Bulk NumericDocValues#longValues (new in 10.5.0) throws on a valid dense segment for one exact request size per byte-aligned width, because its size guard checks the unpadded byte count against Integer.MAX_VALUE and then adds the padding bytes in int (40/48-bit: wraps to Integer.MIN_VALUE, 'read past EOF'; 16/24/56-bit: a byte[] above the JVM limit)
Finding LFV-381Area: codecs / doc values (Lucene90DocValuesProducer dense byte-aligned bulk decode, NumericDocValues#longValues)Priority: lowImpact: crashNovelty: new
AI-assisted finding, confirmed by automated checks, not yet triaged by a human Lucene committer. This site files nothing upstream.
What it shows
You index a dense numeric doc-values field that the default codec stores as one fixed 40-bit width, in a segment of 429,496,729 documents, and read every value with a single NumericDocValues#longValues call, expecting what the advanceExact/longValue loop in its javadoc returns. Instead Lucene90DocValuesProducer.bulkDecodeByteAlignedValues checks only the unpadded byte count (5 x 429,496,729 = 2,147,483,645) against Integer.MAX_VALUE and then adds 3 padding bytes in int arithmetic, which wraps to Integer.MIN_VALUE; the negative length passes the slice and allocation checks, and the call throws EOFException 'read past EOF (pos=0)' on a valid file (ArrayIndexOutOfBoundsException with NIOFS or an in-heap directory). The same wrap happens at 357,913,941 docs for 48-bit values, and at one size each for 16, 24 and 56 bits the padded count asks for a byte[] longer than the JVM allows (OutOfMemoryError 'Requested array size exceeds VM limit'), while the per-document loop and a batch one document smaller return every value.
Expected vs actual
- Expected:
longValuesreturns all 357,913,941 values, as the per-document loop does. - Actual:
EOFException: read past EOF (pos=0)on the valid.dvd, thrown fromLucene90DocValuesProducer.bulkDecodeByteAlignedValues:691, with assertions on and off.
Files
Run it
- Tested on Lucene main e1faeba08bde (2026-11-02): still reproduces.
- Tested on Lucene main ea57360912ed (2026-10-24): reproduced.
- How to read the outcome: a failing test is the finding reproducing (the failure message states expected against actual).
- Seeds: a rerun with the same seed replays the same random choices. Assertions off (-da -dsa -Dtests.asserts=false) is the production-like setting.
# Lucene checkout at the tested commit, built with: ./gradlew --no-daemon assemble (JDK 25)
LUCENE=/path/to/lucene; V=12.0.0-SNAPSHOT
CP=$LUCENE/lucene/core/build/libs/lucene-core-$V.jar
CP=$CP:$LUCENE/lucene/codecs/build/libs/lucene-codecs-$V.jar
CP=$CP:$LUCENE/lucene/test-framework/build/libs/lucene-test-framework-$V.jar
CP=$CP:randomizedtesting-runner-2.9.2.jar:junit-4.13.2.jar:hamcrest-3.0.jar # test-framework dependencies (Maven Central)
CP=$CP:junit-platform-console-standalone-6.1.3.jar # JUnit Platform console launcher (Maven Central)
javac -d classes -cp "$CP" $(find src -name '*.java')
# run 1: assertions off (production-like), seed B0B94
java -da -dsa -Dtests.asserts=false -Xmx12g --add-modules jdk.incubator.vector --enable-native-access=ALL-UNNAMED -Dtests.monster=true -Dbulkdecode.dir=heap -Dtests.seed=B0B94 -Dtests.method=test48BitsWholeSegment \
-cp "classes:$CP" org.junit.platform.console.ConsoleLauncher \
execute --select-class org.apache.lucene.index.TestNumericBulkDecodePaddedSizeGradle alternative: copy the test into the matching module's src/test/ tree of the checkout and run ./gradlew -p lucene/<module> test --tests <class> -Ptests.seed=<seed> (add -Ptests.asserts=false for a run without assertions).
Files shown on the finding page
Formal models and scripts this finding's part of the hunt used: numeric-dv-bulk-values
Feedback
Open an issue in lucene-ai-bug-hunt. Please do not open issues or pull requests in apache/lucene about this finding unless a Lucene committer agrees it should be filed.
Snapshot of 2026-11-04. AI-assisted; not yet triaged by a human Lucene committer. Feedback: open an issue in lucene-ai-bug-hunt.