SLING-11836 - Add Jakarta JSON support to the Sling Starter

[rombert]

Add the glassfish osgi resource locator to the boot feature. This shim is explicitly invoked by the JsonProvider to properly function in OSGi environments 1.

[enapps-enorman]

Perhaps they would accept a patch that uses the osgi service loader mediator instead of that glasssfish workaround? We already have the org.apache.aries.spifly.dynamic.bundle for dealing with those things.

[rombert]

@enapps-enorman - I did try using the SPIFly bundle locally, but I think I did not do it properly. There is some discussion about it at osgi/osgi#434 and also a larger discussion about ServiceLoader at osgi/osgi#372.

I will try again the SPIFly approach.

[enapps-enorman]

@enapps-enorman - I did try using the SPIFly bundle locally, but I think I did not do it properly. There is some discussion about it at osgi/osgi#434 and also a larger discussion about ServiceLoader at osgi/osgi#372.

I will try again the SPIFly approach.

I haven't done much testing, but I think something like this should work: eclipse-ee4j/parsson@master...enapps-enorman:parsson:use-osgi-serviceloader-mediator

[rombert]

@enapps-enorman - thanks for looking into this. I built your patched version of Parsson, applied it to the starter master branch

git diff
diff --git a/src/main/features/boot.json b/src/main/features/boot.json
index fa62734..fe939d7 100644
--- a/src/main/features/boot.json
+++ b/src/main/features/boot.json
@@ -89,7 +89,7 @@
             "start-order":"1"
         },
         {
-            "id":"org.eclipse.parsson:parsson:1.1.1",
+            "id":"org.eclipse.parsson:parsson:1.1.2-SNAPSHOT",
             "start-order":"1"
         },
         {

but the build failed with

0]]
[ERROR] [requirements-capabilities] org.eclipse.parsson:parsson:1.1.2-SNAPSHOT: Artifact org.eclipse.parsson:parsson:1.1.2-SNAPSHOT requires [org.eclipse.parsson/1.1.2.SNAPSHOT] osgi.extender; filter:="(osgi.extender=osgi.serviceloader.registrar)" in start level 1 but no artifact is providing a matching capability in this start level.

What am I missing?

[rombert]

I went down the SPIFly route and it seems to work fine. I filed alternate PR #156 so we can compare the two and decide on a solution on the short-term.

Of course, removing the need for dynamic weaving by patching Parsson would be even better.

[enapps-enorman]

What am I missing?

Looks like you figured this out in #156 by moving the org.apache.aries.spifly.dynamic.bundle bundle (and related bundles) to start level 1.

[rombert]

Closing in favour of the other PR.