[
https://issues.apache.org/jira/browse/HIVE-15649?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Shelukhin updated HIVE-15649:
------------------------------------
Attachment: HIVE-15649.patch
Added a test too, however on MiniLlap/Local it doesn't repro the issue that
we've seen on some cluster where analyze table ... for columns resulted in null
columnIds.
[~prasanth_j] can you take a look?
[~mmccline] FYI I think some schema evolution paths may also assume non-null
column list... e.g. buildConversionFileTypesArray seems to assume in isOk path
that readerIncludes are not null even though they could be.
> LLAP IO may NPE on all-column read
> ----------------------------------
>
> Key: HIVE-15649
> URL: https://issues.apache.org/jira/browse/HIVE-15649
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
> Attachments: HIVE-15649.patch
>
>
> It seems like very few paths use READ_ALL_COLUMNS config, but some do. LLAP
> IO doesn't account for that.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
