The following query can cause a schema mismatch problem between join node and its both child scan nodes. The main cause that '##' col1 are evaluated in both ScanNode. Because a schema cannot include the same column names, one col1 is omitted. So, it causes the schema mismatch problem.
This is a critical bug. So, I've uploaded the urgent workaround bug fix. Later, I'll fix it in more valid and stable manner. I already created the Jira issue (
TAJO-895) for that way.