[
https://issues.apache.org/jira/browse/LUCENE-9501?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Julie Tibshirani updated LUCENE-9501:
-------------------------------------
Description:
In LUCENE-7714 we added a new query to sandbox called
IndexSortSortedNumericDocValuesRangeQuery that optimizes range calculations
when the field is sorted. The query has a bad bug: its DocIdSetIterator can
return an old value for docID() even after advance has returned NO_MORE_DOCS.
This violates the DocIdSetIterator contract and means that it's possible for
DocIdSetIterator#advance to be called when it's already been exhausted (which
can result in invalid reads).
We would have expected this issue to be caught in tests, especially because
classes like AssertingIndexSearcher check for these invariants. As part of this
fix I'll look into improvements to the Asserting* wrapper framework.
was:
In LUCENE-7714 added a new query to sandbox called
IndexSortSortedNumericDocValuesRangeQuery that optimizes range calculations
when the field is sorted. The query has a bad bug: its DocIdSetIterator can
return an old value for docID() even after advance has returned NO_MORE_DOCS.
This violates the DocIdSetIterator contract and means that it's possible for
DocIdSetIterator#advance to be called when it's already been exhausted (which
can result in invalid reads).
We would have expected this issue to be caught in tests, especially because
classes like AssertingIndexSearcher check for these invariants. As part of this
fix I'll look into improvements to the Asserting* wrapper framework.
> IndexSortSortedNumericDocValuesRangeQuery violates iterator invariant.
> ----------------------------------------------------------------------
>
> Key: LUCENE-9501
> URL: https://issues.apache.org/jira/browse/LUCENE-9501
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Julie Tibshirani
> Priority: Major
>
> In LUCENE-7714 we added a new query to sandbox called
> IndexSortSortedNumericDocValuesRangeQuery that optimizes range calculations
> when the field is sorted. The query has a bad bug: its DocIdSetIterator can
> return an old value for docID() even after advance has returned NO_MORE_DOCS.
> This violates the DocIdSetIterator contract and means that it's possible for
> DocIdSetIterator#advance to be called when it's already been exhausted (which
> can result in invalid reads).
> We would have expected this issue to be caught in tests, especially because
> classes like AssertingIndexSearcher check for these invariants. As part of
> this fix I'll look into improvements to the Asserting* wrapper framework.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
