vicclavijo opened a new issue, #2423:
URL: https://github.com/apache/incubator-kie-issues/issues/2423

   Description
   
   Building a Quarkus project that uses drools-quarkus-rules / 
drools-quarkus-decisions (kogito-bom 10.2.0, the latest published version) 
fails at build time once Quarkus is upgraded past 3.36.3. A second, independent 
issue causes codegen to fail with quarkus-rest (RESTEasy Reactive) on every 
Quarkus version we tested, including the last one where the first issue doesn't 
occur.
   
   Environment
   
   - org.kie.kogito:kogito-bom: 10.2.0
   - org.drools:drools-quarkus-rules and org.drools:drools-quarkus-decisions: 
10.2.0
   - Quarkus: tested 3.33.1 through 3.40.1.1 (21 versions)
   - Java: 21 and 25 (both affected equally)
   - Build: mvn clean package -DskipTests
   
   Issue 1 — build-time NoSuchMethodError from Quarkus 3.37.0 onward
   
   [error]: Build step 
org.kie.kogito.quarkus.common.deployment.KogitoAssetsProcessor#generateModel
   threw an exception: java.lang.NoSuchMethodError:
   'void 
io.quarkus.deployment.builditem.nativeimage.ReflectiveClassBuildItem.<init>(boolean,
 boolean, java.lang.String[])'
           at 
org.kie.kogito.quarkus.common.deployment.KogitoAssetsProcessor.registerDataEventsForReflection(...)
           at 
org.kie.kogito.quarkus.common.deployment.KogitoAssetsProcessor.generateModel(...)
   
   Root cause: kogito-quarkus-common:10.2.0's KogitoAssetsProcessor calls the 
3-arg constructor ReflectiveClassBuildItem(boolean, boolean, String...). That 
constructor exists in quarkus-core-deployment through 3.36.3 but was removed 
starting with 3.37.0 (replaced by overloads with 8–10 boolean parameters). This 
is an internal (non-public-API) build-time signature change on the Quarkus 
side, which is presumably why it isn't called out as a breaking change — but it 
means kogito-quarkus-common as published is bound to a Quarkus internal API 
window that closed at 3.37.0.
   
   Verification method: downloaded quarkus-core-deployment-<version>.jar for 21 
versions between 3.33.1 and 3.40.1.1 from Maven Central and checked for the 
constructor via javap. Confirmed present through 3.36.3, confirmed absent from 
3.37.0 onward. Reproduced with a real mvn clean package build on both 3.33.3.1 
(success) and 3.40.1.1 (fails with the trace above) using a minimal 
single-module project (one RuleUnitData + DataStore<String> + one trivial .drl).
   
   Impact: this blocks before model generation even starts — it is not a 
runtime or native-image issue, it affects plain JVM-mode mvn package.
   
   Issue 2 — codegen fails with quarkus-rest (RESTEasy Reactive) independent of 
the Quarkus version
   
   With Quarkus 3.36.3 (the last version where Issue 1 doesn't occur) and 
quarkus-rest instead of quarkus-resteasy:
   
   java.lang.IllegalStateException: Multiple GeneratedClassBuildItem were 
produced for the same
   classes: org.kie.kogito.app.GlobalObjectMapper: 2
   
   The same build succeeds with quarkus-resteasy (classic) unchanged otherwise. 
We previously hit what looks like the same symptom (duplicate/silently-missing 
generated classes) under Quarkus 3.33.x in an internal project, so this does 
not appear to be new in 3.36.x — it looks like Kogito codegen and quarkus-rest 
have never reliably coexisted across the versions we've tested. We have not 
found any Quarkus version where Kogito 10.2.0 builds successfully with 
quarkus-rest.
   
   Summary
   
   ### Summary
   
   | Quarkus version | `quarkus-resteasy` (classic)                | 
`quarkus-rest` (reactive) |
   
|-------------------|-----------------------------------------|------------------------------|
   | 3.33.x - 3.36.3     | ✅ builds                                           | 
❌ `IllegalStateException` (duplicate generated classes) |
   | >= 3.37.0            | ❌ `NoSuchMethodError` (Issue 1)  | ❌ same 
`NoSuchMethodError` |
   
   Questions for maintainers
   
   1. Is there a newer kogito-quarkus-common build (or a branch/snapshot 
post-Apache-KIE-migration) already compiled against a Quarkus version ≥ 3.37.0, 
even unreleased?
   2. Is the quarkus-rest codegen conflict (Issue 2) a known limitation, or is 
RESTEasy classic expected to keep working indefinitely for Kogito-based 
projects regardless of what the rest of a given archetype/BOM standardizes on?
   
   Thanks a lot for maintaining this project, and for taking the time to look 
into this — really appreciate it!
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to