Cause:
1. ExeMgr asks PrimProc to scan the dict data for matches
2. PrimProc return LBID of matches
3. ExeMgr generates a token based on LBID and offset of matches
4. We then add these tokens to the result set. If missing they silently fail as they could be deleted entries.

Part 3/4 is where this fails. We also have a block count in the token now, so the generated token won't match a BLOB/TEXT entry using multiple blocks.

Pull request in develop and regression suite.

pDictionaryScan does the steps outlined in my first comment. This is triggered on a WHERE condition with table that has a single row. pDictionaryScan will never work for BLOB/TEXT in its current form so the patch enforces pDictionaryStep for BLOB/TEXT instead. This is the changes to dbcon/joblist/jlf_execplantojoblist.cpp around filterWithDictionary.

In addition this patch adds BLOB/TEXT support for more of joblist, fixing problems such as error 202 during a join in an UPDATE query with BLOB/TEXT.

Build verified: Github source

[root@localhost mariadb-columnstore-server]# git show
commit 0a6aea3145f55cebffd86fdb6717284a83fbd2b7
Merge: ef372bf d9dbe2f
Author: david hill <david.hill@mariadb.com>
Date: Thu Apr 6 13:56:52 2017 -0500

Merge pull request #42 from mariadb-corporation/MCOL-377

Make audit plugin display columnstore original query in logging

[root@localhost mariadb-columnstore-server]# cd mariadb-columnstore-engine/
[root@localhost mariadb-columnstore-engine]# git show
commit f240c89faf2e2f8a7e27a7d2f9f1ebd2d6eeed90
Merge: f8f045f 6128293
Author: dhall-InfiniDB <david.hall@mariadb.com>
Date: Fri Apr 21 10:06:34 2017 -0500

Merge pull request #156 from mariadb-corporation/MCOL-671

MCOL-671 Fix TEXT/BLOB single row SELECT WHERE

[root@localhost mariadb-columnstore-engine]#