[ https://issues.apache.org/jira/browse/CASSANDRA-15872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17136209#comment-17136209 ]
Caleb Rackliffe commented on CASSANDRA-15872: --------------------------------------------- [~jasonstack] It looks like the fix should address the race between {{SyncTask}} completion and the assertion, but I think we can do this without modifying {{RepairJob}}. One alternative is that we could take {{MeasureableRepairSession}}, provide it a callback (perhaps allow it to be set int the test body), override {{syncComplete()}}, and hit the callback from that before we call {{super.syncComplete()}}. That would allow us to await on a latch in there or even directly make the assertion. (My assumption is that there will be at least one remote job.) Making the test a bit more complex while simplifying the production code seems reasonable here. Thoughts? > Flaky test: testNoTreesRetainedAfterDifference - > org.apache.cassandra.repair.RepairJobTest > ------------------------------------------------------------------------------------------ > > Key: CASSANDRA-15872 > URL: https://issues.apache.org/jira/browse/CASSANDRA-15872 > Project: Cassandra > Issue Type: Task > Components: Consistency/Repair > Reporter: ZhaoYang > Assignee: ZhaoYang > Priority: Normal > Fix For: 4.0-rc > > > link: [https://circleci.com/gh/maedhroz/cassandra/21#tests/containers/95] > > -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org