[dmccoystephenson]

Summary

Stage B (unit-test expansion) cycle: three characterization tests for paths through the 0.7.0 coalescing code in report() that no existing test drove. No production code is changed.

• coalesce_infiniteValuesAreTheSameEventAsNoValue: +Infinity, -Infinity and no value are counted as one event. The existing coalesce_keysOnNameValueAndTagsButNotTagOrder only covers NaN.
• coalesce_aTagWithANullKeyOrValueIsLeftOutOfTheEventAndNeverThrows: a tag map with a null key and a null value goes through the coalescing key (new TreeMap<>(merged), which throws on a null key) without throwing, and is the same event as the map without them. The test acts as a regression guard for the never-throws promise if withVersion() ever stops dropping null tags.
• coalesce_aCountCarriesTheInstallAndServerWideTagsBeforeItsOwn: a count event keeps the install and server-wide tags: entries, with coalesced added last.

All three tests assert current behaviour, and no bug was found.

Triage notes (skipped issues)

All four open issues (#13, #15, #27, #31) ask for a maintainer decision on the wire format or the server-wide opt-out contract. Each one lists options, so none was picked up for autonomous implementation this cycle.

Test plan

• mvn -B verify run locally on Java 21: Tests run: 97, Failures: 0, Errors: 0, Skipped: 0 (94 before this change)
• CI matrix green on Java 8, 17 and 21

No tracking issue: this 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 (scored against the diff and CI run 40833499389):

• Scope: PASS. Only src/test/java/software/stephenson/trace/TraceClientTest.java is modified (+75 lines, 3 new tests). There is no formatting churn.
• Tests-new: PASS (not applicable). No public members were added. The three tests cover existing report() coalescing paths.
• Tests-fix: not applicable. No production change. The tests characterize current behaviour.
• Sibling structure: PASS. The names follow the coalesce_* series, use Arrange/Act/Assert comments, run against the loopback HttpServer fixture, and use @TempDir plus writeServerWideConfig for the server-wide file.
• Sibling renames: not applicable.
• Docs: PASS. There is no behaviour change, so the README and Javadoc are unaffected. The new tests agree with the documented order (the event's own tags, then version, install, server-wide tags and coalesced) and with the documented rule that an infinite value is sent as no value.
• Issue resolution: not applicable (no Closes).
• CI: PASS. build (8), build (17) and build (21) are green on head 3131c72. Locally, mvn -B verify (Java 21) ran 97 tests with 0 failures (94 before).
• No dependencies / single file / Java 8: PASS. pom.xml and src/main are untouched. The test code uses only Java 8 APIs (HashMap, Collections.singletonMap), and the Java 8 leg confirms this.
• Never throws / non-blocking / environment seam / version agreement / opt-out contract: not applicable (no production change).

Observations:

• TraceClientTest.java coalesce_infiniteValuesAreTheSameEventAsNoValue and coalesce_aTagWithANullKeyOrValueIsLeftOutOfTheEventAndNeverThrows: the "nothing more sent at once" check relies on a 200 ms Thread.sleep, as the existing coalesce_sendsTheFirstAtOnce... and coalesce_keysOnNameValue... tests do. A slow runner could only make that check pass more easily, never fail it, because a late count would be caught by the later exact-size assertion after close(). The flake risk therefore stays at the existing level.
• TraceClientTest.java coalesce_aTagWithANullKeyOrValueIsLeftOutOfTheEventAndNeverThrows: the comment says the TreeMap key is safe only because withVersion() drops null tags. This was confirmed against TraceClient.java withVersion (it copies only non-null keys and values) and report (new TreeMap<>(merged)).

Summary: the PR adds tests only, CI is green on all three JDKs, and no blocking findings remain.

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

---

drafted by Claude on behalf of Daniel Stephenson