There is no reason to use this in HTML in addition to the lang attribute. No spec requires it (AFAIK) and it's completely redundant.
174 lines
4.5 KiB
PHP
174 lines
4.5 KiB
PHP
<?php
|
|
|
|
/**
|
|
* Standard output handler for use with ob_start
|
|
*/
|
|
function wfOutputHandler( $s ) {
|
|
global $wgDisableOutputCompression, $wgValidateAllHtml;
|
|
$s = wfMangleFlashPolicy( $s );
|
|
if ( $wgValidateAllHtml ) {
|
|
$headers = apache_response_headers();
|
|
$isHTML = true;
|
|
foreach ( $headers as $name => $value ) {
|
|
if ( strtolower( $name ) == 'content-type' && strpos( $value, 'text/html' ) === false && strpos( $value, 'application/xhtml+xml' ) === false ) {
|
|
$isHTML = false;
|
|
break;
|
|
}
|
|
}
|
|
if ( $isHTML ) {
|
|
$s = wfHtmlValidationHandler( $s );
|
|
}
|
|
}
|
|
if ( !$wgDisableOutputCompression && !ini_get( 'zlib.output_compression' ) ) {
|
|
if ( !defined( 'MW_NO_OUTPUT_COMPRESSION' ) ) {
|
|
$s = wfGzipHandler( $s );
|
|
}
|
|
if ( !ini_get( 'output_handler' ) ) {
|
|
wfDoContentLength( strlen( $s ) );
|
|
}
|
|
}
|
|
return $s;
|
|
}
|
|
|
|
/**
|
|
* Get the "file extension" that some client apps will estimate from
|
|
* the currently-requested URL.
|
|
* This isn't on WebRequest because we need it when things aren't initialized
|
|
* @private
|
|
*/
|
|
function wfRequestExtension() {
|
|
/// @todo Fixme: this sort of dupes some code in WebRequest::getRequestUrl()
|
|
if( isset( $_SERVER['REQUEST_URI'] ) ) {
|
|
// Strip the query string...
|
|
list( $path ) = explode( '?', $_SERVER['REQUEST_URI'], 2 );
|
|
} elseif( isset( $_SERVER['SCRIPT_NAME'] ) ) {
|
|
// Probably IIS. QUERY_STRING appears separately.
|
|
$path = $_SERVER['SCRIPT_NAME'];
|
|
} else {
|
|
// Can't get the path from the server? :(
|
|
return '';
|
|
}
|
|
|
|
$period = strrpos( $path, '.' );
|
|
if( $period !== false ) {
|
|
return strtolower( substr( $path, $period ) );
|
|
}
|
|
return '';
|
|
}
|
|
|
|
/**
|
|
* Handler that compresses data with gzip if allowed by the Accept header.
|
|
* Unlike ob_gzhandler, it works for HEAD requests too.
|
|
*/
|
|
function wfGzipHandler( $s ) {
|
|
if( !function_exists( 'gzencode' ) || headers_sent() ) {
|
|
return $s;
|
|
}
|
|
|
|
$ext = wfRequestExtension();
|
|
if( $ext == '.gz' || $ext == '.tgz' ) {
|
|
// Don't do gzip compression if the URL path ends in .gz or .tgz
|
|
// This confuses Safari and triggers a download of the page,
|
|
// even though it's pretty clearly labeled as viewable HTML.
|
|
// Bad Safari! Bad!
|
|
return $s;
|
|
}
|
|
|
|
if( wfClientAcceptsGzip() ) {
|
|
header( 'Content-Encoding: gzip' );
|
|
$s = gzencode( $s, 6 );
|
|
}
|
|
|
|
// Set vary header if it hasn't been set already
|
|
$headers = headers_list();
|
|
$foundVary = false;
|
|
foreach ( $headers as $header ) {
|
|
if ( substr( $header, 0, 5 ) == 'Vary:' ) {
|
|
$foundVary = true;
|
|
break;
|
|
}
|
|
}
|
|
if ( !$foundVary ) {
|
|
header( 'Vary: Accept-Encoding' );
|
|
header( 'X-Vary-Options: Accept-Encoding;list-contains=gzip' );
|
|
}
|
|
return $s;
|
|
}
|
|
|
|
/**
|
|
* Mangle flash policy tags which open up the site to XSS attacks.
|
|
*/
|
|
function wfMangleFlashPolicy( $s ) {
|
|
# Avoid weird excessive memory usage in PCRE on big articles
|
|
if ( preg_match( '/\<\s*cross-domain-policy\s*\>/i', $s ) ) {
|
|
return preg_replace( '/\<\s*cross-domain-policy\s*\>/i', '<NOT-cross-domain-policy>', $s );
|
|
} else {
|
|
return $s;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Add a Content-Length header if possible. This makes it cooperate with squid better.
|
|
*/
|
|
function wfDoContentLength( $length ) {
|
|
if ( !headers_sent() && $_SERVER['SERVER_PROTOCOL'] == 'HTTP/1.0' ) {
|
|
header( "Content-Length: $length" );
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Replace the output with an error if the HTML is not valid
|
|
*/
|
|
function wfHtmlValidationHandler( $s ) {
|
|
|
|
$errors = '';
|
|
if ( MWTidy::checkErrors( $s, $errors ) ) {
|
|
return $s;
|
|
}
|
|
|
|
header( 'Cache-Control: no-cache' );
|
|
|
|
$out = <<<EOT
|
|
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
|
|
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" dir="ltr">
|
|
<head>
|
|
<title>HTML validation error</title>
|
|
<style>
|
|
.highlight { background-color: #ffc }
|
|
li { white-space: pre }
|
|
</style>
|
|
</head>
|
|
<body>
|
|
<h1>HTML validation error</h1>
|
|
<ul>
|
|
EOT;
|
|
|
|
$error = strtok( $errors, "\n" );
|
|
$badLines = array();
|
|
while ( $error !== false ) {
|
|
if ( preg_match( '/^line (\d+)/', $error, $m ) ) {
|
|
$lineNum = intval( $m[1] );
|
|
$badLines[$lineNum] = true;
|
|
$out .= "<li><a href=\"#line-{$lineNum}\">" . htmlspecialchars( $error ) . "</a></li>\n";
|
|
}
|
|
$error = strtok( "\n" );
|
|
}
|
|
|
|
$out .= '</ul>';
|
|
$out .= '<pre>' . htmlspecialchars( $errors ) . '</pre>';
|
|
$out .= "<ol>\n";
|
|
$line = strtok( $s, "\n" );
|
|
$i = 1;
|
|
while ( $line !== false ) {
|
|
if ( isset( $badLines[$i] ) ) {
|
|
$out .= "<li class=\"highlight\" id=\"line-$i\">";
|
|
} else {
|
|
$out .= '<li>';
|
|
}
|
|
$out .= htmlspecialchars( $line ) . "</li>\n";
|
|
$line = strtok( "\n" );
|
|
$i++;
|
|
}
|
|
$out .= '</ol></body></html>';
|
|
return $out;
|
|
}
|