[
https://issues.apache.org/jira/browse/CASSANDRA-8778?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14328242#comment-14328242
]
Jeff Jirsa commented on CASSANDRA-8778:
---------------------------------------
I know that [~jbellis]] suggested pushing this to 2.0, but 2.0 calculates sha1
digests (when needed), and does so correctly. The real issue I intended to
report in #8778 is that CASSANDRA-4165 incorrectly calculated adler32 digests
in 2.1 and 3.0/trunk
If someone wants to port #4165 to 2.0, I won't really object, but the digests
calculated by 2.0 are technically correct when present. In 2.1+, they were
incorrect.
I have, however, uploaded a new .diff (rather than .patch) if you prefer that
format. I've confirmed this applies cleanly to both 2.1 and trunk.
The one thing I did neglect is that the digest is still named .sha1 rather than
.adler32. That change remains in my patch for #5791, and probably belongs here.
Trivial change, though, I can add it or you can.
> Adler32 Digests for Compressed SSTables are Incorrect
> -----------------------------------------------------
>
> Key: CASSANDRA-8778
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8778
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jeff Jirsa
> Assignee: Jeff Jirsa
> Priority: Trivial
> Attachments: cassandra-8778.diff, cassandra-8778.patch
>
>
> ChecksumWriter append() calculates the checksum of a buffer, then also writes
> that checksum as an int into the provided DataOutput. The final call to
> writeFullChecksum will have the checksum of all of the buffers, but none of
> the intermediate int checksums.
> This issue was discovered (and a fix proposed) in #5791 - if the patch for
> #5791 is not accepted, some fix for
> src/java/org/apache/cassandra/io/util/DataIntegrityMetadata.java should be
> applied.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
