]
Steven Hawkins commented on TEIID-5585:
---------------------------------------
The issue is actually with the auto-cleanup performed by the CommandContext tracking of
created InputStreamFactories - they do not last from one statement to the next. So we
need temporary tables to assume responsibility for any such lobs.
Insert of a disk backed temporary lob into a temp table is
un-readable
----------------------------------------------------------------------
Key: TEIID-5585
URL:
https://issues.jboss.org/browse/TEIID-5585
Project: Teiid
Issue Type: Bug
Components: Query Engine
Reporter: Steven Hawkins
Assignee: Steven Hawkins
Priority: Major
Fix For: 12.0
This can be reproduced with a script like:
insert into #temp select 1 as x, cast(repeat('1', 4000) as clob) y;
insert into #temp select 2, concat((select y from #temp where x = 1), (select y from
#temp where x = 1));
insert into #temp select 3, concat((select y from #temp where x = 2), (select y from
#temp where x = 2));
insert into #temp select 4, concat((select y from #temp where x = 3), (select y from
#temp where x = 3));
insert into #temp select 5, concat((select y from #temp where x = 4), (select y from
#temp where x = 4));
-- will fail
insert into #temp select 6, concat((select y from #temp where x = 5), (select y from
#temp where x = 5))
The #5 clob is unreadable as the backing filestore is removed after insert.