[ https://issues.apache.org/jira/browse/CASSANDRA-15178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16869388#comment-16869388 ]
Sam Tunnicliffe commented on CASSANDRA-15178: --------------------------------------------- The linked branches revert the original CASSANDRA-15086 patch (but retain the test), and include an alternative fix whereby the illegal legacy cells are read from disk as normal, but get filtered out during the collation of LegacyCells into cql Rows. ||branch||CI|| |[15178-3.0|https://github.com/beobal/cassandra/tree/15178-3.0]|[circle|https://circleci.com/gh/beobal/workflows/cassandra/tree/cci%2F15178-3.0]| |[15178-3.11|https://github.com/beobal/cassandra/tree/15178-3.11]|[circle|https://circleci.com/gh/beobal/workflows/cassandra/tree/cci%2F15178-3.11]| > Skipping illegal legacy cells can break reverse iteration of indexed > partitions > ------------------------------------------------------------------------------- > > Key: CASSANDRA-15178 > URL: https://issues.apache.org/jira/browse/CASSANDRA-15178 > Project: Cassandra > Issue Type: Bug > Components: Legacy/Local Write-Read Paths > Reporter: Sam Tunnicliffe > Assignee: Sam Tunnicliffe > Priority: Normal > > The fix for CASSANDRA-15086 interacts badly with the accounting of bytes read > from disk when indexed partitions are read in reverse. The skipped columns > can cause the tracking of where CQL rows span index block boundaries to be > incorrectly calculated, leading to rows being missing from read results. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org