[
https://issues.apache.org/jira/browse/CASSANDRA-10590?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14978473#comment-14978473
]
Sylvain Lebresne commented on CASSANDRA-10590:
----------------------------------------------
That's definitively my bad for missing that test, sorry. The issue has to do
with the fact that the legacy deserializer {{hasNext()}} method deserialize the
full row upfront, and the fact that a row can span multiple index block (again,
in the legacy case). The patch for CASSANDRA-10360 accounted for that but
missed one part, so that we were potentially stopping reading early because a
call to {{hasNext()}} put us on an index block where we knew not row we were
interested in could start, but that made us mistakenly skip the next row (the
one read in the {{hasNext()}} call). Anyway, pushed a patch to fix this
[here|https://github.com/pcmanus/cassandra/commits/10590].
> LegacySSTableTest fails after CASSANDRA-10360
> ---------------------------------------------
>
> Key: CASSANDRA-10590
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10590
> Project: Cassandra
> Issue Type: Bug
> Reporter: Robert Stupp
> Assignee: Sylvain Lebresne
> Fix For: 3.0.0
>
>
> LegacySSTableTest fails reading pre-3.0 sstables (versions {{jb}}, {{ka}},
> {{la}}) with clustering keys and counters.
> [First failing 3.0 testall
> build|http://cassci.datastax.com/job/cassandra-3.0_testall/205/]
> /cc [~slebresne]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
