[SLING-12306] Sling mock: JUnit 5 parallel execution results in unregistered adapter factories

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Testing
• Assignee: Unassigned
• Reporter: Henry Kuijpers
• Created: 29/Apr/24 13:06
• Updated: 29/Apr/24 13:06

Description

We have junit-jupiter.properties:

junit.jupiter.execution.parallel.enabled = true
junit.jupiter.execution.parallel.mode.default = same_thread
junit.jupiter.execution.parallel.mode.classes.default = concurrent 

And this results, randomly everytime, in lots of tests failing because of missing adapter factories.
 
We sometimes see that `boundAdapterFactories` is an empty list. Sometimes we see that `factoryCache` is a filled map, with all value lists empty.
 
We usually see the expected `descriptors` having variable content. One time it has 4 entries for o.a.s.a.r.Resource, the other run it has 5 entries.
 
I believe this is a concurrency issue somehow. I can reproduce it on older versions as well (previously we couldn't). So there could also be something with dependencies involved. 
 

---

More specifically, we seem to get a lot of these stack traces like this:

java.lang.RuntimeException: No page manager.
    at io.wcm.testing.mock.aem.context.AemContextImpl.pageManager(AemContextImpl.java:169)
    at io.wcm.testing.mock.aem.context.AemContextImpl.currentPage(AemContextImpl.java:229) 

(also posting this for findability on search engines)

And also a lot of cases where a Resource cannot be adapted to a ValueMap for example.

Which boils down to the adapter factory that can adapt a ResourceResolver to a PageManager, missing or actually not registered yet.

We do have the proper @ExtendWith etc setup.