oracle
[Native Image] Possible race condition in InterceptJSInvokeTypeFlow causes intermittent build failures
Describe the Issue
Intermittent build failure (~80% failure rate) in native-image --tool:svm-wasm with "Should not reach here" error. The failure is non-deterministic - running the same build multiple times produces different results.
Using the latest version of GraalVM can resolve many issues.
- [x] I tried with the latest version of GraalVM.
GraalVM Version
java 26 2026-03-17 Java(TM) SE Runtime Environment Oracle GraalVM 26-dev+13.1 (build 26+13-jvmci-b01) Java HotSpot(TM) 64-Bit Server VM Oracle GraalVM 26-dev+13.1 (build 26+13-jvmci-b01, mixed mode, sharing)
Operating System and Version
25.1.0 Darwin Kernel Version 25.1.0: Mon Oct 20 19:34:03 PDT 2025; root:xnu-12377.41.6~2/RELEASE_ARM64_T8112 arm64
Build Command
I unfortunately cannot provide a minimal repro -- it appears to trigger after a certain level of complexity, but I can share a repro based on a working project:
git clone https://github.com/willcohen/wasmts
cd wasmts
mvn clean compile
###### Run 10 times - typically ~2 succeed, ~8 fail
for i in {1..10}; do
native-image --tool:svm-wasm -cp "target/classes:target/lib/*" \
net.willcohen.wasmts.API target/wasmts 2>&1 | grep -E "(Finished|Failed)"
done
Expected Behavior
A successful build!
Actual Behavior
My output this last run was:
Finished generating 'wasmts' in 30.9s.
Finished generating 'wasmts' in 27.4s.
Failed generating 'wasmts' after 6.6s.
Finished generating 'wasmts' in 26.7s.
Failed generating 'wasmts' after 8.0s.
Failed generating 'wasmts' after 7.1s.
Failed generating 'wasmts' after 8.8s.
Finished generating 'wasmts' in 31.5s.
Failed generating 'wasmts' after 9.3s.
Failed generating 'wasmts' after 9.5s.
Steps to Reproduce
git clone https://github.com/willcohen/wasmts cd wasmts mvn clean compile
######## Run 10 times - typically ~2 succeed, ~8 fail
for i in {1..10}; do
native-image --tool:svm-wasm -cp "target/classes:target/lib/*"
net.willcohen.wasmts.API target/wasmts 2>&1 | grep -E "(Finished|Failed)"
done
Additional Context
No response
Build Log Output and Error Messages
To see the stack trace
mvn package
native-image --tool:svm-wasm \
-cp "target/classes:target/lib/jts-core-1.20.0.jar:target/lib/webimage-preview-25.0.0.jar" \
net.willcohen.wasmts.API target/wasmts \
-H:+ReportExceptionStackTraces 2>&1
===================================================================================================
GraalVM Native Image: Generating 'wasmts' (executable)...
===================================================================================================
[1/8] Initializing... (3.8s @ 0.12GB)
Builder configuration:
- Java version: 26+13, vendor version: Oracle GraalVM 26-dev+13.1
- Graal compiler: optimization level: 2, target machine: unknown, PGO: off
- C compiler: cc (apple, arm64, 17.0.0)
- Assertions: disabled, system assertions: disabled
- 1 user-specific feature(s):
- com.oracle.svm.thirdparty.gson.GsonFeature
Image configuration:
- Garbage collector: JS-runtime-provided GC (max heap size: unknown)
- Assertions: disabled (class-specific config may apply), system assertions: disabled
---------------------------------------------------------------------------------------------------
Build resources:
- 11.03GiB of memory (45.9% of system memory, using all available memory)
- 8 thread(s) (100.0% of 8 available processor(s), determined at start)
[2/8] Performing analysis... [] (1.5s @ 0.21GB)
1,048 types, 555 fields, and 2,370 methods found reachable
415 types, 7 fields, and 106 methods registered for reflection
9 resource accesses registered with 0B total size
Error: Should not reach here: This method should never be called.
Internal exception: com.oracle.graal.pointsto.util.AnalysisError: Should not reach here: This method should never be called.
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.util.AnalysisError.shouldNotReachHere(AnalysisError.java:181)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.typestate.PrimitiveTypeState.shouldNotReachHere(PrimitiveTypeState.java:171)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.typestate.PrimitiveTypeState.objectsIterator(PrimitiveTypeState.java:157)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.typestate.TypeState.lambda$objects$0(TypeState.java:76)
at org.graalvm.extraimage.builder/com.oracle.svm.hosted.webimage.pointsto.flow.InterceptJSInvokeTypeFlow.onObservedUpdate(InterceptJSInvokeTypeFlow.java:88)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.flow.TypeFlow.lambda$addObserver$0(TypeFlow.java:748)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.util.CompletionExecutor.executeCommand(CompletionExecutor.java:166)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.util.CompletionExecutor.lambda$executeService$0(CompletionExecutor.java:152)
at java.base/java.util.concurrent.ForkJoinTask$RunnableExecuteAction.compute(ForkJoinTask.java:1750)
at java.base/java.util.concurrent.ForkJoinTask$RunnableExecuteAction.compute(ForkJoinTask.java:1742)
at java.base/java.util.concurrent.ForkJoinTask$InterruptibleTask.exec(ForkJoinTask.java:1659)
at java.base/java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:511)
at java.base/java.util.concurrent.ForkJoinPool$WorkQueue.topLevelExec(ForkJoinPool.java:1450)
at java.base/java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:2019)
at java.base/java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:187)
Error: Should not reach here: This method should never be called.
Internal exception: com.oracle.graal.pointsto.util.AnalysisError: Should not reach here: This method should never be called.
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.util.AnalysisError.shouldNotReachHere(AnalysisError.java:181)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.typestate.PrimitiveTypeState.shouldNotReachHere(PrimitiveTypeState.java:171)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.typestate.PrimitiveTypeState.objectsIterator(PrimitiveTypeState.java:157)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.typestate.TypeState.lambda$objects$0(TypeState.java:76)
at org.graalvm.extraimage.builder/com.oracle.svm.hosted.webimage.pointsto.flow.InterceptJSInvokeTypeFlow.onObservedUpdate(InterceptJSInvokeTypeFlow.java:88)
at org.graalvm.extraimage.builder/com.oracle.svm.hosted.webimage.pointsto.flow.InterceptJSInvokeTypeFlow.lambda$onFlowEnabled$0(InterceptJSInvokeTypeFlow.java:78)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.AbstractAnalysisEngine$1.run(AbstractAnalysisEngine.java:337)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.util.CompletionExecutor.executeCommand(CompletionExecutor.java:166)
at org.graalvm.nativeimage.pointsto/com.oracle.graal.pointsto.util.CompletionExecutor.lambda$executeService$0(CompletionExecutor.java:152)
at java.base/java.util.concurrent.ForkJoinTask$RunnableExecuteAction.compute(ForkJoinTask.java:1750)
at java.base/java.util.concurrent.ForkJoinTask$RunnableExecuteAction.compute(ForkJoinTask.java:1742)
at java.base/java.util.concurrent.ForkJoinTask$InterruptibleTask.exec(ForkJoinTask.java:1659)
at java.base/java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:511)
at java.base/java.util.concurrent.ForkJoinPool$WorkQueue.topLevelExec(ForkJoinPool.java:1450)
at java.base/java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:2019)
at java.base/java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:187)
---------------------------------------------------------------------------------------------------
0.3s (4.3% of total time) in 16 GCs | Peak RSS: 904.15MB | CPU load: 3.96
===================================================================================================
Failed generating 'wasmts' after 5.4s.
Thanks for reporting, will look into it. This looks like a small oversight on our side that only gets triggered if the static analysis runs in a specific order.
I still don’t have a small repro but after spending the last few hours doing a large number of rebuilds using the branch from the PR, the issue is resolved with it.
Thanks for providing a fix. I have debugged this on my end and while your solution indeed solves this issue, these primitive states should never even appear in InterceptJSInvokeTypeFlow. I have a fix for the root cause that seems to work, your reproducer has been running successfully in a loop now for over half an hour.
I will let you know once my fix is mirrored to GitHub so that you can try it out