Issue 6036: Support for Jackson3

[jolarsen]

New module json-jackson3 in media, examples, and performance tests. Based on the existing json-jackson module

• Source code from jackson-jakarta-rs-providers - branch 3.x. Using only jakarta - coordinates.
• Adapter-code from json-jackson modified to fit Jackson 3. Moved some code from DefaultProvider to MapperConfigurator due to immutable mappers and builder-pattern.
• Opted for separate version for jackson-annotations in pom.xml beacuse of new pattern - version "2.20" is the latest and needed for Jackson3.
• Updated tests to variations of NoCtorDeserModule since Jdk8Module is no longer an issue and expanded with POST
• Using JsonMapper rather than ObjectMapper
• Replaced JaxRs-references with JakartaRS in comments, classnames and imports

For discussion:

• Drop support for maxStringLength (in JacksonFeature and not look at MessageProperties) ?

Testing: Apart from updated tests in media / json-jackson3, I have done this:

• Tried json-jackson3 in various tests/integration (2322, 2689, 2673) and benchmark
• Tried out a locally built version with another project I am working on (server with 50-100 endpoints)
• FilteringProvider is least tested as there were "few" tests in json-jackson and I do not have any other cases at hand

[reneleonhardt]

Great work, thank you so much for your big contribution! ❤️

Please be aware: Copyright checking failed
https://ci.eclipse.org/jersey/job/ci-cd-integration/job/PR-6048/1/pipeline-overview/?start-byte=0&selected-node=22#log-22-854

[jolarsen]

@reneleonhardt I'm looking into the copyright-warnings. I am not familiar with the practices here, so I'll need some guidance

[reneleonhardt]

I'm not a Jersey maintainer, I guess glassfish-copyright-maven-plugin wants a copyright + license notice in every file, it will probably be enough to copy from another source file and change the year to 2026 for new files.
https://javaee.github.io/glassfish/wiki-archive/Copyrights.html

[jolarsen]

Jenkins stopped working after 1h23 mins - in the middle of jersey-tests-osgi-functional (287/387).
jersey-media-json-jackson3 built OK https://ci.eclipse.org/jersey/job/ci-cd-integration/job/PR-6048/4/consoleFull
Jenkins Error: "Creating placeholder flownodes because failed loading originals.
ERROR: Cannot resume build because FlowNode 22 for FlowHead 1 could not be loaded. "

Can someone with access trigger a new build? Or should I just commit a NOOP-change ?

[reneleonhardt]

Megabytes of logs... I remember when unit tests were only allowed to print dots, good times 😄

[senivam]

@jolarsen Thank you for adding the Jackson3 support! In order to avoid dependency convergence, I've also adopted Jackson 2.20 (my commit is added to your PR). In total, this should work, and all checks should be OK now.

[senivam]

Oops, I had to re-write that because I didn't notice that the 2.20 is only applicable for Jackson annotations.

[jolarsen]

@senivam Jackson3 uses com.fasterxml for jackson-annotations - and needs 2.20 for JsonFormat.POJO
Jackson-annotations started a new release-numbering (2.20 vs 2.20.0). This is a bit confusing and we will have to see what versions of jackson-annotations are included in databind 2.2x.y and modules.
Running dependency-convergence.sh or mvn enforcer locally does not report all problem.

But lets try to find a path to complete the PR. There are many users who want to do the Jackson3-migration - which is a fair amount of changes - new coordinates, immutable mappers, builders, etc.

[senivam]

@jolarsen , I was hoping to solve this without exclusions. Dependency convergence check should work similarly both locally and on Jenkins. It, however, requires previously built artifacts to be prepared for the check. As I'm running it with Jackson 2.20.1 adopted, it does not fail for jackson annotations convergence.

[jolarsen]

When I tried locally - mvn enforcer reported problems with tests/integration/jackson-14 - but not media/json-jackson.
Thus the exclusions to avoid another 2 hour build showing new discoveries. databind is the one pulling inn annotations.
I agree that we should avoid exclusions and keep these 3 jackson-versions in sync (ref jackson-bom version numbers).

[senivam]

@jolarsen OK, clear. I hope this time the build will finally pass OK.

[jolarsen]

Jjenkins resigned with "pool-979-thread-1: Retried waiting for GCLocker too often allocating 1250002 words"

[senivam]

Finally it passed fine

[jolarsen]

@senivam Thanks for helping out !
Now we should get some feedback on the implementation