Tests graalvm-community-jdk21u-issues-28

[Karm]

Embeds a variation on the reproducer @zakkak outlined on the original issue.
This forces the native agent to register e.g. the getNextThreadIdOffset that later causes

/bin/ld: jdkreflections.o:(.data+0x7b8): undefined reference to `Java_java_lang_Thread_getNextThreadIdOffset'

Fails with current Mandrel for JDK 21, passes with e.g. Mandrel for JDK 23.

 mvn clean verify -Dquarkus.version=3.15.1 -Ptestsuite -Dtest=AppReproducersTest#jdkReflectionsTest

or in a container:

mvn clean verify -Ptestsuite-builder-image 
    -Dtest=AppReproducersTest#jdkReflectionsContainerTest -Dquarkus.version=3.15.1
    -Dquarkus.native.builder-image=quay.io/quarkus/ubi-quarkus-mandrel-builder-image:jdk-23
    -Dquarkus.native.container-runtime=podman
    -Drootless.container-runtime=true -Dpodman.with.sudo=false | tee log.log

...swap jdk-23 with jdk-21 to see the fail.

[Karm]

Hmm, it worries me that all tests in CI passes. JDK 21 should have failed. Lemme take a look...

[zakkak]

Hmm, it worries me that all tests in CI passes. JDK 21 should have failed. Lemme take a look...

You set minJDK = "21.0.3" but the CI uses

• Mandrel-23.1.1.0-Final which is Java version: 21.0.1+12-LTS
• Mandrel-23.1.2.0-Final which is Java version: 21.0.2+13-LTS
• Mandrel-24.0.1.0-Final which is Java version: 22.0.1+8 (which includes the fix)

I suggest bumping both the minJDK and the versions used in the pipelines

[zakkak]

LGTM, other than the minJDK version being 21.0.3 while the test is expected to fail with any JDK 21 build.

[jerboaa]

OK (with the suggestions from @zakkak)