[
https://issues.apache.org/jira/browse/CASSANDRA-1367?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12923504#action_12923504
]
Jonathan Ellis commented on CASSANDRA-1367:
-------------------------------------------
MurmurHash is broken, it should be using data.get(index) (it also shouldn't
take a separate length value since BB knows its length).
the use of position() scares me, it's a bug waiting to happen, e.g. in
FBUtilities.hash
block.position()+block.arrayOffset()
all of these these should all just be block.arrayOffset().
Similarly remaining() scares me. We don't "use up" our ByteBuffers on purpose
except in very unusual cases (e.g. your BBUtil.getLong), all? of these should
be limit - offset instead.
I _suspect_ that there is a bug from position/remaining causing the test
failure: it's building the index on the test rows and being rejected at the
row-level bloom filter saying "this row doesn't exist" which is completely
bogus.
style: space after commas and between operators please.
Let's fix the above and see where that gets us.
> Upgrade to Thrift 0.5.0
> -----------------------
>
> Key: CASSANDRA-1367
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1367
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Johan Oskarsson
> Priority: Minor
> Fix For: 0.8
>
> Attachments: 1367_v5.patch, libthrift-0.5.jar
>
>
> There's finally a new thrift release out. This gives us a chance to
> standardize on a release instead of just a revision of thrift trunk.
> http://www.apache.org/dist/incubator/thrift/0.4.0-incubating/thrift-0.4.0.tar.gz
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
