[
https://issues.apache.org/jira/browse/KUDU-565?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15494594#comment-15494594
]
Jean-Daniel Cryans commented on KUDU-565:
-----------------------------------------
[~danburkert] you think your work on non-covering range partitions covers this
perceived lack of test coverage?
> Tests for client where keys fall on tablet boundaries
> -----------------------------------------------------
>
> Key: KUDU-565
> URL: https://issues.apache.org/jira/browse/KUDU-565
> Project: Kudu
> Issue Type: Task
> Components: client, test
> Affects Versions: M5
> Reporter: Todd Lipcon
> Priority: Critical
>
> I think we may have some bugs when a row falls exactly on a tablet boundary.
> Most of our correctness testing uses auto-generated split keys in a big key
> space, but this means it's unlikely we insert something exactly on the
> boundary. Looking at the code, I think there's some sloppiness where in some
> cases we consider upper bounds to be inclusive (eg on scans) and in other
> places exclusive (eg on tablet ranges).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
