[
https://issues.apache.org/jira/browse/CASSANDRA-15560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17096295#comment-17096295
]
Berenguer Blasi commented on CASSANDRA-15560:
---------------------------------------------
[~brandon.williams] thank you for reviewing this. Indeed the dtests were very
noisy and I was worried about it so any guidance is welcomed. That particular
test I ran locally and it passed before I pushed. I tested again and it passes
in my branch:
{quote}git branch
* CASSANDRA-15560-4.0
trunk
===Flaky Test Report===
test_disable_compression_cql passed 1 out of the required 1 times. Success!
test_compression_cql_options passed 1 out of the required 1 times. Success!
test_compression_cql_disabled_with_alter passed 1 out of the required 1 times.
Success!
test_compression_cql_enabled_with_alter passed 1 out of the required 1 times.
Success!
===End Flaky Test Report===
{quote}
but fails on your branch and even the commit before mine
{quote}git branch
* (HEAD detached at f0526fa368)
CASSANDRA-15560-4.0
trunk
===Flaky Test Report===
test_disable_compression_cql passed 1 out of the required 1 times. Success!
test_compression_cql_options failed and was not selected for rerun.
<class 'AssertionError'>
assert 'DEFLATE' == 'UNKNOWN'
- DEFLATE
+ UNKNOWN
[<TracebackEntry /home/bereng/work/repos/bdpWS/dtests/compression_test.py:112>]
test_compression_cql_disabled_with_alter passed 1 out of the required 1 times.
Success!
test_compression_cql_enabled_with_alter passed 1 out of the required 1 times.
Success!{quote}
I bisected it to
[9c1bbf3ac913f9bdf7a0e0922106804af42d2c1e|https://github.com/apache/cassandra/commit/9c1bbf3ac913f9bdf7a0e0922106804af42d2c1e]
CASSANDRA-15379. The dtests in there, if I managed to make myself around
circleci correctly, don't seem to have ran the compression test.
Also I looked at your CI results and tried to match them to
[jenkins|https://ci-cassandra.apache.org/job/Cassandra-trunk-dtest-novnode/]
but the latest runs seem to have failed. I did run locally some of the failing
tests against trunck and they also failed. I _think_ we're safe to merge this
PR and open a new ticket for the compression test failure due to 15379. But my
circleci/dtests reasonable failures knowledge is very limited. Wdyt?
> Change io.compressor.LZ4Compressor to LZ4SafeDecompressor
> ---------------------------------------------------------
>
> Key: CASSANDRA-15560
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15560
> Project: Cassandra
> Issue Type: Improvement
> Components: Feature/Compression
> Reporter: Jordan West
> Assignee: Berenguer Blasi
> Priority: Normal
> Fix For: 4.0, 4.0-rc
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> CASSANDRA-15556 and related tickets showed that LZ4FastDecompressor can crash
> the JVM and that LZ4SafeDecompressor performs better w/o the crash risk — its
> also not deprecated. While we protect ourselves by checksumming the
> compressed data but that doesn’t mean we should leave deprecated code that
> can segfault the jvm (providing a potential DDOS vector among other things)
> in crucial places like io.compress.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
