[ 
https://issues.apache.org/jira/browse/CASSANDRA-10471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14958024#comment-14958024
 ] 

Ariel Weisberg edited comment on CASSANDRA-10471 at 10/14/15 11:50 PM:
-----------------------------------------------------------------------

bq. But that's not true at least for empty IN queries on older nodes,
As a reviewer I didn't figure out how to verify that this statement is true or 
why. I mucked about with StatementRestrictions and family and found where IN 
restrictions are expressed, but it's all pretty big in scope. Do you have any 
pointers?

I don't get the difference between null vs PartitionColumns.NONE either. As far 
as I can tell null isn't used as a signal anywhere although maybe it was in the 
past? canSkipValue might have depended on it before this change. Maybe also 
serialization which saves writing the empty selection.

I verified the dtest is fixed, and that your fix matches your description and 
makes the assertion that was removed always true.

There is no unit test for ColumnFilter expressing these expected behaviors and 
configurations of ColumnFilter and ColumnFilter.Builder.


was (Author: aweisberg):
bq. But that's not true at least for empty IN queries on older nodes,
As a reviewer I didn't figure out how to verify that this statement is true or 
why. I mucked about with StatementRestrictions and family and found where IN 
restrictions are expressed, but it's all pretty big in scope. Do you have any 
pointers?

I don't get the difference between null vs PartitionColumns.NONE

I verified the dtest is fixed, and that your fix matches your description and 
makes the assertion that was removed always true.

There is no unit test for ColumnFilter expressing these expected behaviors and 
configurations of ColumnFilter and ColumnFilter.Builder.

> fix flapping empty_in_test dtest
> --------------------------------
>
>                 Key: CASSANDRA-10471
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-10471
>             Project: Cassandra
>          Issue Type: Sub-task
>            Reporter: Jim Witschey
>            Assignee: Sylvain Lebresne
>             Fix For: 3.0.0 rc2
>
>
> {{upgrade_tests/cql_tests.py:TestCQL.empty_in_test}} fails about half the 
> time on the upgrade path from 2.2 to 3.0:
> http://cassci.datastax.com/view/Upgrades/job/storage_engine_upgrade_dtest-22_tarball-30_HEAD/42/testReport/upgrade_tests.cql_tests/TestCQL/empty_in_test/history/
> Once [this dtest PR|https://github.com/riptano/cassandra-dtest/pull/586] is 
> merged, these tests should also run with this upgrade path on normal 3.0 
> jobs. Until then, you can run it with the following command:
> {code}
> SKIP=false CASSANDRA_VERSION=binary:2.2.0 UPGRADE_TO=git:cassandra-3.0 
> nosetests 2>&1 upgrade_tests/cql_tests.py:TestCQL.empty_in_test
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to