Normalize inserted text using UNorm

For some reason, the client was sending the server a Unicode-normalized
version of inserted strings. So if for example we inserted the string
'ä' (i.e. \x61\xCC\x88) into the document, what would be sent to the
server would be 'ä' (i.e. \xC3\xA4).

This wouldn't be a problem on its own. BUT JavaScript reports that the
length of the first string is 2, while the length of the second one is
1.

So the command that was being sent to the server was 'Z:1>2*0+1$ä', when
it should really be 'Z:1>1*0+1$ä'. When the `checkRep` method checks the
length of the inserted string, it finds an inconsistency, and
disconnects the client.

We now normalize the inserted string before the command is generated, so
the length is always correct.

1 files changed, 2 insertions, 1 deletions

diff --git a/src/static/js/contentcollector.js b/src/static/js/contentcollector.js
index 777b3421..b27dfc5e 100644
--- a/src/static/js/contentcollector.js
+++ b/src/static/js/contentcollector.js
@@ -25,13 +25,14 @@
 
 var _MAX_LIST_LEVEL = 8;
 
+var UNorm = require('./unorm');
 var Changeset = require('./Changeset');
 var hooks = require('./pluginfw/hooks');
 var _ = require('./underscore');
 
 function sanitizeUnicode(s)
 {
-  return s.replace(/[\uffff\ufffe\ufeff\ufdd0-\ufdef\ud800-\udfff]/g, '?');
+  return UNorm.nfc(s).replace(/[\uffff\ufffe\ufeff\ufdd0-\ufdef\ud800-\udfff]/g, '?');
 }
 
 function makeContentCollector(collectStyles, browser, apool, domInterface, className2Author)