On main, CheckIndex's HNSW graph test reports valid segments written by Lucene 9.12.0-9.12.1 and 10.0.0-10.1.0 as broken ('has repeated neighbors'), because the check has no version gate

Finding LFV-543Area: index / CheckIndex HNSW graph checkPriority: lowImpact: wrong errorNovelty: known upstream

AI-assisted finding, confirmed by automated checks, not yet triaged by a human Lucene committer. This site files nothing upstream.

What it shows

You open, on Lucene main, an index that Lucene 9.12.0, 9.12.1, 10.0.0 or 10.1.0 wrote with the default codec, and run CheckIndex at level 2 or 3. You expect it to pass: main opens, reads, searches and merges the index correctly, and upstream calls repeated HNSW neighbours benign. Instead, when an old unmerged segment's graph lists the same neighbour twice (those releases' connectComponents step could add an existing link again, and their writers did not de-duplicate), CheckIndex reports 'has repeated neighbors': with failFast it throws CheckIndexException ('Segment ... check failed'), otherwise it prints 'test: hnsw graphs.........ERROR', because the rule applies to every segment whatever version wrote it.

Expected vs actual

Expected: You expect it to pass: main opens, reads, searches and merges the index correctly, and upstream calls repeated HNSW neighbours benign.
Actual: When an old unmerged segment's graph lists the same neighbour twice (those releases' connectComponents step could add an existing link again, and their writers did not de-duplicate), CheckIndex reports 'has repeated neighbors': with failFast it throws CheckIndexException ('Segment ... check failed'), otherwise it prints 'test: hnsw graphs.........ERROR', because the rule applies to every segment whatever version wrote it.

Files

src/org/apache/lucene/index/TestCheckIndexOldGraphRepeatedNeighbors.java (JUnit test)
fixture-src/DupGen3.java (writes the fixture the test reads, as Run it says; not compiled with the test)

Run it

Tested on Lucene main e1faeba08bde (2026-11-02): still reproduces.
Tested on Lucene main ea57360912ed (2026-10-24): reproduced.
Fixtures: the test reads repeated-neighbors.9.12.0.zip, repeated-neighbors.10.0.0.zip and repeated-neighbors.10.1.0.zip, indexes of 1,000 one-dimensional vectors written by those releases whose HNSW graphs list a neighbour twice, from its package on the class path. This site ships no binary index files, so the recipe first makes them with fixture-src/DupGen3.java and each release's lucene-core jar (Maven Central): it tries seeds 0 to 99 and keeps the first index with a repeated neighbour, and the zips go to src/, which the run puts on the class path; without them every test fails for another reason.
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=$CP:$LUCENE/lucene/backward-codecs/build/libs/lucene-backward-codecs-$V.jar
CP=$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)

# fixtures repeated-neighbors.<version>.zip, written by the 9.12.0, 10.0.0 and 10.1.0 releases: lucene-core-<version>.jar
# of each (Maven Central)
for v in 9.12.0 10.0.0 10.1.0; do
  javac -d fixture-classes-$v -cp lucene-core-$v.jar fixture-src/DupGen3.java
  java -cp fixture-classes-$v:lucene-core-$v.jar DupGen3 fixture-$v 0 100 1 12 32 661 0.4185 18
  (cd fixture-$v/seed-* && zip -q -X ../../src/org/apache/lucene/index/repeated-neighbors.$v.zip *)
done

javac -d classes -cp "$CP" $(find src -name '*.java')

# run 1: assertions on, seed DEADBEEF
java -ea -esa --add-modules jdk.incubator.vector --enable-native-access=ALL-UNNAMED -Dtests.seed=DEADBEEF \
     -cp "classes:src:$CP" org.junit.platform.console.ConsoleLauncher \
     execute --select-class org.apache.lucene.index.TestCheckIndexOldGraphRepeatedNeighbors

Gradle 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).

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.