[jira] [Updated] (CASSANDRA-3084) o.a.c.dht.Range.differenceToFetch() doesn't handle all cases correctly

["Tyler Hobbs (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CASSANDRA-3084?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Tyler Hobbs updated CASSANDRA-3084:
-----------------------------------

    Attachment: 3084.txt
                3084-unit-test.txt

3084.txt covers all cases of possible range differences.

3084-unit-test.txt provides unit tests that fail before the patch.

The logic is definitely a bit hairy here, so any thoughts on how to simplify it are welcome.

> o.a.c.dht.Range.differenceToFetch() doesn't handle all cases correctly
> ----------------------------------------------------------------------
>
>                 Key: CASSANDRA-3084
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-3084
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.8.4
>            Reporter: Tyler Hobbs
>         Attachments: 3084-unit-test.txt, 3084.txt
>
>
> It's possible that differenceToFetch is making implicit assumptions about the relationship between the two ranges, but the following cases are not handled correctly (the old range is (A, B], the new is (C, D]:
> {noformat}
> --C--A-----B--D--
> {noformat}
> Here, the result will be (C, A] and (D, B], instead of (C, A] and (B, D].
> {noformat}
> --C--A-----D--B--
> {noformat}
> The result will be (C, D] instead of just (C, A].
> {noformat}
> --A--C-----D--B--
> {noformat}
> The result will be (B, D] when nothing needs to be transfered.
> If there is some kind of implicit assumption that these cases won't arise, it either needs to be explicit (assertions, exceptions) or the cases need to be handled.  It should be easy to cover this with unit tests.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira