AnalyzingSuggester and FuzzySuggester take the new input's payload flag before build() can fail, so after a failed rebuild (or during a rebuild) lookups decode the previous FST in the wrong payload format: the payload glued to the suggestion text, or the suggestion returned as the payload

Finding LFV-784Area: suggest / analyzing suggester (AnalyzingSuggester.build failure atomicity: hasPayloads is published before the new FST; inherited by FuzzySuggester)Priority: lowImpact: silent wrong resultsNovelty: new

Finding page for LFV-784

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 an AnalyzingSuggester (or FuzzySuggester) from input with payloads, for example 'foo' with weight 10 and payload 'p1', then rebuild the same instance from a dictionary without payloads whose input throws an IOException partway through, and you expect lookups to keep returning the previous build ('foo', payload 'p1') or nothing. Instead lookup('fo') returns the text 'foo<U+001F>p1' with payload null, because build() stores the new input's payload flag at its start (AnalyzingSuggester.java:394) but replaces the FST only at its end (:589-590), so the old FST is read with the new flag; in the opposite direction (old build without payloads, failed rebuild with payloads) lookups return empty text with 'foo' as the payload, or trip the suggester's own assertion with -ea. A lookup made while such a rebuild is still reading its input sees the same mixture, and store() writes it to disk, so load() brings it back.

The test shows that AnalyzingSuggester.build() stores the new input's payload flag before it can fail, so after a rebuild without payloads that throws an IOException, the previous FST (built with payloads) is decoded in the wrong format.

Expected vs actual

Expected: lookup("fo") returns foo with payload p1 (the previous build). Actual (-ea and -da): key foo<U+001F>p1 with payload null.

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.analyzing.TestAnalyzingSuggesterFailedRebuild

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 2026-11-03. AI-assisted; not yet triaged by a human Lucene committer. Feedback: open an issue in lucene-ai-bug-hunt.