Bad CodepointIterator::_next Error from MarkLogic Server

Getting a "Bad CodepointIterator::_next" error from a query in MarkLogic? Check you configuration at the app server level.

We were seeing this error on our production instance, but not our development instance. In our ASP.NET app (through XCC) we would see a generic message, but when we ran the query in cq we saw the more specific version. After a wasted day or so trying to track down the root cause, we thought we had it narrowed down to UTF-8 characters, but we couldn't explain why it was fine on one server but not the other. Then on a whim I compared the configurations looking for anything out of sync and noticed that the "collation" field on the server where we had the error was set to "http://marklogic.com/collation" while on our development instance it was "http://marklogic.com/collation/codepoint". We set them both to use /codepoint and the problem was solved.

If you're using both the HTTP and XDBC layers, make sure to check them both. We fixed it on our HTTP layer, which eliminated the bug in cq, but we were still getting it in the ASP.NET application until we changed the setting in both.