HTML files created for converting online submissions to pdf for feedback/annotation (unoconv) are wrongly encoded (See attachment).

To reproduce (unoconv must be configured correctly on your moodle instance) :

• As a teacher, create an online text assignment
• As a student, submit a text with special characters like 'éèÉàÀ'
• As the teacher, go to the grade form of this student's assignment.
  You should see that the text in the annotation is wrongly encoded

Dug the code a bit and found out what the problem is :

• The method called strip_image in "/mod/assign/feedback/editpdf/classes/document_services.php" is using DOMDocument::loadHTML which causes the problem.
• It seems that by default, DOMDOcument::loadHTML treat the string as being in ISO-8859-1 and not UTF-8. See http://stackoverflow.com/questions/8218230/php-domdocument-loadhtml-not-encoding-utf-8-correctly

Function format_text in "lib/weblib.php" is doing one of the proposed solution by appending "'<?xml version="1.0" encoding="UTF-8" ?>' ." to the string, to force loadHTML to work in UTF-8.

The following code modif would fix the issue :

diff --git a/mod/assign/feedback/editpdf/classes/document_services.php b/mod/assign/feedback/editpdf/classes/document_services.php

index 4ada0a7..e348bc5 100644

--- a/mod/assign/feedback/editpdf/classes/document_services.php

+++ b/mod/assign/feedback/editpdf/classes/document_services.php

@@ -119,7 +119,7 @@ EOD;

      */

     protected static function strip_images($html) {

         $dom = new DOMDocument();

-        $dom->loadHTML($html);

+        $dom->loadHTML('<?xml version="1.0" encoding="UTF-8" ?>' . $html);

         $images = $dom->getElementsByTagName('img');

         $i = 0;

@@ -135,7 +135,7 @@ EOD;

             $text = $dom->createTextNode($replacement);

             $node->parentNode->replaceChild($text, $node);

         }

-        return $dom->saveHTML();

+        return trim(preg_replace('~<(?:!DOCTYPE|/?(?:html|body))[^>]*>\s*~i', '', $dom->saveHTML($dom->documentElement)));

     }

     /**