FELIX-6870 - Run the integration tests against the code being built

[paulrutter]

Fixes FELIX-6870.

Stacked on #433, so it targets feature/FELIX-6759-Java-25-LTS rather than master. See Why not master below.

The problem

Modules across the repository pin released Felix sibling bundles as dependencies of their tests, so pax-exam and the bnd resolver exercise released artifacts instead of the code in this repository. A regression introduced here can pass CI unnoticed — which is exactly how the FilterImpl.WrapperCapability bug fixed in #554 survived on master until the HTTP tests were pointed at the local build in #571.

The change

Every such pin becomes a property that still defaults to the latest release, so each module keeps building on its own. CI builds and installs a producing module whenever it or one of its consumers changed, and overrides the property with the version built here:

producer	consumers	pin was
framework	configadmin, scr, http, gogo itests	7.0.5 / 7.0.1 / 6.0.1
configadmin	scr	1.9.0
utils	resolver	1.8.0
gogo bundles	gogo/jline, itest-shell, itest-jline	1.1.2 / 1.1.4 / 1.1.8

Two findings beyond the original survey:

• http/itest still pinned framework 7.0.5 — missed by FELIX-6869 - Run the HTTP integration tests on a current JDK and on this framework #571, which only covered jetty and jetty12.
• The gogo itest modules resolved against none of the code in this repository. They take their versions from an imported BOM that pins released bundles, and the import is itself stale (gogo.bom:1.0.3-SNAPSHOT against 1.0.7-SNAPSHOT in the reactor). They now carry explicit, overridable versions.

The version-resolving step evaluates the gogo modules inside the gogo reactor, because gogo-parent:7-SNAPSHOT is not published anywhere (verified: 404 on repository.apache.org/snapshots) and a standalone help:evaluate would fail on a clean runner.

Deliberately left alone

• framework.security stays pinned at 2.8.3: it is no longer built here (removed in the 8.0.0 framework line, FELIX-6807) and its only consumer, ConfigAdminSecurityTest, is already excluded.
• ipojo, since its removal is already proposed separately.

Why not master

resolver and configadmin cannot compile on JDK 21 or 25 on master — felix.java.version 6 and felix-parent 6 (release 7) respectively. CI hides this because those steps only run when the module itself changed, and neither has changed in a long time. Touching their poms makes CI run them and go red on two of the three matrix JDKs. #433 already raises both baselines, so stacking there is what makes this ticket implementable at all.

Verification

Run locally on JDK 25 against the in-repo bundles:

check	result
http reactor on framework 8.0.0-SNAPSHOT	✅ all 13 modules, 162 itest tests pass (was 7.0.5)
gogo reactor, all overrides incl. framework 8.0.0-SNAPSHOT	✅ BUILD SUCCESS; bnd resolver resolves both itest modules
resolver on utils 1.11.11-SNAPSHOT	✅ BUILD SUCCESS, 28 tests
help:evaluate for all 7 producer poms	✅ reactor-scoped form resolves every one

configadmin and scr integration tests could not be verified locally: on Windows they fail in pax-tinybundles with IOException: Access is denied from File.createTempFile, and control runs on the original pins fail identically (configadmin: 44 errors either way). That is a harness limitation of this machine, not a result of these overrides — CI on Linux is the real verdict for those two.

🤖 Generated with Claude Code