[dmccoystephenson]

Summary

Stage B (unit-test expansion) cycle: two characterization tests are added for wire-format behavior that was previously unexercised. No production code is changed.

• test_a_base_url_with_a_path_keeps_it_in_front_of_the_endpoint — a base URL under a path prefix (a trace server behind a reverse proxy, …/trace//) is posted to /trace/api/metrics: the prefix is kept and every trailing slash is stripped. This locks in the README's POST {base_url}/api/metrics for more than a bare host.
• test_non_ascii_names_and_tags_arrive_intact_as_utf8_json — a non-ASCII event name and tag values (Latin-1 accents and an astral-plane emoji) decode on the server to exactly what was reported, sent with the exact Content-Type: application/json; charset=utf-8 header.

Test plan

• python3 -m unittest locally — Ran 64 tests, OK
• CI Build — test (3.8), test (3.10), test (3.12)

Issues

No tracking issue — gap found during triage. #16 is skipped: it remains blocked on an owner decision about the README's "tags always holds at least version" promise at the MAX_TAGS cap and on a coordinated change in trace-client-java (see the comment on #16).

trace-client-java

No Java-visible behavior changed (tests only).

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 (performed inline in the same session; not an independent review):

• Scope: PASS — gh pr diff 22 touches only tests/test_trace_client.py (+20 lines, two new test methods).
• Tests-new: PASS (n/a) — no new public names; __all__ unchanged.
• Tests-fix: PASS (n/a) — no bug fix; both tests are characterization tests of current behavior and pass against unchanged trace_client/trace_client.py.
• Sibling structure: PASS — both tests follow the neighbors: TraceClient(self.base_url, ...), capture.arrived.wait(5), close() before asserting, assertion messages naming the promise.
• Sibling renames: PASS (n/a) — nothing renamed.
• Docs: PASS — README "The wire format" (POST {base_url}/api/metrics) already describes the behavior locked in; no doc change needed.
• Issue resolution: PASS (n/a) — no Closes; The version tag is added even when it pushes an event past MAX_TAGS #16's skip reason is recorded in the PR body.
• CI: PASS — test (3.8), test (3.10), test (3.12) green on the PR head (run 39770539380).
• Stdlib-only: PASS — no import lines added.
• Py3.8 floor: PASS — test (3.8) green.
• Never-raises / never-blocks: PASS (n/a) — no production code changed.
• Log level: PASS (n/a) — no _LOG calls added.
• Version agreement: PASS — no version string touched.
• Exports: PASS (n/a).
• Java parity noted: PASS — PR body states no Java-visible behavior changed.

Observation (judgment, non-blocking):

• tests/test_trace_client.py:116 — test_non_ascii_names_and_tags_arrive_intact_as_utf8_json: json.dumps keeps its default ensure_ascii=True, so the body bytes are actually ASCII with \u escapes (which is still valid UTF-8). The test asserts what the server decodes and the exact Content-Type header, not the byte encoding. That is deliberate: a later switch to raw UTF-8 bytes would be equally correct on the wire and should not break this test.

Verdict: tests-only, CI green, no do-not-auto-merge path touched.

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

---

drafted by Claude on behalf of Daniel Stephenson