T-T3QJ-retire-tests-parity-harness

Status: closed/done · Impact: low · Complexity: small

Goal

tests/parity/ is a finished Python→TypeScript migration-parity harness whose runner was deleted months ago. Its fixtures are orphaned — nothing in the test suite, quality_checks, lefthook, or CI loads them — yet six source-file comments still describe it as a live gate, which misleads readers and rots as the tree moves. Delete the directory and the stale references.

Today

Location	Role today
`tests/parity/`	Orphaned migration-parity fixtures. The `run_parity.py` runner that consumed them was deleted 2026-06-01; the validator goldens were dropped 2026-06-13. Nothing loads what remains.
`tests/parity/canonical/`	17 `.input.json` / `.expected.txt` pairs capturing `canonical_json` byte-equivalence — dead data with no harness reading it.
`tests/parity/exit_codes.json`	Static lease exit-code reference table; no code reads it.
`tests/parity/README.md`	Claims the fixtures are “replayed by the `bun:test` suite” — inaccurate; they are not loaded by any test.
`plugin/lib/util/naming.ts`	Docstring (line 7) cites “a parity test in tests/parity/”.
`plugin/lib/model/ops/validate.ts`	Docstring (line 28) cites the byte-identical copy as “locked by tests/parity/”; a second comment (line 671) attributes the exit-2→1 delta to “the parity” gate.
`plugin/lib/services/lease/canonical_json.ts`	Docstring (lines 17-18) says “the parity harness verifies this implementation against” `tests/parity/canonical/`.
`plugin/lib/services/lease/canonical_json_cli.ts`	Docstring (line 8) says it is “backed by the tests/parity/canonical/ fixtures”.
`plugin/lib/services/lease/tree.ts`	Docstring (line 5) references “the parity-gate harness”.
`plugin/lib/services/lease/tests/canonical_json.test.ts`	Comment (line 7) says the fixtures are “gated by run_parity.py” — a file that no longer exists.

Proposed

tests/parity/ is gone. No .ts source comment references tests/parity or run_parity, and no comment describes the deleted parity harness/gate as if it were live. The byte-equivalence coverage that matters survives where it already lives: the hardcoded unit tests in plugin/lib/services/lease/tests/canonical_json.test.ts. bun test and bunx tsc --noEmit stay green.

Approach

Delete the tests/parity/ directory (canonical/, exit_codes.json, README.md).
Edit the six source comments so each describes the surviving unit-test coverage instead of the deleted harness — drop every tests/parity/ / run_parity.py path citation and every “parity harness/gate” mention that implies a live gate. Leave the legitimate “byte parity” wording in canonical_json.ts (line 24) that explains why the serializer is hand-rolled.
Run bun test and bunx tsc --noEmit to confirm nothing depended on the fixtures.

Files to touch

Location	Kind	Change
`tests/parity/`	delete	Remove the orphaned parity harness directory (fixtures, `exit_codes.json`, `README.md`).
`plugin/lib/util/naming.ts`	modify	Drop the “parity test in tests/parity/” citation from the module docstring.
`plugin/lib/model/ops/validate.ts`	modify	Remove the “locked by tests/parity/” citation (line 28) and the “locked by the parity” gate attribution (line 671); keep the documented exit-2→1 config-error delta note, without the dead-harness reference.
`plugin/lib/services/lease/canonical_json.ts`	modify	Rewrite the docstring (lines 17-18) to point at the surviving unit tests instead of the deleted parity harness; keep the line-24 byte-parity serialization rationale.
`plugin/lib/services/lease/canonical_json_cli.ts`	modify	Drop the “backed by the tests/parity/canonical/ fixtures” citation.
`plugin/lib/services/lease/tree.ts`	modify	Remove the “parity-gate harness” reference from the docstring.
`plugin/lib/services/lease/tests/canonical_json.test.ts`	modify	Replace the “gated by run_parity.py” comment with a note that these hardcoded tests are now the canonical `canonical_json` coverage.

Acceptance criteria

AC-1: tests/parity/ no longer exists (test ! -e tests/parity succeeds).
AC-2: grep -rn -E "tests/parity|run_parity" plugin/ --include="*.ts" returns no matches.
AC-3: bun test passes — the hardcoded canonical_json unit tests in plugin/lib/services/lease/tests/canonical_json.test.ts stay green.
AC-4: bunx tsc --noEmit passes.

Out of scope

The closed/done task documents that mention tests/parity as historical context — T-DHUF, T-JO4I, T-LIYW, T-P3HA, T-YNJO — and their generated site/ mirrors. They are immutable historical records, not live references; rewriting them is revisionism, not dangling-reference cleanup.
The canonical_json implementation and its unit-test assertions. Only comments change here; behavior and coverage are untouched.

Dependencies

none

Discovery context

Surfaced 2026-06-28 while reviewing the monorepo migration. A check of tests/parity/ confirmed the run_parity.py runner was deleted 2026-06-01 and the validator goldens dropped 2026-06-13, leaving the directory orphaned. The closed T-YBKU retired plugin/scripts and plugin/validators but never removed this directory, despite a validate.ts comment implying it would.

Post-mortem

Captured by /sdlc:task-work on 2026-06-28. PR: pending.

Acceptance criteria coverage

AC-1: agent-manual — test ! -e tests/parity succeeded after git rm -r tests/parity (35 files removed).
AC-2: auto — command grep -rn -E "tests/parity|run_parity" plugin/ --include="*.ts" returned no matches; a follow-up sweep for “parity harness”/“parity-gate” phrasing in the six edited files also found none.
AC-3: auto — bun test plugin/lib/services/lease/tests/canonical_json.test.ts passes 10/10. The full suite’s lone failure (task_auto_define behavioural test) is a pre-existing environmental flake — it fails identically from the main checkout with this branch’s files untouched, on git fetch origin inside a throwaway temp repo — not a regression from this change.
AC-4: auto — bunx tsc --noEmit exits 0.

What worked

The deterministic relevance check immediately caught that the misleadingly-named PR #502 (chore/retire-tests-parity) had only added this task’s spec, not done the work — so the task was genuinely actionable.
The change was exactly as scoped: pure comment edits plus deletion of orphaned, unloaded fixtures. tsc and the canonical_json unit suite stayed green with zero adjustment, confirming nothing actually depended on the fixtures.

Friction and automation gaps

The baseline-gated quality gate reported 3 false-positive new-drift findings — a bun test dashboard line whose PID is not normalized (24505 vs the baseline’s 89733), and two rumdl summary-count lines (Found 30 issues in 10/648 files) that shifted only because this branch legitimately deleted one markdown file (tests/parity/README.md), changing the file-count denominator. The gate’s output normalizer should also mask process IDs and treat tool summary/count lines (issue counts, N/M files) as non-gating, so a legitimate file deletion or a non-deterministic PID does not surface as new drift. → T-BQRU-quality-normalize-ports-pids-timings (PID facet), T-BCNP-quality-gate-ignores-summary-and-corpus-lines (summary/count-line facet)
The task_auto_define behavioural test needs a reachable origin remote inside a throwaway temp git repo; in this sandbox that git fetch origin fails, so the full bun test verb exits non-zero on every run regardless of the change under test. That makes bun test an unreliable gate signal here — the test should provision a local bare origin for its temp repo (or skip when no network/remote is available) so the suite is hermetic. → T-03JW-task-auto-define-test-hermetic-origin

Spawned follow-up tasks

T-BQRU-quality-normalize-ports-pids-timings — linked (existing open/ready owner of the PID-normalization facet of friction bullet #1).
T-BCNP-quality-gate-ignores-summary-and-corpus-lines — linked (existing owner of the summary/count-line facet of friction bullet #1).
T-03JW-task-auto-define-test-hermetic-origin (https://github.com/sksizer/dev/pull/505) — spawned (Upstream-plugin / sdlc-meta) for friction bullet #2.

← Back to Tasks