Catch exception in link normalizer on broken unicode sequence

lib/common/utils.js

@@ -132,6 +132,13 @@ function escapeHtml(str) {
 
 // Incoming link can be partially encoded. Convert possible combinations to
 // unified form.
+//
+// TODO: Rewrite it. Should use:
+//
+// - encodeURIComponent for query
+// - encodeURI for path
+// - (?) punicode for domain mame (but encodeURI seems to work in real world)
+//
 function normalizeLink(url) {
   var normalized = replaceEntities(url);
 
@@ -141,7 +148,14 @@ function normalizeLink(url) {
     normalized = decodeURI(normalized);
   } catch (__) {}
 
-  return encodeURI(normalized);
+  // Encoder throws exception on broken unicode sequence.
+  // Kill suspicious data for the safety.
+  //
+  try {
+    return encodeURI(normalized);
+  } catch (__) {
+    return '';
+  }
 }
 
 ////////////////////////////////////////////////////////////////////////////////