[
https://issues.apache.org/jira/browse/SOLR-18624?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
ASF GitHub Bot updated SOLR-18624:
----------------------------------
Labels: pull-request-available (was: )
> Flaky TestMultiWordSynonyms.testPf failure with tied scores and no explicit
> sort
> --------------------------------------------------------------------------------
>
> Key: SOLR-18624
> URL: https://issues.apache.org/jira/browse/SOLR-18624
> Project: Solr
> Issue Type: Bug
> Components: Tests
> Reporter: Nick Shanin
> Priority: Major
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> h2. Observed failure
> {code}
> org.apache.solr.search.TestMultiWordSynonyms.testPf
> Expected first document: 10
> Actual result order: 13, 12, 11, 10
> Scores: 7.7074103 for all four documents
> {code}
> The request does not specify an explicit secondary sort. In this failing run,
> the compared documents have equal scores, so their relative order is not
> stabilized by the request; the assertion nevertheless expects document 10
> first.
> h2. Test context
> This test was added under SOLR-12243 for eDisMax phrase queries involving
> multiword synonyms. It is intended to verify that phrase proximity can
> outweigh {{bf=boost_d}}.
> The same failure predates PR 5291, and PR 5291 does not change this test.
> Develocity reports 1 flaky run in 273 executions. This ticket records the
> observed flakiness; it does not establish a regression in phrase-query
> behavior or attribute the failure to PR 5291.
> h2. Investigation direction
> Determine how to keep the test's phrase-proximity-versus-boost coverage while
> avoiding reliance on the incidental order of equal-score results, or
> otherwise make the tie-break deterministic if that ordering is part of the
> intended behavior.
> SOLR-17976 concerns a separate SolrCloud distributed-search shard tie-break
> issue; this report is specifically about {{TestMultiWordSynonyms.testPf}}.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
