[
https://issues.apache.org/jira/browse/CASSANDRA-21276?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18112521#comment-18112521
]
Niyam Siwach commented on CASSANDRA-21276:
------------------------------------------
Hi, following up on this issue. The changes are limited to test portability and
stability on s390x:
- wait for asynchronous materialized-view mutations in ViewPKTest
- use little-endian buffers consistently with Memory semantics
- increase the helper-thread stack for the recursive partition-index test on
s390x
Could you please review? I’m also happy address any suggested changes.
> Fix s390x test failures in ViewPKTest, PartitionIndexTest, and MemoryTest on
> s390x
> ----------------------------------------------------------------------------------
>
> Key: CASSANDRA-21276
> URL: https://issues.apache.org/jira/browse/CASSANDRA-21276
> Project: Apache Cassandra
> Issue Type: Bug
> Components: Test/unit
> Reporter: Niyam Siwach
> Assignee: Niyam Siwach
> Priority: Normal
> Attachments: 21276-unit-tests.patch
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> ViewPKTest.testNullInClusteringColumns performs base-table mutations with
> execute / executeNet, but materialized view updates are asynchronous.
> Switching these statements to updateView() makes the test wait for pending
> view mutations and avoids race-driven failures on slower s390x runs.
> PartitionIndexTest creates a helper thread with a 32 KB stack, which is too
> small on s390x and can fail in this environment. Increasing the stack size to
> 256 KB makes the test stable.
> MemoryTest currently relies on native-endian buffers, but Memory is defined
> as little-endian. On big-endian s390x, the test must compare against
> LITTLE_ENDIAN buffers explicitly, otherwise it validates the wrong byte order.
> This will be a test-only fix to make the existing test suite behave correctly
> on s390x without changing production behavior.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
