[PIsberg]

Prepares 1.14.1. Merging publishes nothing: the tag does that, and it waits for the owner.

Why this is a patch

The owner chose the patch number on 2026-11-04, knowing docs/SUPPORT_POLICY.md classes this content as minor. About 10 Fixed entries are new detections, and four changes can turn a green build red, on purpose: tests that never ran, a benchmark check that never failed, a licence key that was never verified, and an update that can be lost. The changelog's Upgrading section and a new policy paragraph say this, so nobody learns it from a red build. japicmp cannot see any of it, because the signatures stay compatible.

Known limit, shipped on purpose

ATOMIC_NON_ATOMIC_UPDATE also reports a correct log(gauge.get()); gauge.set(queue.size());. Reproduced through the agent with a temporary corpus row: fired (1, HIGH). It takes the agent, an explicit failOn of HIGH or lower, and minTrust below VERDICT; the default failOn = NONE is unaffected. No small fix is correct, since runtime values look the same as a real balance.set(balance.get() + deposit). Tracked in #1187.

What changed

• Version: bumped to 1.14.1 across the reactor, versions/pom.xml, the fixtures, the corpus and the install docs.
• japicmp baseline: 1.13.2 to 1.14.0. A stale baseline protects nothing.
• @since: four tags said 1.15.0, a version this release is not; they now say 1.14.1.
• bump-version.sh (separate commit): it matched async-test-lib: only, so README's agent line stayed on the previous release, in the 1.14.0 bump and again here. A reader copying that snippet got a library and an agent from two releases.

How it was verified

• mvn --batch-mode clean verify "-Dlicense.mock.mode=true": BUILD SUCCESS in 19:44 on JDK 26, Maven 3.9.16, before the final docs edits. japicmp ran against the 1.14.0 from Central: ~/.m2 held local builds under those coordinates, moved aside first.
• Docs edits after that run: DocsIndexCoverageTest, DetectorCatalogCoverageTest, SatelliteVersionsTest, JapicmpBaselineFreshnessTest re-run and green.
• SatelliteVersionsTest.installSnippetsNameTheReactorVersion was seen failing on README.md: ...async-test-agent:1.14.0 before the fix. A trial bump to 1.14.2 with the fixed script moved that line and left the baseline alone.
• mvn -f corpus-eval/pom.xml test against this tree's jars: green with the probe row removed.
• Not run: the downstream sweep, the Gradle examples, and JDK 21 and 25 locally; CI covers the JDKs.

Docs

docs/CHANGELOG.md, docs/SUPPORT_POLICY.md, docs/DETECTOR_CATALOG.md, README.md, docs/QUICK_REFERENCE.md, docs/DISTRIBUTION.md, .claude/SKILL.md.

Guardrails

No @AI* annotation changed. The only main code touched is four @since tags, and none of it adds a fact a guardrail should carry.

Still open after this

• [P2] ATOMIC_NON_ATOMIC_UPDATE reports a get then set whose stored value does not depend on the read #1187: the false positive above.
• [P3] Gradle plugin 1.14.0 is not on the Plugin Portal: GRADLE_PUBLISH_KEY/SECRET secrets are missing #1114: the Gradle plugin needs GRADLE_PUBLISH_KEY and GRADLE_PUBLISH_SECRET, or 1.14.1 will miss the Plugin Portal as 1.14.0 did.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!