[dmccoystephenson]

Stage B — unit-test expansion (the server-wide tag merge and the enabled: line reader). Test code only; no production change.

Summary

• withServerWideTags_dropsNullTagsLetsTheEventWinAndNeverModifiesTheCallersMap: the helper's own Javadoc promises (null key/value counts as absent, the event's tag wins, the caller's map is untouched) were asserted only end to end, and its null-dropping cannot be reached through report() at all, because withVersion drops nulls first. The key order (event tags first, then server-wide) and an event with null tags are now pinned too.
• withServerWideTags_addsNoneToAnEventAlreadyAtOrOverMaxTags: the cap was covered only for an event below MAX_TAGS. This pins that an event at the limit gains nothing, that one already over it (as withVersion produces today, see withVersion pushes an event already at MAX_TAGS over the server's tag limit #13) is neither trimmed nor grown, and that one short of it takes exactly the first server-wide tag.
• serverWideConfig_aValuelessEnabledLineIsSkippedButACommentOnlyOneIsTheSwitch: a characterization test. A bare enabled: is passed over and a later enabled: false disables, but enabled: # ... captures #, becomes the switch, and a later enabled: false is ignored (fail-open). Filed as A comment-only enabled: line in the server-wide config shadows a later enabled: false #27 for a maintainer decision rather than changed here. The test pins the current behaviour and should be updated by whichever PR settles A comment-only enabled: line in the server-wide config shadows a later enabled: false #27.

Test plan

• mvn -B verify locally on JDK 21: Tests run: 84, Failures: 0, Errors: 0, Skipped: 0 (81 before, 3 new)
• CI matrix build (8), build (17), build (21) green on the PR head (only java.util.ArrayList is added, which is available on Java 8)

Deferred issues

• withVersion pushes an event already at MAX_TAGS over the server's tag limit #13 (withVersion ignores MAX_TAGS) and A quoted enabled: "false" in the server-wide config does not turn reporting off #15 (quoted enabled: "false") were skipped. Both change the wire format or the opt-out contract and are waiting on a maintainer decision.
• A comment-only enabled: line in the server-wide config shadows a later enabled: false #27 was filed this cycle and is deferred for the same reason. A fix running the enabled: value through scalar() would likely settle A quoted enabled: "false" in the server-wide config does not turn reporting off #15 and A comment-only enabled: line in the server-wide config shadows a later enabled: false #27 together.

No tracking issue for the coverage work. The gap was found during triage.

This PR description was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

🤖 Generated with Claude Code

---

drafted by Claude on behalf of Daniel Stephenson

[dmccoystephenson]

Self-review rubric (head bcc67be):

• Scope: PASS. git diff --name-only origin/main...HEAD lists only src/test/java/software/stephenson/trace/TraceClientTest.java: three tests and one java.util.ArrayList import.
• Tests-new: PASS (not applicable). No public method was added; the three new tests cover the package-private withServerWideTags and parseServerWideConfig.
• Tests-fix: not applicable. No bug fix is in the PR. The one apparent defect is pinned as current behaviour and filed as A comment-only enabled: line in the server-wide config shadows a later enabled: false #27 rather than changed.
• Sibling structure: PASS. Names follow subject_behaviourInPlainWords, Arrange/Act/Assert sections match withVersion_dropsNullKeysAndValuesAndKeepsTheEventsOwnVersion, and the characterization comment follows the withVersion_addsTheVersionEvenToAnEventAlreadyAtMaxTags pattern ("Characterizes current behaviour, which issue #N questions").
• Sibling renames: not applicable. Nothing was renamed.
• Docs: PASS. No behaviour changed, so no README table or Javadoc moved out of step. The existing imprecision in the parseServerWideConfig Javadoc ("The first enabled: line ... is the switch", although a bare enabled: is skipped) is recorded in A comment-only enabled: line in the server-wide config shadows a later enabled: false #27 and not edited under a test-only cycle.
• Issue resolution: not applicable. No Closes reference. withVersion pushes an event already at MAX_TAGS over the server's tag limit #13, A quoted enabled: "false" in the server-wide config does not turn reporting off #15 and A comment-only enabled: line in the server-wide config shadows a later enabled: false #27 are deferred as maintainer decisions.
• CI: PASS. build (8), build (17) and build (21) are all green on bcc67be. Locally, mvn -B verify on JDK 21 ran Tests run: 84, Failures: 0, Errors: 0, Skipped: 0.
• No dependencies / Single file / Environment seam / Version agreement / Opt-out contract: PASS. pom.xml and src/main are untouched.
• Java 8: PASS. The build (8) leg is green, and only ArrayList (Java 1.2) and the existing LinkedHashMap APIs are used.

Finding fixed during review: TraceClientTest.java:400 had a characterization comment line that ran well past the file's usual comment width. It was reflowed in bcc67be.

Judgment call for the maintainer: TraceClientTest.java:407, the assertFalse(... "enabled: # set below", "enabled: false" ...) assertion, pins a fail-open behaviour on purpose. Whichever PR settles #27 (or #15, if the scalar() route is taken) is expected to flip it.

This review comment was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

---

drafted by Claude on behalf of Daniel Stephenson