[FIX] tools: prevent html_translate crash with malformed HTML

If an user stores malformed HTML code in a *translatable HTML field*,
the ORM won't prevent it (this could be a fix for master as it breaks
existing behaviours) and it could be rendered anyway by browsers.

When reading/prefetching the record, the translation function was
crashing because malformed HTML is obviously malformed XML (the
translation function uses XML internals).

This fix will return the source value as a fallback if the HTML field is
malformed (and thus not translatable) + log the error.

openerp/tools/translate.py

@@ -305,12 +305,16 @@ def html_translate(callback, value):
     if not value:
         return value
 
-    parser = etree.HTMLParser(encoding='utf-8')
-    trans = XMLTranslator(callback, 'html', parser)
-    wrapped = "<div>%s</div>" % encode(value)
-    root = etree.fromstring(wrapped, parser)
-    trans.process(root[0][0])               # html > body > div
-    return trans.get_done()[5:-6]           # remove tags <div> and </div>
+    try:
+        parser = etree.HTMLParser(encoding='utf-8')
+        trans = XMLTranslator(callback, 'html', parser)
+        wrapped = "<div>%s</div>" % encode(value)
+        root = etree.fromstring(wrapped, parser)
+        trans.process(root[0][0])               # html > body > div
+        value = trans.get_done()[5:-6]           # remove tags <div> and </div>
+    except ValueError:
+        _logger.exception("Cannot translate malformed HTML, using source value instead")
+    return value
 
 
 #