
 215                 'Oslash'   => 216,
 216                 'oslash'   => 248,
 217                 'Otilde'   => 213,
 218                 'otilde'   => 245,
 219                 'otimes'   => 8855,
 220                 'Ouml'     => 214,
 221                 'ouml'     => 246,
 222                 'para'     => 182,
 223                 'part'     => 8706,
 224                 'permil'   => 8240,
 225                 'perp'     => 8869,
 226                 'Phi'      => 934,
 227                 'phi'      => 966,
 228                 'Pi'       => 928,
 229                 'pi'       => 960,
 230                 'piv'      => 982,
 231                 'plusmn'   => 177,
 232                 'pound'    => 163,
 233                 'prime'    => 8242,
 234                 'Prime'    => 8243,
 235                 'prod'     => 8719,
 236                 'prop'     => 8733,
 237                 'Psi'      => 936,
 238                 'psi'      => 968,
 239                 'quot'     => 34,
 240                 'radic'    => 8730,
 241                 'rang'     => 9002,
 242                 'raquo'    => 187,
 243                 'rarr'     => 8594,
 244                 'rArr'     => 8658,
 245                 'rceil'    => 8969,
 246                 'rdquo'    => 8221,
 247                 'real'     => 8476,
 248                 'reg'      => 174,
 249                 'rfloor'   => 8971,
 250                 'Rho'      => 929,
 251                 'rho'      => 961,
 252                 'rlm'      => 8207,
 253                 'rsaquo'   => 8250,
 254                 'rsquo'    => 8217,
 255                 'sbquo'    => 8218,
 256                 'Scaron'   => 352,
 257                 'scaron'   => 353,
 258                 'sdot'     => 8901,
 259                 'sect'     => 167,
 260                 'shy'      => 173,
 261                 'Sigma'    => 931,
 262                 'sigma'    => 963,
 263                 'sigmaf'   => 962,
 264                 'sim'      => 8764,
 265                 'spades'   => 9824,
 266                 'sub'      => 8834,
 267                 'sube'     => 8838,
 268                 'sum'      => 8721,
 269                 'sup'      => 8835,
 270                 'sup1'     => 185,
 271                 'sup2'     => 178,
 272                 'sup3'     => 179,
 273                 'supe'     => 8839,
 274                 'szlig'    => 223,
 275                 'Tau'      => 932,
 276                 'tau'      => 964,
 277                 'there4'   => 8756,
 278                 'Theta'    => 920,
 279                 'theta'    => 952,
 280                 'thetasym' => 977,
 281                 'thinsp'   => 8201,
 282                 'THORN'    => 222,
 283                 'thorn'    => 254,
 284                 'tilde'    => 732,
 285                 'times'    => 215,
 286                 'trade'    => 8482,
 287                 'Uacute'   => 218,
 288                 'uacute'   => 250,
 289                 'uarr'     => 8593,
 290                 'uArr'     => 8657,
 291                 'Ucirc'    => 219,
 292                 'ucirc'    => 251,
 293                 'Ugrave'   => 217,
 294                 'ugrave'   => 249,
 295                 'uml'      => 168,
 296                 'upsih'    => 978,
 297                 'Upsilon'  => 933,
 298                 'upsilon'  => 965,
 299                 'Uuml'     => 220,
 300                 'uuml'     => 252,
 301                 'weierp'   => 8472,
 302                 'Xi'       => 926,
 303                 'xi'       => 958,
 304                 'Yacute'   => 221,
 305                 'yacute'   => 253,
 306                 'yen'      => 165,
 307                 'Yuml'     => 376,
 308                 'yuml'     => 255,
 309                 'Zeta'     => 918,
 310                 'zeta'     => 950,
 311                 'zwj'      => 8205,
 312                 'zwnj'     => 8204
 313         );
 314
 315         /**
 316          * Character entity aliases accepted by MediaWiki
 317          */
 318         static $htmlEntityAliases = array(
 319                 'רלמ' => 'rlm',
 320                 'رلم' => 'rlm',
 321         );
 322
 323         /**
 324          * Lazy-initialised attributes regex, see getAttribsRegex()
 325          */
 326         static $attribsRegex;
 327
 328         /**
 329          * Regular expression to match HTML/XML attribute pairs within a tag.
 330          * Allows some... latitude.
 331          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 332          */
 333         static function getAttribsRegex() {
 334                 if ( self::$attribsRegex === null ) {
 335                         $attribFirst = '[:A-Z_a-z0-9]';
 336                         $attrib = '[:A-Z_a-z-.0-9]';
 337                         $space = '[\x09\x0a\x0d\x20]';
 338                         self::$attribsRegex =
 339                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 340                                   ($space*=$space*
 341                                         (?:
 342                                          # The attribute value: quoted or alone
 343                                           \"([^<\"]*)\"
 344                                          | '([^<']*)'
 345                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 346                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 347                                                                                  # colors are specified like this.
 348                                                                                  # We'll be normalizing it.
 349                                         )
 350                                 )?(?=$space|\$)/sx";
 351                 }
 352                 return self::$attribsRegex;
 353         }
 354
 355         /**
 356          * Cleans up HTML, removes dangerous tags and attributes, and
 357          * removes HTML comments
 358          * @private
 359          * @param $text String
 360          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 361          * @param $args Array for the processing callback
 362          * @param $extratags Array for any extra tags to include
 363          * @param $removetags Array for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 367                 global $wgUseTidy, $wgHtml5, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 // Base our staticInitialised variable off of the global config state so that if the globals
 375                 // are changed (like in the secrewed up test system) we will re-initialise the settings.
 376                 $globalContext = implode( '-', compact( 'wgHtml5', 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 377                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 378
 379                         $htmlpairsStatic = array( # Tags that must be closed
 380                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 381                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 382                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 383                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 384                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 385                                 'kbd', 'samp'
 386                         );
 387                         if ( $wgHtml5 ) {
 388                                 $htmlpairsStatic = array_merge( $htmlpairsStatic, array( 'data', 'time' ) );
 389                         }
 390                         $htmlsingle = array(
 391                                 'br', 'hr', 'li', 'dt', 'dd'
 392                         );
 393                         $htmlsingleonly = array( # Elements that cannot have close tags
 394                                 'br', 'hr'
 395                         );
 396                         if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
 397                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 398                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 399                         }
 400                         $htmlnest = array( # Tags that can be nested--??
 401                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 402                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 403                         );
 404                         $tabletags = array( # Can only appear inside table, we will close them
 405                                 'td', 'th', 'tr',
 406                         );
 407                         $htmllist = array( # Tags used by list
 408                                 'ul','ol',
 409                         );
 410                         $listtags = array( # Tags that can appear in a list
 411                                 'li',
 412                         );
 413
 414                         if ( $wgAllowImageTag ) {
 415                                 $htmlsingle[] = 'img';
 416                                 $htmlsingleonly[] = 'img';
 417                         }
 418
 419                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 420                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 421
 422                         # Convert them all to hashtables for faster lookup
 423                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 424                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 425                         foreach ( $vars as $var ) {
 426                                 $$var = array_flip( $$var );
 427                         }
 428                         $staticInitialised = $globalContext;
 429                 }
 430                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 431                 $extratags = array_flip( $extratags );
 432                 $removetags = array_flip( $removetags );
 433                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 434                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 435
 436                 # Remove HTML comments
 437                 $text = Sanitizer::removeHTMLcomments( $text );
 438                 $bits = explode( '<', $text );
 439                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 440                 if ( !$wgUseTidy ) {
 441                         $tagstack = $tablestack = array();
 442                         foreach ( $bits as $x ) {
 443                                 $regs = array();
 444                                 # $slash: Does the current element start with a '/'?
 445                                 # $t: Current element name
 446                                 # $params: String between element name and >
 447                                 # $brace: Ending '>' or '/>'
 448                                 # $rest: Everything until the next element of $bits
 449                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 450                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 451                                 } else {
 452                                         $slash = $t = $params = $brace = $rest = null;
 453                                 }
 454
 455                                 $badtag = false;
 456                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 457                                         # Check our stack
 458                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 459                                                 $badtag = true;
 460                                         } elseif ( $slash ) {
 461                                                 # Closing a tag... is it the one we just opened?
 462                                                 $ot = @array_pop( $tagstack );
 463                                                 if ( $ot != $t ) {
 464                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 465                                                                 # Pop all elements with an optional close tag
 466                                                                 # and see if we find a match below them
 467                                                                 $optstack = array();
 468                                                                 array_push( $optstack, $ot );
 469                                                                 wfSuppressWarnings();
 470                                                                 $ot = array_pop( $tagstack );
 471                                                                 wfRestoreWarnings();
 472                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 473                                                                         array_push( $optstack, $ot );
 474                                                                         wfSuppressWarnings();
 475                                                                         $ot = array_pop( $tagstack );
 476                                                                         wfRestoreWarnings();
 477                                                                 }
 478                                                                 if ( $t != $ot ) {
 479                                                                         # No match. Push the optional elements back again
 480                                                                         $badtag = true;
 481                                                                         wfSuppressWarnings();
 482                                                                         $ot = array_pop( $optstack );
 483                                                                         wfRestoreWarnings();
 484                                                                         while ( $ot ) {
 485                                                                                 array_push( $tagstack, $ot );
 486                                                                                 wfSuppressWarnings();
 487                                                                                 $ot = array_pop( $optstack );
 488                                                                                 wfRestoreWarnings();
 489                                                                         }
 490                                                                 }
 491                                                         } else {
 492                                                                 @array_push( $tagstack, $ot );
 493                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 494                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 495                                                                         $badtag = true;
 496                                                                 }
 497                                                         }
 498                                                 } else {
 499                                                         if ( $t == 'table' ) {
 500                                                                 $tagstack = array_pop( $tablestack );
 501                                                         }
 502                                                 }
 503                                                 $newparams = '';
 504                                         } else {
 505                                                 # Keep track for later
 506                                                 if ( isset( $tabletags[$t] ) &&
 507                                                 !in_array( 'table', $tagstack ) ) {
 508                                                         $badtag = true;
 509                                                 } elseif ( in_array( $t, $tagstack ) &&
 510                                                 !isset( $htmlnest [$t ] ) ) {
 511                                                         $badtag = true;
 512                                                 # Is it a self closed htmlpair ? (bug 5487)
 513                                                 } elseif ( $brace == '/>' &&
 514                                                 isset( $htmlpairs[$t] ) ) {
 515                                                         $badtag = true;
 516                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 517                                                         # Hack to force empty tag for uncloseable elements
 518                                                         $brace = '/>';
 519                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 520                                                         # Hack to not close $htmlsingle tags
 521                                                         $brace = null;
 522                                                 } elseif ( isset( $tabletags[$t] )
 523                                                 && in_array( $t, $tagstack ) ) {
 524                                                         // New table tag but forgot to close the previous one
 525                                                         $text .= "</$t>";
 526                                                 } else {
 527                                                         if ( $t == 'table' ) {
 528                                                                 array_push( $tablestack, $tagstack );
 529                                                                 $tagstack = array();
 530                                                         }
 531                                                         array_push( $tagstack, $t );
 532                                                 }
 533
 534                                                 # Replace any variables or template parameters with
 535                                                 # plaintext results.
 536                                                 if( is_callable( $processCallback ) ) {
 537                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 538                                                 }
 539
 540                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 541                                                         $badtag = true;
 542                                                 }
 543
 544                                                 # Strip non-approved attributes from the tag
 545                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 546                                         }
 547                                         if ( !$badtag ) {
 548                                                 $rest = str_replace( '>', '&gt;', $rest );
 549                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 550                                                 $text .= "<$slash$t$newparams$close>$rest";
 551                                                 continue;
 552                                         }
 553                                 }
 554                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 555                         }
 556                         # Close off any remaining tags
 557                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 558                                 $text .= "</$t>\n";
 559                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 560                         }
 561                 } else {
 562                         # this might be possible using tidy itself
 563                         foreach ( $bits as $x ) {
 564                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 565                                 $x, $regs );
 566                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 567                                 $badtag = false;
 568                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 569                                         if( is_callable( $processCallback ) ) {
 570                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 571                                         }
 572
 573                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 574                                                 $badtag = true;
 575                                         }
 576
 577                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 578                                         if ( !$badtag ) {
 579                                                 $rest = str_replace( '>', '&gt;', $rest );
 580                                                 $text .= "<$slash$t$newparams$brace$rest";
 581                                                 continue;
 582                                         }
 583                                 }
 584                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 585                         }
 586                 }
 587                 wfProfileOut( __METHOD__ );
 588                 return $text;
 589         }
 590
 591         /**
 592          * Remove '<!--', '-->', and everything between.
 593          * To avoid leaving blank lines, when a comment is both preceded
 594          * and followed by a newline (ignoring spaces), trim leading and
 595          * trailing spaces and one of the newlines.
 596          *
 597          * @private
 598          * @param $text String
 599          * @return string
 600          */
 601         static function removeHTMLcomments( $text ) {
 602                 wfProfileIn( __METHOD__ );
 603                 while (($start = strpos($text, '<!--')) !== false) {
 604                         $end = strpos($text, '-->', $start + 4);
 605                         if ($end === false) {
 606                                 # Unterminated comment; bail out
 607                                 break;
 608                         }
 609
 610                         $end += 3;
 611
 612                         # Trim space and newline if the comment is both
 613                         # preceded and followed by a newline
 614                         $spaceStart = max($start - 1, 0);
 615                         $spaceLen = $end - $spaceStart;
 616                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 617                                 $spaceStart--;
 618                                 $spaceLen++;
 619                         }
 620                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 621                                 $spaceLen++;
 622                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 623                                 # Remove the comment, leading and trailing
 624                                 # spaces, and leave only one newline.
 625                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 626                         }
 627                         else {
 628                                 # Remove just the comment.
 629                                 $text = substr_replace($text, '', $start, $end - $start);
 630                         }
 631                 }
 632                 wfProfileOut( __METHOD__ );
 633                 return $text;
 634         }
 635
 636         /**
 637          * Take an array of attribute names and values and fix some deprecated values
 638          * for the given element type.
 639          * This does not validate properties, so you should ensure that you call
 640          * validateTagAttributes AFTER this to ensure that the resulting style rule
 641          * this may add is safe.
 642          *
 643          * - Converts most presentational attributes like align into inline css
 644          *
 645          * @param $attribs Array
 646          * @param $element String
 647          * @return Array
 648          */
 649         static function fixDeprecatedAttributes( $attribs, $element ) {
 650                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 651
 652                 // presentational attributes were removed from html5, we can leave them
 653                 // in when html5 is turned off
 654                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 655                         return $attribs;
 656                 }
 657
 658                 $table = array( 'table' );
 659                 $cells = array( 'td', 'th' );
 660                 $colls = array( 'col', 'colgroup' );
 661                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 662                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 663
 664                 $presentationalAttribs = array(
 665                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 666                         'clear' => array( 'clear', array( 'br' ) ),
 667                         'height' => array( 'height', $cells ),
 668                         'nowrap' => array( 'white-space', $cells ),
 669                         'size' => array( 'height', array( 'hr' ) ),
 670                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 671                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 672                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 673                 );
 674
 675                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 676                 foreach ( $attribs as $attribute => $value ) {
 677                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 678                                 $attribs[strtolower( $attribute )] = $value;
 679                                 unset( $attribs[$attribute] );
 680                         }
 681                 }
 682
 683                 $style = "";
 684                 foreach ( $presentationalAttribs as $attribute => $info ) {
 685                         list( $property, $elements ) = $info;
 686
 687                         // Skip if this attribute is not relevant to this element
 688                         if ( !in_array( $element, $elements ) ) {
 689                                 continue;
 690                         }
 691
 692                         // Skip if the attribute is not used
 693                         if ( !array_key_exists( $attribute, $attribs ) ) {
 694                                 continue;
 695                         }
 696
 697                         $value = $attribs[$attribute];
 698
 699                         // For nowrap the value should be nowrap instead of whatever text is in the value
 700                         if ( $attribute === 'nowrap' ) {
 701                                 $value = 'nowrap';
 702                         }
 703
 704                         // clear="all" is clear: both; in css
 705                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 706                                 $value = 'both';
 707                         }
 708
 709                         // Size based properties should have px applied to them if they have no unit
 710                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 711                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 712                                         $value = "{$value}px";
 713                                 }
 714                         }
 715
 716                         $style .= " $property: $value;";
 717
 718                         unset( $attribs[$attribute] );
 719                 }
 720
 721                 if ( $style ) {
 722                         // Prepend our style rules so that they can be overridden by user css
 723                         if ( isset($attribs['style']) ) {
 724                                 $style .= " " . $attribs['style'];
 725                         }
 726                         $attribs['style'] = trim($style);
 727                 }
 728
 729                 return $attribs;
 730         }
 731
 732         /**
 733          * Takes attribute names and values for a tag and the tah name and
 734          * validates that the tag is allowed to be present.
 735          * This DOES NOT validate the attributes, nor does it validate the
 736          * tags themselves. This method only handles the special circumstances
 737          * where we may want to allow a tag within content but ONLY when it has
 738          * specific attributes set.
 739          *
 740          * @param $
 741          */
 742         static function validateTag( $params, $element ) {
 743                 $params = Sanitizer::decodeTagAttributes( $params );
 744
 745                 if ( $element == 'meta' || $element == 'link' ) {
 746                         if ( !isset( $params['itemprop'] ) ) {
 747                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 748                                 return false;
 749                         }
 750                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 751                                 // <meta> must have a content="" for the itemprop
 752                                 return false;
 753                         }
 754                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 755                                 // <link> must have an associated href=""
 756                                 return false;
 757                         }
 758                 }
 759
 760                 return true;
 761         }
 762
 763         /**
 764          * Take an array of attribute names and values and normalize or discard
 765          * illegal values for the given element type.
 766          *
 767          * - Discards attributes not on a whitelist for the given element
 768          * - Unsafe style attributes are discarded
 769          * - Invalid id attributes are reencoded
 770          *
 771          * @param $attribs Array
 772          * @param $element String
 773          * @return Array
 774          *
 775          * @todo Check for legal values where the DTD limits things.
 776          * @todo Check for unique id attribute :P
 777          */
 778         static function validateTagAttributes( $attribs, $element ) {
 779                 return Sanitizer::validateAttributes( $attribs,
 780                         Sanitizer::attributeWhitelist( $element ) );
 781         }
 782
 783         /**
 784          * Take an array of attribute names and values and normalize or discard
 785          * illegal values for the given whitelist.
 786          *
 787          * - Discards attributes not the given whitelist
 788          * - Unsafe style attributes are discarded
 789          * - Invalid id attributes are reencoded
 790          *
 791          * @param $attribs Array
 792          * @param $whitelist Array: list of allowed attribute names
 793          * @return Array
 794          *
 795          * @todo Check for legal values where the DTD limits things.
 796          * @todo Check for unique id attribute :P
 797          */
 798         static function validateAttributes( $attribs, $whitelist ) {
 799                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 800
 801                 $whitelist = array_flip( $whitelist );
 802                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 803
 804                 $out = array();
 805                 foreach( $attribs as $attribute => $value ) {
 806                         #allow XML namespace declaration if RDFa is enabled
 807                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 808                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 809                                         $out[$attribute] = $value;
 810                                 }
 811
 812                                 continue;
 813                         }
 814
 815                         # Allow any attribute beginning with "data-", if in HTML5 mode
 816                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 817                                 continue;
 818                         }
 819
 820                         # Strip javascript "expression" from stylesheets.
 821                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 822                         if( $attribute == 'style' ) {
 823                                 $value = Sanitizer::checkCss( $value );
 824                         }
 825
 826                         if ( $attribute === 'id' ) {
 827                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 828                         }
 829
 830                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 831                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 832                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 833                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 834                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 835                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 836
 837                                 //Paranoia. Allow "simple" values but suppress javascript
 838                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 839                                         continue;
 840                                 }
 841                         }
 842
 843                         # NOTE: even though elements using href/src are not allowed directly, supply
 844                         #       validation code that can be used by tag hook handlers, etc
 845                         if ( $attribute === 'href' || $attribute === 'src' ) {
 846                                 if ( !preg_match( $hrefExp, $value ) ) {
 847                                         continue; //drop any href or src attributes not using an allowed protocol.
 848                                                   //NOTE: this also drops all relative URLs
 849                                 }
 850                         }
 851
 852                         // If this attribute was previously set, override it.
 853                         // Output should only have one attribute of each name.
 854                         $out[$attribute] = $value;
 855                 }
 856
 857                 if ( $wgAllowMicrodataAttributes ) {
 858                         # itemtype, itemid, itemref don't make sense without itemscope
 859                         if ( !array_key_exists( 'itemscope', $out ) ) {
 860                                 unset( $out['itemtype'] );
 861                                 unset( $out['itemid'] );
 862                                 unset( $out['itemref'] );
 863                         }
 864                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 865                 }
 866                 return $out;
 867         }
 868
 869         /**
 870          * Merge two sets of HTML attributes.  Conflicting items in the second set
 871          * will override those in the first, except for 'class' attributes which
 872          * will be combined (if they're both strings).
 873          *
 874          * @todo implement merging for other attributes such as style
 875          * @param $a Array
 876          * @param $b Array
 877          * @return array
 878          */
 879         static function mergeAttributes( $a, $b ) {
 880                 $out = array_merge( $a, $b );
 881                 if( isset( $a['class'] ) && isset( $b['class'] )
 882                 && is_string( $a['class'] ) && is_string( $b['class'] )
 883                 && $a['class'] !== $b['class'] ) {
 884                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 885                                 -1, PREG_SPLIT_NO_EMPTY );
 886                         $out['class'] = implode( ' ', array_unique( $classes ) );
 887                 }
 888                 return $out;
 889         }
 890
 891         /**
 892          * Pick apart some CSS and check it for forbidden or unsafe structures.
 893          * Returns a sanitized string. This sanitized string will have
 894          * character references and escape sequences decoded, and comments
 895          * stripped. If the input is just too evil, only a comment complaining
 896          * about evilness will be returned.
 897          *
 898          * Currently URL references, 'expression', 'tps' are forbidden.
 899          *
 900          * NOTE: Despite the fact that character references are decoded, the
 901          * returned string may contain character references given certain
 902          * clever input strings. These character references must
 903          * be escaped before the return value is embedded in HTML.
 904          *
 905          * @param $value String
 906          * @return String
 907          */
 908         static function checkCss( $value ) {
 909                 // Decode character references like &#123;
 910                 $value = Sanitizer::decodeCharReferences( $value );
 911
 912                 // Decode escape sequences and line continuation
 913                 // See the grammar in the CSS 2 spec, appendix D.
 914                 // This has to be done AFTER decoding character references.
 915                 // This means it isn't possible for this function to return
 916                 // unsanitized escape sequences. It is possible to manufacture
 917                 // input that contains character references that decode to
 918                 // escape sequences that decode to character references, but
 919                 // it's OK for the return value to contain character references
 920                 // because the caller is supposed to escape those anyway.
 921                 static $decodeRegex;
 922                 if ( !$decodeRegex ) {
 923                         $space = '[\\x20\\t\\r\\n\\f]';
 924                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 925                         $backslash = '\\\\';
 926                         $decodeRegex = "/ $backslash
 927                                 (?:
 928                                         ($nl) |  # 1. Line continuation
 929                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 930                                         (.) | # 3. backslash cancelling special meaning
 931                                         () | # 4. backslash at end of string
 932                                 )/xu";
 933                 }
 934                 $value = preg_replace_callback( $decodeRegex,
 935                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 936
 937                 // Remove any comments; IE gets token splitting wrong
 938                 // This must be done AFTER decoding character references and
 939                 // escape sequences, because those steps can introduce comments
 940                 // This step cannot introduce character references or escape
 941                 // sequences, because it replaces comments with spaces rather
 942                 // than removing them completely.
 943                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 944
 945                 // Remove anything after a comment-start token, to guard against
 946                 // incorrect client implementations.
 947                 $commentPos = strpos( $value, '/*' );
 948                 if ( $commentPos !== false ) {
 949                         $value = substr( $value, 0, $commentPos );
 950                 }
 951
 952                 // Reject problematic keywords and control characters
 953                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 954                         return '/* invalid control char */';
 955                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 956                         return '/* insecure input */';
 957                 }
 958                 return $value;
 959         }
 960
 961         /**
 962          * @param $matches array
 963          * @return String
 964          */
 965         static function cssDecodeCallback( $matches ) {
 966                 if ( $matches[1] !== '' ) {
 967                         // Line continuation
 968                         return '';
 969                 } elseif ( $matches[2] !== '' ) {
 970                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 971                 } elseif ( $matches[3] !== '' ) {
 972                         $char = $matches[3];
 973                 } else {
 974                         $char = '\\';
 975                 }
 976                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 977                         // These characters need to be escaped in strings
 978                         // Clean up the escape sequence to avoid parsing errors by clients
 979                         return '\\' . dechex( ord( $char ) ) . ' ';
 980                 } else {
 981                         // Decode unnecessary escape
 982                         return $char;
 983                 }
 984         }
 985
 986         /**
 987          * Take a tag soup fragment listing an HTML element's attributes
 988          * and normalize it to well-formed XML, discarding unwanted attributes.
 989          * Output is safe for further wikitext processing, with escaping of
 990          * values that could trigger problems.
 991          *
 992          * - Normalizes attribute names to lowercase
 993          * - Discards attributes not on a whitelist for the given element
 994          * - Turns broken or invalid entities into plaintext
 995          * - Double-quotes all attribute values
 996          * - Attributes without values are given the name as attribute
 997          * - Double attributes are discarded
 998          * - Unsafe style attributes are discarded
 999          * - Prepends space if there are attributes.
1000          *
1001          * @param $text String
1002          * @param $element String
1003          * @return String
1004          */
1005         static function fixTagAttributes( $text, $element ) {
1006                 if( trim( $text ) == '' ) {
1007                         return '';
1008                 }
1009
1010                 $decoded = Sanitizer::decodeTagAttributes( $text );
1011                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
1012                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1013
1014                 $attribs = array();
1015                 foreach( $stripped as $attribute => $value ) {
1016                         $encAttribute = htmlspecialchars( $attribute );
1017                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1018
1019                         $attribs[] = "$encAttribute=\"$encValue\"";
1020                 }
1021                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1022         }
1023
1024         /**
1025          * Encode an attribute value for HTML output.
1026          * @param $text String
1027          * @return HTML-encoded text fragment
1028          */
1029         static function encodeAttribute( $text ) {
1030                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1031
1032                 // Whitespace is normalized during attribute decoding,
1033                 // so if we've been passed non-spaces we must encode them
1034                 // ahead of time or they won't be preserved.
1035                 $encValue = strtr( $encValue, array(
1036                         "\n" => '&#10;',
1037                         "\r" => '&#13;',
1038                         "\t" => '&#9;',
1039                 ) );
1040
1041                 return $encValue;
1042         }
1043
1044         /**
1045          * Encode an attribute value for HTML tags, with extra armoring
1046          * against further wiki processing.
1047          * @param $text String
1048          * @return HTML-encoded text fragment
1049          */
1050         static function safeEncodeAttribute( $text ) {
1051                 $encValue = Sanitizer::encodeAttribute( $text );
1052
1053                 # Templates and links may be expanded in later parsing,
1054                 # creating invalid or dangerous output. Suppress this.
1055                 $encValue = strtr( $encValue, array(
1056                         '<'    => '&lt;',   // This should never happen,
1057                         '>'    => '&gt;',   // we've received invalid input
1058                         '"'    => '&quot;', // which should have been escaped.
1059                         '{'    => '&#123;',
1060                         '['    => '&#91;',
1061                         "''"   => '&#39;&#39;',
1062                         'ISBN' => '&#73;SBN',
1063                         'RFC'  => '&#82;FC',
1064                         'PMID' => '&#80;MID',
1065                         '|'    => '&#124;',
1066                         '__'   => '&#95;_',
1067                 ) );
1068
1069                 # Stupid hack
1070                 $encValue = preg_replace_callback(
1071                         '/(' . wfUrlProtocols() . ')/',
1072                         array( 'Sanitizer', 'armorLinksCallback' ),
1073                         $encValue );
1074                 return $encValue;
1075         }
1076
1077         /**
1078          * Given a value, escape it so that it can be used in an id attribute and
1079          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1080          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1081          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1082          * escaped with lots of dots.
1083          *
1084          * To ensure we don't have to bother escaping anything, we also strip ', ",
1085          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1086          * We also strip # because it upsets IE, and % because it could be
1087          * ambiguous if it's part of something that looks like a percent escape
1088          * (which don't work reliably in fragments cross-browser).
1089          *
1090          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1091          *                                                          in the id and
1092          *                                                          name attributes
1093          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1094          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1095          *   HTML5 definition of id attribute
1096          *
1097          * @param $id String: id to escape
1098          * @param $options Mixed: string or array of strings (default is array()):
1099          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1100          *       so don't pay attention if the first character isn't valid at the
1101          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1102          *       false.
1103          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1104          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1105          *       anchors and links won't break.
1106          * @return String
1107          */
1108         static function escapeId( $id, $options = array() ) {
1109                 global $wgHtml5, $wgExperimentalHtmlIds;
1110                 $options = (array)$options;
1111
1112                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1113                         $id = Sanitizer::decodeCharReferences( $id );
1114                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1115                         $id = trim( $id, '_' );
1116                         if ( $id === '' ) {
1117                                 # Must have been all whitespace to start with.
1118                                 return '_';
1119                         } else {
1120                                 return $id;
1121                         }
1122                 }
1123
1124                 # HTML4-style escaping
1125                 static $replace = array(
1126                         '%3A' => ':',
1127                         '%' => '.'
1128                 );
1129
1130                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1131                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1132
1133                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1134                 && !in_array( 'noninitial', $options ) )  {
1135                         // Initial character must be a letter!
1136                         $id = "x$id";
1137                 }
1138                 return $id;
1139         }
1140
1141         /**
1142          * Given a value, escape it so that it can be used as a CSS class and
1143          * return it.
1144          *
1145          * @todo For extra validity, input should be validated UTF-8.
1146          *
1147          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1148          *
1149          * @param $class String
1150          * @return String
1151          */
1152         static function escapeClass( $class ) {
1153                 // Convert ugly stuff to underscores and kill underscores in ugly places
1154                 return rtrim(preg_replace(
1155                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1156                         '_',
1157                         $class ), '_');
1158         }
1159
1160         /**
1161          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1162          * This allows (generally harmless) entities like &#160; to survive.
1163          *
1164          * @param $html String to escape
1165          * @return String: escaped input
1166          */
1167         static function escapeHtmlAllowEntities( $html ) {
1168                 $html = Sanitizer::decodeCharReferences( $html );
1169                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1170                 # hurt.
1171                 $html = htmlspecialchars( $html, ENT_QUOTES );
1172                 return $html;
1173         }
1174
1175         /**
1176          * Regex replace callback for armoring links against further processing.
1177          * @param $matches Array
1178          * @return string
1179          */
1180         private static function armorLinksCallback( $matches ) {
1181                 return str_replace( ':', '&#58;', $matches[1] );
1182         }
1183
1184         /**
1185          * Return an associative array of attribute names and values from
1186          * a partial tag string. Attribute names are forces to lowercase,
1187          * character references are decoded to UTF-8 text.
1188          *
1189          * @param $text String
1190          * @return Array
1191          */
1192         public static function decodeTagAttributes( $text ) {
1193                 if( trim( $text ) == '' ) {
1194                         return array();
1195                 }
1196
1197                 $attribs = array();
1198                 $pairs = array();
1199                 if( !preg_match_all(
1200                         self::getAttribsRegex(),
1201                         $text,
1202                         $pairs,
1203                         PREG_SET_ORDER ) ) {
1204                         return $attribs;
1205                 }
1206
1207                 foreach( $pairs as $set ) {
1208                         $attribute = strtolower( $set[1] );
1209                         $value = Sanitizer::getTagAttributeCallback( $set );
1210
1211                         // Normalize whitespace
1212                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1213                         $value = trim( $value );
1214
1215                         // Decode character references
1216                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1217                 }
1218                 return $attribs;
1219         }
1220
1221         /**
1222          * Pick the appropriate attribute value from a match set from the
1223          * attribs regex matches.
1224          *
1225          * @param $set Array
1226          * @return String
1227          */
1228         private static function getTagAttributeCallback( $set ) {
1229                 if( isset( $set[6] ) ) {
1230                         # Illegal #XXXXXX color with no quotes.
1231                         return $set[6];
1232                 } elseif( isset( $set[5] ) ) {
1233                         # No quotes.
1234                         return $set[5];
1235                 } elseif( isset( $set[4] ) ) {
1236                         # Single-quoted
1237                         return $set[4];
1238                 } elseif( isset( $set[3] ) ) {
1239                         # Double-quoted
1240                         return $set[3];
1241                 } elseif( !isset( $set[2] ) ) {
1242                         # In XHTML, attributes must have a value.
1243                         # For 'reduced' form, return explicitly the attribute name here.
1244                         return $set[1];
1245                 } else {
1246                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1247                 }
1248         }
1249
1250         /**
1251          * Normalize whitespace and character references in an XML source-
1252          * encoded text for an attribute value.
1253          *
1254          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1255          * but note that we're not returning the value, but are returning
1256          * XML source fragments that will be slapped into output.
1257          *
1258          * @param $text String
1259          * @return String
1260          */
1261         private static function normalizeAttributeValue( $text ) {
1262                 return str_replace( '"', '&quot;',
1263                         self::normalizeWhitespace(
1264                                 Sanitizer::normalizeCharReferences( $text ) ) );
1265         }
1266
1267         /**
1268          * @param $text string
1269          * @return mixed
1270          */
1271         private static function normalizeWhitespace( $text ) {
1272                 return preg_replace(
1273                         '/\r\n|[\x20\x0d\x0a\x09]/',
1274                         ' ',
1275                         $text );
1276         }
1277
1278         /**
1279          * Normalizes whitespace in a section name, such as might be returned
1280          * by Parser::stripSectionName(), for use in the id's that are used for
1281          * section links.
1282          *
1283          * @param $section String
1284          * @return String
1285          */
1286         static function normalizeSectionNameWhitespace( $section ) {
1287                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1288         }
1289
1290         /**
1291          * Ensure that any entities and character references are legal
1292          * for XML and XHTML specifically. Any stray bits will be
1293          * &amp;-escaped to result in a valid text fragment.
1294          *
1295          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1296          *   numericized (this way we're well-formed even without a DTD)
1297          * b. any numeric char refs must be legal chars, not invalid or forbidden
1298          * c. use &#x, not &#X
1299          * d. fix or reject non-valid attributes
1300          *
1301          * @param $text String
1302          * @return String
1303          * @private
1304          */
1305         static function normalizeCharReferences( $text ) {
1306                 return preg_replace_callback(
1307                         self::CHAR_REFS_REGEX,
1308                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1309                         $text );
1310         }
1311         /**
1312          * @param $matches String
1313          * @return String
1314          */
1315         static function normalizeCharReferencesCallback( $matches ) {
1316                 $ret = null;
1317                 if( $matches[1] != '' ) {
1318                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1319                 } elseif( $matches[2] != '' ) {
1320                         $ret = Sanitizer::decCharReference( $matches[2] );
1321                 } elseif( $matches[3] != ''  ) {
1322                         $ret = Sanitizer::hexCharReference( $matches[3] );
1323                 }
1324                 if( is_null( $ret ) ) {
1325                         return htmlspecialchars( $matches[0] );
1326                 } else {
1327                         return $ret;
1328                 }
1329         }
1330
1331         /**
1332          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1333          * return the equivalent numeric entity reference (except for the core &lt;
1334          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1335          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1336          * pseudo-entity source (eg &amp;foo;)
1337          *
1338          * @param $name String
1339          * @return String
1340          */
1341         static function normalizeEntity( $name ) {
1342                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1343                         return '&' . self::$htmlEntityAliases[$name] . ';';
1344                 } elseif ( in_array( $name,
1345                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1346                         return "&$name;";
1347                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1348                         return '&#' . self::$htmlEntities[$name] . ';';
1349                 } else {
1350                         return "&amp;$name;";
1351                 }
1352         }
1353
1354         /**
1355          * @param $codepoint
1356          * @return null|string
1357          */
1358         static function decCharReference( $codepoint ) {
1359                 $point = intval( $codepoint );
1360                 if( Sanitizer::validateCodepoint( $point ) ) {
1361                         return sprintf( '&#%d;', $point );
1362                 } else {
1363                         return null;
1364                 }
1365         }
1366
1367         /**
1368          * @param $codepoint
1369          * @return null|string
1370          */
1371         static function hexCharReference( $codepoint ) {
1372                 $point = hexdec( $codepoint );
1373                 if( Sanitizer::validateCodepoint( $point ) ) {
1374                         return sprintf( '&#x%x;', $point );
1375                 } else {
1376                         return null;
1377                 }
1378         }
1379
1380         /**
1381          * Returns true if a given Unicode codepoint is a valid character in XML.
1382          * @param $codepoint Integer
1383          * @return Boolean
1384          */
1385         private static function validateCodepoint( $codepoint ) {
1386                 return ($codepoint ==    0x09)
1387                         || ($codepoint ==    0x0a)
1388                         || ($codepoint ==    0x0d)
1389                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1390                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1391                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1392         }
1393
1394         /**
1395          * Decode any character references, numeric or named entities,
1396          * in the text and return a UTF-8 string.
1397          *
1398          * @param $text String
1399          * @return String
1400          */
1401         public static function decodeCharReferences( $text ) {
1402                 return preg_replace_callback(
1403                         self::CHAR_REFS_REGEX,
1404                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1405                         $text );
1406         }
1407
1408         /**
1409          * Decode any character references, numeric or named entities,
1410          * in the next and normalize the resulting string. (bug 14952)
1411          *
1412          * This is useful for page titles, not for text to be displayed,
1413          * MediaWiki allows HTML entities to escape normalization as a feature.
1414          *
1415          * @param $text String (already normalized, containing entities)
1416          * @return String (still normalized, without entities)
1417          */
1418         public static function decodeCharReferencesAndNormalize( $text ) {
1419                 global $wgContLang;
1420                 $text = preg_replace_callback(
1421                         self::CHAR_REFS_REGEX,
1422                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1423                         $text, /* limit */ -1, $count );
1424
1425                 if ( $count ) {
1426                         return $wgContLang->normalize( $text );
1427                 } else {
1428                         return $text;
1429                 }
1430         }
1431
1432         /**
1433          * @param $matches String
1434          * @return String
1435          */
1436         static function decodeCharReferencesCallback( $matches ) {
1437                 if( $matches[1] != '' ) {
1438                         return Sanitizer::decodeEntity( $matches[1] );
1439                 } elseif( $matches[2] != '' ) {
1440                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1441                 } elseif( $matches[3] != ''  ) {
1442                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1443                 }
1444                 # Last case should be an ampersand by itself
1445                 return $matches[0];
1446         }
1447
1448         /**
1449          * Return UTF-8 string for a codepoint if that is a valid
1450          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1451          * @param $codepoint Integer
1452          * @return String
1453          * @private
1454          */
1455         static function decodeChar( $codepoint ) {
1456                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1457                         return codepointToUtf8( $codepoint );
1458                 } else {
1459                         return UTF8_REPLACEMENT;
1460                 }
1461         }
1462
1463         /**
1464          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1465          * return the UTF-8 encoding of that character. Otherwise, returns
1466          * pseudo-entity source (eg &foo;)
1467          *
1468          * @param $name String
1469          * @return String
1470          */
1471         static function decodeEntity( $name ) {
1472                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1473                         $name = self::$htmlEntityAliases[$name];
1474                 }
1475                 if( isset( self::$htmlEntities[$name] ) ) {
1476                         return codepointToUtf8( self::$htmlEntities[$name] );
1477                 } else {
1478                         return "&$name;";
1479                 }
1480         }
1481
1482         /**
1483          * Fetch the whitelist of acceptable attributes for a given element name.
1484          *
1485          * @param $element String
1486          * @return Array
1487          */
1488         static function attributeWhitelist( $element ) {
1489                 $list = Sanitizer::setupAttributeWhitelist();
1490                 return isset( $list[$element] )
1491                         ? $list[$element]
1492                         : array();
1493         }
1494
1495         /**
1496          * Foreach array key (an allowed HTML element), return an array
1497          * of allowed attributes
1498          * @return Array
1499          */
1500         static function setupAttributeWhitelist() {
1501                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1502
1503                 static $whitelist, $staticInitialised;
1504                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1505
1506                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1507                         return $whitelist;
1508                 }
1509
1510                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1511
1512                 if ( $wgAllowRdfaAttributes ) {
1513                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1514                         $common = array_merge( $common, array(
1515                             'about', 'property', 'resource', 'datatype', 'typeof',
1516                         ) );
1517                 }
1518
1519                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1520                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1521                         $common = array_merge( $common, array(
1522                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1523                         ) );
1524                 }
1525
1526                 $block = array_merge( $common, array( 'align' ) );
1527                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1528                 $tablecell = array( 'abbr',
1529                                     'axis',
1530                                     'headers',
1531                                     'scope',
1532                                     'rowspan',
1533                                     'colspan',
1534                                     'nowrap', # deprecated
1535                                     'width',  # deprecated
1536                                     'height', # deprecated
1537                                     'bgcolor' # deprecated
1538                                     );
1539
1540                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1541                 # See: http://www.w3.org/TR/html4/
1542                 $whitelist = array(
1543                         # 7.5.4
1544                         'div'        => $block,
1545                         'center'     => $common, # deprecated
1546                         'span'       => $block, # ??
1547
1548                         # 7.5.5
1549                         'h1'         => $block,
1550                         'h2'         => $block,
1551                         'h3'         => $block,
1552                         'h4'         => $block,
1553                         'h5'         => $block,
1554                         'h6'         => $block,
1555
1556                         # 7.5.6
1557                         # address
1558
1559                         # 8.2.4
1560                         # bdo
1561
1562                         # 9.2.1
1563                         'em'         => $common,
1564                         'strong'     => $common,
1565                         'cite'       => $common,
1566                         'dfn'        => $common,
1567                         'code'       => $common,
1568                         'samp'       => $common,
1569                         'kbd'        => $common,
1570                         'var'        => $common,
1571                         'abbr'       => $common,
1572                         # acronym
1573
1574                         # 9.2.2
1575                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1576                         # q
1577
1578                         # 9.2.3
1579                         'sub'        => $common,
1580                         'sup'        => $common,
1581
1582                         # 9.3.1
1583                         'p'          => $block,
1584
1585                         # 9.3.2
1586                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1587
1588                         # 9.3.4
1589                         'pre'        => array_merge( $common, array( 'width' ) ),
1590
1591                         # 9.4
1592                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1593                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1594
1595                         # 10.2
1596                         'ul'         => array_merge( $common, array( 'type' ) ),
1597                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1598                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1599
1600                         # 10.3
1601                         'dl'         => $common,
1602                         'dd'         => $common,
1603                         'dt'         => $common,
1604
1605                         # 11.2.1
1606                         'table'      => array_merge( $common,
1607                                                                 array( 'summary', 'width', 'border', 'frame',
1608                                                                                 'rules', 'cellspacing', 'cellpadding',
1609                                                                                 'align', 'bgcolor',
1610                                                                 ) ),
1611
1612                         # 11.2.2
1613                         'caption'    => array_merge( $common, array( 'align' ) ),
1614
1615                         # 11.2.3
1616                         'thead'      => array_merge( $common, $tablealign ),
1617                         'tfoot'      => array_merge( $common, $tablealign ),
1618                         'tbody'      => array_merge( $common, $tablealign ),
1619
1620                         # 11.2.4
1621                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1622                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1623
1624                         # 11.2.5
1625                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1626
1627                         # 11.2.6
1628                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1629                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1630
1631                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1632                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1633
1634                         # 13.2
1635                         # Not usually allowed, but may be used for extension-style hooks
1636                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1637                         # true
1638                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1639
1640                         # 15.2.1
1641                         'tt'         => $common,
1642                         'b'          => $common,
1643                         'i'          => $common,
1644                         'big'        => $common,
1645                         'small'      => $common,
1646                         'strike'     => $common,
1647                         's'          => $common,
1648                         'u'          => $common,
1649
1650                         # 15.2.2
1651                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1652                         # basefont
1653
1654                         # 15.3
1655                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1656
1657                         # XHTML Ruby annotation text module, simple ruby only.
1658                         # http://www.w3c.org/TR/ruby/
1659                         'ruby'       => $common,
1660                         # rbc
1661                         # rtc
1662                         'rb'         => $common,
1663                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1664                         'rp'         => $common,
1665
1666                         # MathML root element, where used for extensions
1667                         # 'title' may not be 100% valid here; it's XHTML
1668                         # http://www.w3.org/TR/REC-MathML/
1669                         'math'       => array( 'class', 'style', 'id', 'title' ),
1670                 );
1671
1672                 if ( $wgHtml5 ) {
1673                         # HTML5 elements, defined by:
1674                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1675                         $whitelist += array(
1676                                 'data' => array_merge( $common, array( 'value' ) ),
1677                                 'time' => array_merge( $common, array( 'datetime' ) ),
1678
1679                                 // meta and link are only present when Microdata is allowed anyways
1680                                 // so we don't bother adding another condition here
1681                                 // meta and link are only valid for use as Microdata so we do not
1682                                 // allow the common attributes here.
1683                                 'meta' => array( 'itemprop', 'content' ),
1684                                 'link' => array( 'itemprop', 'href' ),
1685                         );
1686                 }
1687
1688                 $staticInitialised = $globalContext;
1689
1690                 return $whitelist;
1691         }
1692
1693         /**
1694          * Take a fragment of (potentially invalid) HTML and return
1695          * a version with any tags removed, encoded as plain text.
1696          *
1697          * Warning: this return value must be further escaped for literal
1698          * inclusion in HTML output as of 1.10!
1699          *
1700          * @param $text String: HTML fragment
1701          * @return String
1702          */
1703         static function stripAllTags( $text ) {
1704                 # Actual <tags>
1705                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1706
1707                 # Normalize &entities and whitespace
1708                 $text = self::decodeCharReferences( $text );
1709                 $text = self::normalizeWhitespace( $text );
1710
1711                 return $text;
1712         }
1713
1714         /**
1715          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1716          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1717          * PHP 5.1 doesn't.
1718          *
1719          * Use for passing XHTML fragments to PHP's XML parsing functions
1720          *
1721          * @return String
1722          */
1723         static function hackDocType() {
1724                 $out = "<!DOCTYPE html [\n";
1725                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1726                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1727                 }
1728                 $out .= "]>\n";
1729                 return $out;
1730         }
1731
1732         /**
1733          * @param $url string
1734          * @return mixed|string
1735          */
1736         static function cleanUrl( $url ) {
1737                 # Normalize any HTML entities in input. They will be
1738                 # re-escaped by makeExternalLink().
1739                 $url = Sanitizer::decodeCharReferences( $url );
1740
1741                 # Escape any control characters introduced by the above step
1742                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1743                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1744
1745                 # Validate hostname portion
1746                 $matches = array();
1747                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1748                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1749
1750                         // Characters that will be ignored in IDNs.
1751                         // http://tools.ietf.org/html/3454#section-3.1
1752                         // Strip them before further processing so blacklists and such work.
1753                         $strip = "/
1754                                 \\s|          # general whitespace
1755                                 \xc2\xad|     # 00ad SOFT HYPHEN
1756                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1757                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1758                                 \xe2\x81\xa0| # 2060 WORD JOINER
1759                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1760                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1761                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1762                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1763                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1764                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1765                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1766                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1767                                 /xuD";
1768
1769                         $host = preg_replace( $strip, '', $host );
1770
1771                         // @todo FIXME: Validate hostnames here
1772
1773                         return $protocol . $host . $rest;
1774                 } else {
1775                         return $url;
1776                 }
1777         }
1778
1779         /**
1780          * @param $matches array
1781          * @return string
1782          */
1783         static function cleanUrlCallback( $matches ) {
1784                 return urlencode( $matches[0] );
1785         }
1786
1787         /**
1788          * Does a string look like an e-mail address?
1789          *
1790          * This validates an email address using an HTML5 specification found at:
1791          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1792          * Which as of 2011-01-24 says:
1793          *
1794          *   A valid e-mail address is a string that matches the ABNF production
1795          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1796          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1797          *   3.5.
1798          *
1799          * This function is an implementation of the specification as requested in
1800          * bug 22449.
1801          *
1802          * Client-side forms will use the same standard validation rules via JS or
1803          * HTML 5 validation; additional restrictions can be enforced server-side
1804          * by extensions via the 'isValidEmailAddr' hook.
1805          *
1806          * Note that this validation doesn't 100% match RFC 2822, but is believed
1807          * to be liberal enough for wide use. Some invalid addresses will still
1808          * pass validation here.
1809          *
1810          * @since 1.18
1811          *
1812          * @param $addr String E-mail address
1813          * @return Bool
1814          */
1815         public static function validateEmail( $addr ) {
1816                 $result = null;
1817                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1818                         return $result;
1819                 }
1820
1821                 // Please note strings below are enclosed in brackets [], this make the
1822                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1823                 // See bug 26948
1824                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1825                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1826
1827                 $HTML5_email_regexp = "/
1828                 ^                      # start of string
1829                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1830                 @                      # 'apostrophe'
1831                 [$rfc1034_ldh_str]+       # First domain part
1832                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1833                 $                      # End of string
1834                 /ix" ; // case Insensitive, eXtended
1835
1836                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1837         }
1838 }