Restrict character class for charset capture in Util\ParseUrl

- Address https://github.com/friendica/friendica/issues/9252#issuecomment-698742359

diff --git a/src/Util/ParseUrl.php b/src/Util/ParseUrl.php
index ce2a0ea9669591242651081d6c7d24c64e369fcf..1596e015bef1415fb1858cd3d45565eaf0dbf1ed 100644 (file)
--- a/src/Util/ParseUrl.php
+++ b/src/Util/ParseUrl.php
@@ -204,7 +204,7 @@ class ParseUrl
                $charset = '';
                // Look for a charset, first in headers
                // Expected form: Content-Type: text/html; charset=ISO-8859-4
-               if (preg_match('/charset=(.+?)\s/', $header, $matches)) {
+               if (preg_match('/charset=([a-z0-9-_.\/]+)/i', $header, $matches)) {
                        $charset = trim(trim(trim(array_pop($matches)), ';,'));
                }
 
@@ -212,14 +212,18 @@ class ParseUrl
                // Expected forms:
                // - <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
                // - <meta charset="utf-8">
+               // - <meta charset=utf-8>
+               // - <meta charSet="utf-8">
                // We escape <style> and <script> tags since they can contain irrelevant charset information
                // (see https://github.com/friendica/friendica/issues/9251#issuecomment-698636806)
                Strings::performWithEscapedBlocks($body, '#<(?:style|script).*?</(?:style|script)>#ism', function ($body) use (&$charset) {
-                       if (preg_match('/charset=["\']?([^\',"]*?)[\'"]/', $body, $matches)) {
+                       if (preg_match('/charset=["\']?([a-z0-9-_.\/]+)/i', $body, $matches)) {
                                $charset = trim(trim(trim(array_pop($matches)), ';,'));
                        }
                });
 
+               $siteinfo['charset'] = $charset;
+
                if ($charset && strtoupper($charset) != 'UTF-8') {
                        // See https://github.com/friendica/friendica/issues/5470#issuecomment-418351211
                        $charset = str_ireplace('latin-1', 'latin1', $charset);