search

quarkusio / quarkus

Quarkus: Supersonic Subatomic Java.
https://quarkus.io
Apache License 2.0
13.82k stars 2.69k forks source link

Cannot mock/spy class error, after adding quarkus-junit5-mockito dependency #37927

Open ennishol opened 10 months ago

ennishol commented 10 months ago

Describe the bug

After adding quarkus-junit5-mockito dependency, tests start to fail with Cannot mock/spy class error

Expected behavior

Test should work

Actual behavior

Tests fail

How to Reproduce?

con-testing.tar.gz

  1. Run ./mvnw quarkus:dev and press r to start testing
  2. Tests are failing
  3. When quarkus-junit5-mockito is removed with the test using @InjectMock, the failing tests become green

Output of uname -a or ver

No response

Output of java -version

openjdk version "21.0.1" 2023-10-17 LTS Corretto-21.0.1.12.1 (build 21.0.1+12-LTS)

Quarkus version or git rev

5.6.4

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

Apache Maven 3.9.6

Additional information

No response

quarkus-bot[bot] commented 10 months ago

/cc @geoand (testing)

famod commented 10 months ago

The problem is that Quarkus 3.x is fixating the old subclass mockmaker, which cannot mock e.g. record (like you do in your example). There is a hint and a workaround (!) in the 3.0 migration guide: https://github.com/quarkusio/quarkus/wiki/Migration-Guide-3.0#fixation-of-the-mockito-subclass-mockmaker

FTR, this also happens with Java 17.

@geoand are you planning on supporting the default inline mockmaker properly? I fear that the issue of Quarkus relying on the old and inferior mockmaker will only become worse. Until then I think the testing guide should contain a similar hint as the 3.0 migration guide (see above).

ennishol commented 10 months ago

@famod but then I will run again in the memory usage problem, right? #31251 I'll give it a try

ennishol commented 10 months ago

@famod unfortunately exclusion of mockito-subclass leads to OOM in integration tests, But at least for unit tests I can use it as a workaround

geoand commented 10 months ago

Right, we'll have to tackle the memory leak at some point and support inline properly.

famod commented 2 weeks ago

FWIW, I was just hit by subclass mock maker limitations when trying to @InjectSpy a spring-data-jpa JpaRepository that has a default method - error from mockito: "Cannot call abstract real method on java object!"

So +1 (again) to move on to inline.

gsmet commented 2 weeks ago

Could we install a callback somewhere to clean the mocks after test execution?