Patch test op: distinguish JSON boolean from number in Python

[DevChiniwala]

Summary

• Bug: The RFC 6902 test operation in both Python references (coordinator-py, chap-analytics) used bare != to compare values. Python's bool is a subclass of int, so True == 1 and False == 0 evaluate as equal — but JSON distinguishes them: true is a boolean, 1 is a number. The TypeScript reference's deepEqual correctly distinguishes types via typeof.
• Impact: A decide.override carrying a test op that compares a boolean against a number (e.g. {"op": "test", "path": "/confident", "value": true} against actual value 1) succeeds in Python but fails in TypeScript — a cross-language consensus violation that produces divergent audit chains from identical inputs.
• Fix: Add _json_equal() — a JSON-type-aware deep comparator that checks isinstance(…, bool) before falling through to ==, mirroring the TS deepEqual. Applied in both coordinator-py and chap-analytics. Added 7 conformance vectors (2 accept, 5 reject) to json-patch-vectors.json so both references are held to the same rule.

Reproducer

from chap_coordinator.patch import apply_json_patch
# Before fix: passes (wrong). After fix: raises PatchError (correct).
apply_json_patch({"status": 1}, [{"op": "test", "path": "/status", "value": True}])

import { applyJsonPatch } from './packages/coordinator/src/patch.js';
// Already correct: throws PatchError("'test' failed at /status")
applyJsonPatch({status: 1}, [{op: 'test', path: '/status', value: true}]);

End-to-end through the coordinator: the same decide.override envelope produces {"result": {"state": "completed"}} in Python but {"error": {"code": -32012}} in TypeScript.

Prior art in the repo

The whisper profile's _same_value() (whisper.py:29-42) already handles this pattern correctly — it checks isinstance(a, bool) before comparing. The JSON Patch test op was the remaining site where the check was missing.

Test plan

• All 44 Python patch conformance tests pass (including 7 new vectors)
• All 11 TS patch conformance tests pass (including 7 new vectors)
• 493 Python tests pass, no regressions (1 pre-existing Windows-only Unicode failure)
• End-to-end decide.override now agrees across both coordinators

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[DevChiniwala]

The current-head CI workflow is awaiting maintainer approval and has not run. Could a maintainer approve the validation workflow so the shared Python/TypeScript patch vectors can be checked upstream? The Copilot quota notice is not a code-review result, and no automated review pass is being claimed.

[brightbeamarsalan]

Thanks Dev, good catch. I reproduced it on main: a decide.override whose test op compares true with 1 completed in the Python coordinator and was refused with -32012 in TypeScript. With this change both coordinators refuse it with the same code and message, and the new vectors hold both to that rule. The Python coordinator, chap-analytics and TypeScript patch tests pass locally. I’ll squash-merge once CI is green, and the fix will ship in 0.3.3 and chap-analytics 0.2.4.