gnodet commented on PR #13945:
URL: https://github.com/apache/maven/pull/13945#issuecomment-6834701999

   <!-- gnodet-bot review -->
   
   CI analysis for the latest run (run 
[43286914157](https://github.com/apache/maven/actions/runs/43286914157)):
   
   **Two failures, different root causes:**
   
   ### 1. `integration-tests (ubuntu-26.04, 25)` — REAL REGRESSION ❌
   
   **Test:** `MavenITmng5102MixinsTest.testWithPath`  
   **Error:** `groupId cannot be null` during 
`maven-install-plugin:3.2.0:install` on the `child` project.
   
   The child POM in `mng-5102-mixins/path/child` has no `groupId` — it inherits 
it from the parent via the 4.2.0 POM model. The install plugin failing with 
`groupId cannot be null` means the effective model for the child is not 
inheriting `groupId` from its parent.
   
   This failure does **not** occur on `master`'s CI (run 43287919968 — 
`integration-tests (ubuntu-26.04, 25)` → SUCCESS), so it is introduced by this 
PR.
   
   The suspect is the caching change in 
`DefaultModelBuilder.clearRequestScopedCache()` — specifically the new 
behaviour that **preserves** `SourceCacheKey` entries while clearing other 
request-scoped entries. If a `SourceCacheKey` cached from a parent model lookup 
is returned for the child model lookup, the child would get a model without an 
inherited `groupId`. The issue is JDK-25-specific (passes on JDK 17 and JDK 
27), suggesting a timing or execution-order dependency in how the caches are 
populated.
   
   Needs investigation: does `clearRequestScopedCache` correctly handle 
multi-module builds where the child project's cache lookup could match a 
`SourceCacheKey` for a parent-relative path?
   
   ### 2. `integration-tests (macos-latest, 17)` — LIKELY FLAKY ⚠️
   
   **Test:** 
`MavenITmng8445DiDumpTest.dumpIsPassiveWithSequentialAndParallelBuilders()[13]` 
 
   **Error:** `Expected more than one Maven worker` — the test asserts parallel 
build workers exist, but got sequential.
   
   This is a timing-sensitive parallel-build test. The same test passed on 
macOS JDK 25 and JDK 27 in this same run. This looks like pre-existing 
flakiness (CI runner load on macOS JDK 17 caused sequential execution). Can be 
re-run to confirm.
   
   ---
   
   The macOS failure can be retried, but the ubuntu-25 MNG-5102 failure needs 
to be fixed before this PR can merge.


-- 
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]

Reply via email to