Continuous testing is not picking up custom JUnit extension from same module

[famod]

Describe the bug

A Junit extension, e.g. a org.junit.jupiter.api.extension.BeforeAllCallback, that is registered via src/test/resources/META-INF/services/org.junit.jupiter.api.extension.Extension is not picked up when mvn quarkus:test (CT) is run in that module.

If CT is run in another module that depends on the module that contains the extension, everything works as expected!

Expected behavior

"local" extension is run

Actual behavior

"local" extension is not run/found

How to Reproduce?

q_ct-junitext.zip

• mvn verify ✔️
• run test in Eclipse ✔️
• mvn quarkus:test ❌

Output of uname -a or ver

No response

Output of java -version

17.0.2, vendor: BellSoft

GraalVM version (if different from Java)

No response

Quarkus version or git rev

2.7.3.Final and 2.7.4.Final

Build tool (ie. output of mvnw --version or gradlew --version)

Apache Maven 3.8.4

Additional information

ℹ️ There is no issue when the extension is registered via @ExtendWith(...)!

[quarkus-bot]

/cc @stuartwdouglas

[famod]

/cc @geoand @aloubyansky (FTR)

[famod]

Turns out that the "Deployment Class Loader" that is used for discovering the tests does know the META-INF/services/org.junit.jupiter.api.extension.Extension, but the "Augmentation Class Loader" which is used for the actual execution doesn't.

[famod]

@aloubyansky @stuartwdouglas any thoughts/hints?
This inconsistency is pretty annoying when you rely on global JUnit extensions (for all test types and modes).
I'm under the impression that this problem and the one from #24257 would vanish if that serviceloader file would be known to the augmentation classloader...?

[aloubyansky]

aloubyansky@a0d8601 appears to fix it. It replaces the current TCCL which is the augment CL with the deployment CL. @stuartwdouglas what do you think about that?

[famod]

Interesting! I was thinking about that as well, but after seeing that ModuleTestRunner is using the augmentation classloader to run the Runnable from JunitTestRunner, I was under the impression that has to be/stay like that.
I'll give it a spin in my real project.

[famod]

@aloubyansky your change does seem to fix this issue! 👍
The annotation issue in #24257 requires more work/fixes, though.

[stuartwdouglas]

That approach seems correct, the fact it was not using it must have been an oversight on my part.

[aloubyansky]

Thanks @stuartwdouglas I'll open a PR then