[
https://issues.apache.org/jira/browse/LUCENE-10472?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17508982#comment-17508982
]
ASF subversion and git services commented on LUCENE-10472:
----------------------------------------------------------
Commit bb7568d865ca5f6932c705f1ae3b5adb159ae8ec in lucene's branch
refs/heads/main from Luca Cavanna
[ https://gitbox.apache.org/repos/asf?p=lucene.git;h=bb7568d ]
LUCENE-10472: Fix TestMatchAllDocsQuery#testEarlyTermination (#753)
As part of #716 I moved the test to use a collector manager, but I forgot to
update one of the assertions.
We can't rely on totalHits being accurate when the search is executed my
multiple threads and early terminated.
> TestMatchAllDocsQuery#testEarlyTermination fails total hits assertion
> ---------------------------------------------------------------------
>
> Key: LUCENE-10472
> URL: https://issues.apache.org/jira/browse/LUCENE-10472
> Project: Lucene - Core
> Issue Type: Task
> Reporter: Julie Tibshirani
> Priority: Minor
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> Very rarely TestMatchAllDocsQuery can fail with an assertion error:
> {code}
> > java.lang.AssertionError: expected:<201> but was:<241>
> > at
> __randomizedtesting.SeedInfo.seed([B4F2350E137110E3:7B6574512B1D80E0]:0)
> > at org.junit.Assert.fail(Assert.java:89)
> > at org.junit.Assert.failNotEquals(Assert.java:835)
> > at org.junit.Assert.assertEquals(Assert.java:647)
> > at org.junit.Assert.assertEquals(Assert.java:633)
> > at
> org.apache.lucene.search.TestMatchAllDocsQuery.testEarlyTermination(TestMatchAllDocsQuery.java:124)
> {code}
> We are expecting the totalHits.value to be exactly totalHitsThreshold + 1,
> but it is sometimes more:
> {code}
> assertEquals(totalHitsThreshold + 1, topDocs.totalHits.value);
> assertEquals(TotalHits.Relation.GREATER_THAN_OR_EQUAL_TO,
> topDocs.totalHits.relation);
> {code}
> The failures only seem to happen when we choose to use multiple threads in
> IndexSearcher. I was able to reproduce this regularly by forcing
> IndexSearcher to use multiple threads and running the test 100 times. This
> started failing after https://github.com/apache/lucene/pull/716, where we
> updated tests to use a collector manager instead of a collector, which
> introduced the possibility of multiple threads.
> I am not sure this actually indicates a bug, or if the test just needs to be
> tweaked. It seems okay to return more than totalHitsThreshold + 1 sometimes,
> that doesn't seem to violate any contract?
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]