[ https://issues.apache.org/jira/browse/CASSANDRA-7279?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sylvain Lebresne updated CASSANDRA-7279: ---------------------------------------- Attachment: 7279_alternative.txt Attaching a patch for how I would do this. It only deoverlap/sort slices for thrift, but does validate (in an assertion) the slices for CQL "to avoid whatever future bug everyone seems to assume we'll introduce". I'll note that said validation is more thorough than what the de-overlapping method would protect against since it validates that each slices in the proper order (while the de-overlapping method would have silently do the wrong thing in that case, potentially hiding a bug deeper rather than protecting against one). The patch also fix 2 bugs in the handling of the empty finish bounds by the de-overlapping function: 1) in the initial sorting, it is the finish bound that should be special cased, the start bound is automatically handled by the comparator, and 2) later the finish bound also needs special casing when testing for inclusion. Both are covered by additional unit tests. > MultiSliceTest.test_with_overlap* unit tests failing in trunk > ------------------------------------------------------------- > > Key: CASSANDRA-7279 > URL: https://issues.apache.org/jira/browse/CASSANDRA-7279 > Project: Cassandra > Issue Type: Bug > Components: Tests > Reporter: Michael Shuler > Assignee: T Jake Luciani > Priority: Minor > Fix For: 2.1 rc1, 3.0 > > Attachments: 7279-trunk.txt, 7279-trunkv2.txt, 7279-trunkv3.txt, > 7279-trunkv4.txt, 7279_alternative.txt > > > Example: > https://cassci.datastax.com/job/trunk_utest/623/testReport/org.apache.cassandra.thrift/MultiSliceTest/ -- This message was sent by Atlassian JIRA (v6.2#6252)