Merge pull request #5599 from edx/adam/fix-matlab-invalid-char

gracefully fail when parsing xml with non-compatible string (TNL-347)

common/lib/capa/capa/responsetypes.py

@@ -2111,7 +2111,13 @@ class CodeResponse(LoncapaResponse):
         except etree.XMLSyntaxError as _err:
             # If `html` contains attrs with no values, like `controls` in <audio controls src='smth'/>,
             # XML parser will raise exception, so wee fallback to html5parser, which will set empty "" values for such attrs.
-            parsed = html5lib.parseFragment(msg, treebuilder='lxml', namespaceHTMLElements=False)
+            try:
+                parsed = html5lib.parseFragment(msg, treebuilder='lxml', namespaceHTMLElements=False)
+            except ValueError:
+                # the parsed message might contain strings that are not
+                # xml compatible, in which case, throw the error message
+                parsed = False
+
             if not parsed:
                 log.error("Unable to parse external grader message as valid"
                       " XML: score_msg['msg']=%s", msg)

common/lib/capa/capa/tests/test_responsetypes.py

@@ -1006,6 +1006,7 @@ class CodeResponseTest(ResponseTest):
 
         invalid_grader_msgs = [
             '<audio',  # invalid XML and HTML5
+            '<p>\b</p>',  # invalid special character
         ]
 
         answer_ids = sorted(self.problem.get_question_answers())