Suggest Lookup.load() assigns its count and model fields while it is still reading the stream, so a load that throws (a stored copy cut short, a read failure, or FreeTextSuggester rejecting a model saved with other settings) leaves a mix of the previous and the new model: TSTLookup serves part of the new tree and some lookups throw NullPointerException, AnalyzingSuggester/FuzzySuggester can decode the new FST with the old payload flag, FreeTextSuggester scores the old FST with the new token total, and FST/WFSTCompletionLookup report the new count over the old FST

Finding LFV-979Area: suggest / Lookup persistence (load() failure atomicity: TSTLookup, AnalyzingSuggester, FuzzySuggester, FreeTextSuggester, FSTCompletionLookup, WFSTCompletionLookup)Priority: lowImpact: minorNovelty: new

Finding page for LFV-979

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

What it shows

You build a TSTLookup from {aa, ab, ba, bb, cア}, then call load() with a stored copy of another model that is cut short or whose stream throws IOException part-way, and catch the exception. You expect the lookup to hold either its previous data or no data, but it reports the new model's getCount() and serves part of the new tree, and some lookups throw NullPointerException; the same happens when you load into a newly constructed TSTLookup. The other Lookups do the same in their own way, because each load() assigns count and model fields one after another as it reads: AnalyzingSuggester/FuzzySuggester can decode the new FST with the old payload flag (payload glued to the key), FreeTextSuggester scores the old FST with the new token total and changes getCount() even when it rejects a complete model saved with other grams/separator settings, and FST/WFSTCompletionLookup report the new count over the old FST.

The test shows that TSTLookup.load() sets its count before it reads the tree, so a load of a stored copy cut short by one byte throws but leaves a new TSTLookup reporting the stored copy's count.

Expected vs actual

Expected: after the failed load(), getCount() is 0 (the new lookup's previous, empty state). Actual: getCount() after a failed load() expected:<0> but was:<2>.

Files

Run it

# 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/analysis/common/build/libs/lucene-analysis-common-$V.jar
CP=$CP:$LUCENE/lucene/suggest/build/libs/lucene-suggest-$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 on, seed DEADBEEF
java -ea -esa --add-modules jdk.incubator.vector --enable-native-access=ALL-UNNAMED -Dtests.seed=DEADBEEF \
     -cp "classes:$CP" org.junit.platform.console.ConsoleLauncher \
     execute --select-class org.apache.lucene.search.suggest.TestTSTLookupFailedLoad

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.