Fix circular dependency deadlock in TestSecrets initialization by danmuzi · Pull Request #15864 · apache/lucene

danmuzi · 2026-03-23T16:09:26Z

Related Issues

Deadlock between IndexWriter, IndexWriterConfig, ConcurrentMergeScheduler #15119
IndexWriter and ConcurrentMergeScheduler and SegmentReader can cause static initialization deadlock #12419

Overview

A deadlock occurred during static initialization due to the JVM Class Initialization Lock.
This issue occurs through mutual references between org.apache.lucene.index.IndexWriter(CMS, SegmentReader, FilterIndexInput also) and org.apache.lucene.internal.tests.TestSecrets.

Cause

Circular Dependency

IndexWriter invokes TestSecrets during its initialization, while TestSecrets conversely attempts to force-load IndexWriter during its own initialization.

JVM Class Loading Lock

When the JVM loads a class for the first time, it acquires a global lock on that class object.
A deadlock occurs when two threads each hold a lock on different classes and wait for the other thread to release the lock for the class they are trying to load.

Code structure

IndexWriter.java, ConcurrentMergeScheduler.java, SegmentReader.java, FilterIndexInput.java

static {
    // Trigger initialization by calling TestSecrets
    TestSecrets.set...Access(new ...Access() { ... });
}

TestSecrets.java

static {
    ...
    ensureInitialized.accept(ConcurrentMergeScheduler.class);
    ensureInitialized.accept(SegmentReader.class);
    ensureInitialized.accept(IndexWriter.class); // Deadlock trigger point
    ensureInitialized.accept(FilterIndexInput.class);
}

Deadlock Scenario

This deadlock occurs in a multi-threaded environment when two classes are accessed for the first time almost simultaneously.

Step	Thread A	Thread B
1	First access to `IndexWriter` class.	First access to `SegmentReader` class.
2	Acquires lock on `IndexWriter.class` and starts initialization.	Acquires lock on `SegmentReader.class` and starts initialization.
3	Calls `TestSecrets` → Acquires lock on `TestSecrets.class`.	Attempts to call `TestSecrets`.
4	(Initialization in progress)	Blocked: Waiting for `TestSecrets` lock (held by Thread A).
5	`TestSecrets` static block calls `Class.forName("SegmentReader")`.	(Waiting...)
6	Blocked: Waiting for `SegmentReader` lock (held by Thread B).	(Waiting...)

Result: A permanent deadlock occurs where Thread A waits for the SegmentReader lock while Thread B waits for the TestSecrets lock.

Steps to Reproduce

Without applying my TestSecrets fix:

Run the main() of the DeadlockTest class in TestTestSecrets.
Run testDeadlock() in TestTestSecrets.

Solution - Lazy Initialization

All Class.forName calls that forcibly load target classes within the static initialization block of TestSecrets have been removed.
Instead, the logic has been updated to perform initialization individually within each Getter method at the actual time of invocation.

TestSecrets.java

public final class TestSecrets {
    private static IndexWriterAccess indexWriterAccess;
    private static ConcurrentMergeSchedulerAccess cmsAccess;
    private static SegmentReaderAccess segmentReaderAccess;

    // All ensureInitialized calls within the static initialization block have been removed.

    public static IndexWriterAccess getIndexWriterAccess() {
        ensureCallerForGetter();
        if (indexWriterAccess == null) {
            ensureInitialized.accept(IndexWriter.class);
        }
        return Objects.requireNonNull(indexWriterAccess);
    }
    // Same pattern applies to other getters (CMS, SegmentReader, etc.)
}

Diagram

graph TD
    subgraph "Thread A (Initializes IndexWriter)"
        A1[IndexWriter class load] --> A2["IndexWriter.<clinit> (Lock IndexWriter.class)"]
        A2 --> A3["Call TestSecrets.setIndexWriterAccess()"]
        A3 --> A4["Trigger TestSecrets.<clinit> (Lock TestSecrets.class)"]
        A4 --> A5["TestSecrets.<clinit> calls Class.forName('SegmentReader')"]
        A5 -- "Wait for SegmentReader.class lock" --> B2
    end

    subgraph "Thread B (Initializes SegmentReader)"
        B1[SegmentReader class load] --> B2["SegmentReader.<clinit> (Lock SegmentReader.class)"]
        B2 --> B3["Call TestSecrets.setSegmentReaderAccess()"]
        B3 -- "Wait for TestSecrets.class lock" --> A4
    end

    A5 -. Deadlock .-> B2
    B3 -. Deadlock .-> A4

Signed-off-by: Namgyu Kim <namgyu@apache.org>

uschindler

See comments inline.

uschindler · 2026-03-23T16:27:23Z

    expectThrows(AssertionError.class, () -> TestSecrets.setSegmentReaderAccess(null));
  }
+
+  public void testDeadlock() throws Exception {


This test is fine, but I'd like to avoaid new tests spawning childs processes as much as possible. It is good to reproduce, but maybe @dweiss has an opinion.

This is rather complicated and even with a barrier there's no guarantee this will always deadlock... I'm not sure we need this, to be honest.

We have a test for the codec/postingsformat deadlock:

lucene/lucene/core/src/test/org/apache/lucene/codecs/TestCodecLoadingDeadlock.java

Line 50 in 48379f0

public class TestCodecLoadingDeadlock extends Assert {

This one is using some randomness to trigger different combinations of initialization. Maybe we can include the test here. It is realted, as most class loading deadlocks occur in Lucene around loading IndexWriter together with codecs/postings/docvalues format and.... TestSecrents

I think we can add this case there in the thread:

update switch statement and add a new case that loads TestSecrets and calls one of the getter methods

raise the threadcount as described at beginning of tests (2 ztimes the module/number of cases). I'd improve the test to use a common variable for that, it is a bit tricky

I was able to beef-up that test. Works much better and tests the whole deadloding around IndexWriter. I will post a patch for the other test, or should I commit directly?

TestCodecLoadingDeadlock > testDeadlock FAILED java.lang.AssertionError: Process died abnormally? expected:<0> but was:<1> at __randomizedtesting.SeedInfo.seed([A5DDBF322DCE93C1:A8B65E262B943E17]:0) at org.junit.Assert.fail(Assert.java:89) at org.junit.Assert.failNotEquals(Assert.java:835) at org.junit.Assert.assertEquals(Assert.java:647) at org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:88) at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104) at java.base/java.lang.reflect.Method.invoke(Method.java:565) at com.carrotsearch.randomizedtesting.RandomizedRunner.invoke(RandomizedRunner.java:1763) at com.carrotsearch.randomizedtesting.RandomizedRunner$8.evaluate(RandomizedRunner.java:946) at com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(StatementAdapter.java:36) at com.carrotsearch.randomizedtesting.ThreadLeakControl$StatementRunner.run(ThreadLeakControl.java:390) at com.carrotsearch.randomizedtesting.ThreadLeakControl.forkTimeoutingTask(ThreadLeakControl.java:843) at com.carrotsearch.randomizedtesting.ThreadLeakControl$3.evaluate(ThreadLeakControl.java:490) at com.carrotsearch.randomizedtesting.RandomizedRunner.runSingleTest(RandomizedRunner.java:955) at com.carrotsearch.randomizedtesting.RandomizedRunner$5.evaluate(RandomizedRunner.java:840) at com.carrotsearch.randomizedtesting.RandomizedRunner$6.evaluate(RandomizedRunner.java:891) at com.carrotsearch.randomizedtesting.RandomizedRunner$7.evaluate(RandomizedRunner.java:902) at com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(StatementAdapter.java:36) at com.carrotsearch.randomizedtesting.ThreadLeakControl$StatementRunner.run(ThreadLeakControl.java:390) at com.carrotsearch.randomizedtesting.ThreadLeakControl.forkTimeoutingTask(ThreadLeakControl.java:843) at com.carrotsearch.randomizedtesting.ThreadLeakControl$2.evaluate(ThreadLeakControl.java:426) at com.carrotsearch.randomizedtesting.RandomizedRunner.runSuite(RandomizedRunner.java:716) at com.carrotsearch.randomizedtesting.RandomizedRunner.access$200(RandomizedRunner.java:138) at com.carrotsearch.randomizedtesting.RandomizedRunner$2.run(RandomizedRunner.java:637) org.apache.lucene.codecs.TestCodecLoadingDeadlock > test suite's output saved to C:\Users\Uwe Schindler\Projects\lucene\lucene\lucene\core\build\test-results\test_1\outputs\OUTPUT-org.apache.lucene.codecs.TestCodecLoadingDeadlock.txt, copied below: 2> M├ñrz 24, 2026 9:10:47 AM org.apache.lucene.internal.vectorization.VectorizationProvider lookup 2> WARNUNG: Vector bitsize enforcement; using default vectorization provider outside of testMode 1> codec: FastDecompressionCompressingStoredFieldsData, pf: MockRandom, dvf: Lucene90 1> Subprocess emitted the following output: 1> M´┐¢rz 24, 2026 9:10:48 AM org.apache.lucene.internal.vectorization.VectorizationProvider lookup 1> WARNUNG: Java vector incubator module is not readable. For optimal vector performance, pass '--add-modules jdk.incubator.vector' to enable Vector API. 1> Pool didn't return after 30 seconds, classloader deadlock? Dumping stack traces. 1> # Thread: Thread[#26,deadlockchecker-1-thread-1,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#27,deadlockchecker-1-thread-2,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#28,deadlockchecker-1-thread-3,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#29,deadlockchecker-1-thread-4,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#30,deadlockchecker-1-thread-5,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#31,deadlockchecker-1-thread-6,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#32,deadlockchecker-1-thread-7,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#33,deadlockchecker-1-thread-8,5,main], state: RUNNABLE, stack: 1> org.apache.lucene.index.IndexWriter.<clinit>(IndexWriter.java:6575) java.base/jdk.internal.misc.Unsafe.ensureClassInitialized0(Native Method) java.base/jdk.internal.misc.Unsafe.ensureClassInitialized(Unsafe.java:1169) java.base/java.lang.invoke.MethodHandles$Lookup.ensureInitialized(MethodHandles.java:2748) org.apache.lucene.codecs.TestCodecLoadingDeadlock.lambda$main$1(TestCodecLoadingDeadlock.java:157) java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1090) java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:614) java.base/java.lang.Thread.run(Thread.java:1474) 1> 1> # Thread: Thread[#34,deadlockchecker-1-thread-9,5,main], state: RUNNABLE, stack: 1> java.base/jdk.internal.misc.Unsafe.ensureClassInitialized0(Native Method) java.base/jdk.internal.misc.Unsafe.ensureClassInitialized(Unsafe.java:1169) java.base/java.lang.invoke.MethodHandles$Lookup.ensureInitialized(MethodHandles.java:2748) org.apache.lucene.codecs.TestCodecLoadingDeadlock.lambda$main$1(TestCodecLoadingDeadlock.java:160) java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1090) java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:614) java.base/java.lang.Thread.run(Thread.java:1474) 1> 1> # Thread: Thread[#35,deadlockchecker-1-thread-10,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#36,deadlockchecker-1-thread-11,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#37,deadlockchecker-1-thread-12,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#38,deadlockchecker-1-thread-13,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#39,deadlockchecker-1-thread-14,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#40,deadlockchecker-1-thread-15,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#41,deadlockchecker-1-thread-16,5,], state: TERMINATED, stack: 1> 1> 1> # Thread: Thread[#42,deadlockchecker-1-thread-17,5,main], state: RUNNABLE, stack: 1> java.base/jdk.internal.misc.Unsafe.ensureClassInitialized0(Native Method) java.base/jdk.internal.misc.Unsafe.ensureClassInitialized(Unsafe.java:1169) java.base/java.lang.invoke.MethodHandles$Lookup.ensureInitialized(MethodHandles.java:2748) org.apache.lucene.codecs.TestCodecLoadingDeadlock.lambda$main$1(TestCodecLoadingDeadlock.java:157) java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1090) java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:614) java.base/java.lang.Thread.run(Thread.java:1474) 1> 1> # Thread: Thread[#43,deadlockchecker-1-thread-18,5,main], state: RUNNABLE, stack: 1> java.base/java.lang.Class.forName0(Native Method) java.base/java.lang.Class.forName(Class.java:467) java.base/java.lang.Class.forName(Class.java:458) org.apache.lucene.internal.tests.TestSecrets.lambda$static$0(TestSecrets.java:43) org.apache.lucene.internal.tests.TestSecrets.<clinit>(TestSecrets.java:51) org.apache.lucene.index.SegmentReader.<clinit>(SegmentReader.java:402) java.base/jdk.internal.misc.Unsafe.ensureClassInitialized0(Native Method) java.base/jdk.internal.misc.Unsafe.ensureClassInitialized(Unsafe.java:1169) java.base/java.lang.invoke.MethodHandles$Lookup.ensureInitialized(MethodHandles.java:2748) org.apache.lucene.codecs.TestCodecLoadingDeadlock.lambda$main$1(TestCodecLoadingDeadlock.java:160) java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1090) java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:614) java.base/java.lang.Thread.run(Thread.java:1474) 1> > java.lang.AssertionError: Process died abnormally? expected:<0> but was:<1> > at __randomizedtesting.SeedInfo.seed([A5DDBF322DCE93C1:A8B65E262B943E17]:0) > at org.junit.Assert.fail(Assert.java:89) > at org.junit.Assert.failNotEquals(Assert.java:835) > at org.junit.Assert.assertEquals(Assert.java:647) > at org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:88) > at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104) > at java.base/java.lang.reflect.Method.invoke(Method.java:565) > at com.carrotsearch.randomizedtesting.RandomizedRunner.invoke(RandomizedRunner.java:1763) > at com.carrotsearch.randomizedtesting.RandomizedRunner$8.evaluate(RandomizedRunner.java:946) > at com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(StatementAdapter.java:36) > at com.carrotsearch.randomizedtesting.ThreadLeakControl$StatementRunner.run(ThreadLeakControl.java:390) > at com.carrotsearch.randomizedtesting.ThreadLeakControl.forkTimeoutingTask(ThreadLeakControl.java:843) > at com.carrotsearch.randomizedtesting.ThreadLeakControl$3.evaluate(ThreadLeakControl.java:490) > at com.carrotsearch.randomizedtesting.RandomizedRunner.runSingleTest(RandomizedRunner.java:955) > at com.carrotsearch.randomizedtesting.RandomizedRunner$5.evaluate(RandomizedRunner.java:840) > at com.carrotsearch.randomizedtesting.RandomizedRunner$6.evaluate(RandomizedRunner.java:891) > at com.carrotsearch.randomizedtesting.RandomizedRunner$7.evaluate(RandomizedRunner.java:902) > at com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(StatementAdapter.java:36) > at com.carrotsearch.randomizedtesting.ThreadLeakControl$StatementRunner.run(ThreadLeakControl.java:390) > at com.carrotsearch.randomizedtesting.ThreadLeakControl.forkTimeoutingTask(ThreadLeakControl.java:843) > at com.carrotsearch.randomizedtesting.ThreadLeakControl$2.evaluate(ThreadLeakControl.java:426) > at com.carrotsearch.randomizedtesting.RandomizedRunner.runSuite(RandomizedRunner.java:716) > at com.carrotsearch.randomizedtesting.RandomizedRunner.access$200(RandomizedRunner.java:138) > at com.carrotsearch.randomizedtesting.RandomizedRunner$2.run(RandomizedRunner.java:637) :lucene:core:test_1 (FAILURE): 1 test, 1 failure 1 test completed, 1 failed > Task :lucene:core:test_1 FAILED

Patch:

diff --git a/lucene/core/src/test/org/apache/lucene/codecs/TestCodecLoadingDeadlock.java b/lucene/core/src/test/org/apache/lucene/codecs/TestCodecLoadingDeadlock.java index 0e0c3941bec..3705c8e6d58 100644 --- a/lucene/core/src/test/org/apache/lucene/codecs/TestCodecLoadingDeadlock.java +++ b/lucene/core/src/test/org/apache/lucene/codecs/TestCodecLoadingDeadlock.java @@ -20,6 +20,8 @@ import com.carrotsearch.randomizedtesting.LifecycleScope; import com.carrotsearch.randomizedtesting.RandomizedContext; import com.carrotsearch.randomizedtesting.RandomizedRunner; import com.carrotsearch.randomizedtesting.RandomizedTest; + +import java.lang.invoke.MethodHandles; import java.nio.file.Files; import java.nio.file.Path; import java.nio.file.Paths; @@ -36,6 +38,10 @@ import java.util.concurrent.Executors; import java.util.concurrent.TimeUnit; import java.util.stream.Collectors; import java.util.stream.IntStream; + +import org.apache.lucene.index.IndexWriter; +import org.apache.lucene.index.SegmentReader; +import org.apache.lucene.internal.tests.TestSecrets; import org.apache.lucene.tests.util.LuceneTestCase; import org.apache.lucene.util.NamedThreadFactory; import org.apache.lucene.util.SuppressForbidden; @@ -99,7 +105,8 @@ public class TestCodecLoadingDeadlock extends Assert { final String pfName = args[1]; final String dvfName = args[2]; - final int numThreads = 14; // two times the modulo in switch statement below + final int numTasks = 9; // add number of catch cases in switch below + final int numThreads = numTasks * 2; // two times the modulo in switch statement below final CopyOnWriteArrayList<Thread> allThreads = new CopyOnWriteArrayList<>(); final ExecutorService pool = Executors.newFixedThreadPool( @@ -114,6 +121,7 @@ public class TestCodecLoadingDeadlock extends Assert { }); final CyclicBarrier barrier = new CyclicBarrier(numThreads); + final var lookup = MethodHandles.lookup(); IntStream.range(0, numThreads) .forEach( taskNo -> @@ -123,7 +131,7 @@ public class TestCodecLoadingDeadlock extends Assert { // Await a common barrier point for all threads and then // run racy code. This is intentional. barrier.await(); - switch (taskNo % 7) { + switch (taskNo % numTasks) { case 0: Codec.getDefault(); break; @@ -145,6 +153,12 @@ public class TestCodecLoadingDeadlock extends Assert { case 6: DocValuesFormat.availableDocValuesFormats(); break; + case 7: + lookup.ensureInitialized(IndexWriter.class); + break; + case 8: + lookup.ensureInitialized(SegmentReader.class); + break; default: throw new AssertionError(); }

It fails with current main, have not yet applied to this branch. I can do this an disable your test.

The "beefed" up test now checks better for deadlocks by really treating not only codec loading but everything around IndexWriter / IndexReader.

I removed the test and included the test in the more general deadlock testcase which covers also codec deadloads.

The test failed (see above) and passes with this PR.

uschindler · 2026-03-23T16:32:30Z

Hi,
this should be backported to 10.x branch, so add changes to 10.x section of changes list. The bot will change the milestone automatically.

uschindler · 2026-03-23T16:33:08Z

Please run gradlew tidy.

Signed-off-by: Namgyu Kim <namgyu@apache.org>

…lerForSetter Signed-off-by: Namgyu Kim <namgyu@apache.org>

uschindler

I am not sure if the test is really needed, otherwise looks perfect.

Back porting should be easy.

danmuzi · 2026-03-23T17:43:28Z

@uschindler
Uwe,
I also find that test a bit ambiguous. So I expect Dawid(@dweiss) will review it. (#15864 (comment))
Since I have committer rights anyway, I'll update the changes after the review and cherry-pick them to the 10x branch.

uschindler · 2026-03-23T19:01:21Z

We have a similar test for Service loader deadlock with codecs. Maybe we can adapt that one to test both?

dweiss · 2026-03-23T20:12:57Z

    expectThrows(AssertionError.class, () -> TestSecrets.setSegmentReaderAccess(null));
  }
+
+  public void testDeadlock() throws Exception {


This is rather complicated and even with a barrier there's no guarantee this will always deadlock... I'm not sure we need this, to be honest.

…imporve that test

uschindler · 2026-03-24T10:46:14Z

Hi @danmuzi ,
I replaced the test and extended the already existing test that checks for deadlocks on codec loading. The test is better written and was further improved now. I hit the issue here (it was just missing to load IndexWriter and SegmentReader), but now it was beefed up and passes.

I also fixed one more Exception in the setters.

In another commit I will now move and rename the test to be more general.TestCodecLoadingDeadlock => TestClassloaderDeadlocks in index package

uschindler · 2026-03-24T10:50:21Z

Test moved. Have fun!

# Conflicts: # lucene/CHANGES.txt

uschindler · 2026-03-24T11:01:54Z

This passes:

$ gradlew :lucene:core:beast --tests TestClassloadingDeadlock -Ptests.dups=100

uschindler · 2026-03-24T13:32:57Z

I think we should replace all UnsupportedOperationException by IllegalCallerException which is exactly made for that use case.

I would do this for here and the other places in Lucene in a separate PR.

danmuzi · 2026-03-24T16:15:30Z

Hi @uschindler ,

I've tested your latest changes with the updated TestClassloadingDeadlock.
I confirmed that a deadlock occurs when running it against the commit prior to my fix.
Therefore, your patch looks good to me.
I'll submit this PR and cherry-pick it to the 10x branch.

Thanks for your detailed review!

Signed-off-by: Namgyu Kim <namgyu@apache.org> Co-authored-by: Uwe Schindler <uschindler@apache.org>

uschindler · 2026-03-24T17:11:25Z

Perfect, thanks. I will open a small PR for the global replacement of the Exception.

fix deadlock in TestSecrets

4a4c974

Signed-off-by: Namgyu Kim <namgyu@apache.org>

danmuzi requested review from dweiss and uschindler March 23, 2026 16:09

github-actions Bot added the module:core/other label Mar 23, 2026

add Changes and fix main method

2ee92b5

Signed-off-by: Namgyu Kim <namgyu@apache.org>

github-actions Bot added this to the 11.0.0 milestone Mar 23, 2026

danmuzi added module:core/index module:core/search blocker A severe issue that should be resolved before the released specified in its Milestone. labels Mar 23, 2026

uschindler requested changes Mar 23, 2026

View reviewed changes

fix code format

4f79e40

Signed-off-by: Namgyu Kim <namgyu@apache.org>

github-actions Bot removed module:core/index module:core/search labels Mar 23, 2026

move version to 10.5.0 in CHANGES

14de99a

Signed-off-by: Namgyu Kim <namgyu@apache.org>

github-actions Bot modified the milestones: 11.0.0, 10.5.0 Mar 23, 2026

change ensureInitialized from lambda to static method

3b7902b

Signed-off-by: Namgyu Kim <namgyu@apache.org>

uschindler reviewed Mar 23, 2026

View reviewed changes

Comment thread lucene/core/src/java/org/apache/lucene/internal/tests/TestSecrets.java Outdated

uschindler reviewed Mar 23, 2026

View reviewed changes

Comment thread lucene/core/src/java/org/apache/lucene/internal/tests/TestSecrets.java Outdated

uschindler reviewed Mar 23, 2026

View reviewed changes

Comment thread lucene/core/src/java/org/apache/lucene/internal/tests/TestSecrets.java Outdated

make MethodHandles.Lookup to static and change Exception on ensureCal…

efcc9f5

…lerForSetter Signed-off-by: Namgyu Kim <namgyu@apache.org>

uschindler approved these changes Mar 23, 2026

View reviewed changes

dweiss approved these changes Mar 23, 2026

View reviewed changes

Remove test and integrate in the general codec loading dadlock test; …

1672a6d

…imporve that test

github-actions Bot added module:core/codecs and removed module:core/other labels Mar 24, 2026

update exception type to be in line with other code

736c050

Move and rename test

3f5b43d

github-actions Bot added module:core/index and removed module:core/codecs labels Mar 24, 2026

Merge branch 'main' into fix-testsecrets-deadlock

a1d3fc6

# Conflicts: # lucene/CHANGES.txt

github-actions Bot modified the milestones: 10.5.0, 11.0.0 Mar 24, 2026

update CHANGES.txt

362a05b

danmuzi modified the milestones: 11.0.0, 10.5.0 Mar 24, 2026

danmuzi merged commit 83e3f9a into apache:main Mar 24, 2026
17 of 18 checks passed

danmuzi added a commit that referenced this pull request Mar 24, 2026

Fix circular dependency deadlock in TestSecrets initialization (#15864)

bae2a5b

Signed-off-by: Namgyu Kim <namgyu@apache.org> Co-authored-by: Uwe Schindler <uschindler@apache.org>

This was referenced Mar 24, 2026

Deadlock between IndexWriter, IndexWriterConfig, ConcurrentMergeScheduler #15119

Closed

IndexWriter and ConcurrentMergeScheduler and SegmentReader can cause static initialization deadlock #12419

Closed

uschindler mentioned this pull request Mar 25, 2026

Use IllegalCallerException to signal wrong caller in internal classes (VectorizationProvider, TestSecrets) #15877

Merged

Conversation

danmuzi commented Mar 23, 2026

Related Issues

Overview

Cause

Circular Dependency

JVM Class Loading Lock

Code structure

Deadlock Scenario

Steps to Reproduce

Solution - Lazy Initialization

Diagram

Uh oh!

uschindler left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

uschindler commented Mar 23, 2026

Uh oh!

uschindler commented Mar 23, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

uschindler left a comment

Choose a reason for hiding this comment

Uh oh!

danmuzi commented Mar 23, 2026

Uh oh!

uschindler commented Mar 23, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

uschindler commented Mar 24, 2026

Uh oh!

uschindler commented Mar 24, 2026

Uh oh!

uschindler commented Mar 24, 2026

Uh oh!

uschindler commented Mar 24, 2026

Uh oh!

danmuzi commented Mar 24, 2026

Uh oh!

Uh oh!

uschindler commented Mar 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants