40 lines
1.4 KiB
Markdown
40 lines
1.4 KiB
Markdown
|
# Sprint 153 Plan: Multi-Language Determinism and Replay Qualification
|
||
|
|
|
||
|
|
## Context
|
||
|
|
|
||
|
|
Sprint 153 formalizes replay and determinism qualification across languages so
|
||
|
|
constructive loop behavior can be trusted and audited before GA.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Goals
|
||
|
|
|
||
|
|
1. Build replay corpus and golden outcomes for multiple languages
|
||
|
|
2. Detect divergence in AST, text, merge, and diagnostics outputs
|
||
|
|
3. Classify determinism failures with actionable root causes
|
||
|
|
4. Publish qualification status consumable by release gates
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Steps
|
||
|
|
|
||
|
|
### Step 1769: Multi-language replay corpus schema (10 tests)
|
||
|
|
### Step 1770: Golden outcome comparator model (10 tests)
|
||
|
|
### Step 1771: Divergence classification and triage model (10 tests)
|
||
|
|
### Step 1772: Determinism gate policy model (10 tests)
|
||
|
|
### Step 1773: `whetstone_run_constructive_replay_suite` MCP tool (8 tests)
|
||
|
|
### Step 1774: `whetstone_compare_constructive_replays` MCP tool (8 tests)
|
||
|
|
### Step 1775: `whetstone_get_determinism_gate_status` MCP tool (8 tests)
|
||
|
|
### Step 1776: Determinism qualification packet model (8 tests)
|
||
|
|
### Step 1777: Replay and determinism report artifact (8 tests)
|
||
|
|
### Step 1778: Sprint 153 integration summary + regression (8 tests)
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Architecture Gate
|
||
|
|
|
||
|
|
- Replay must be reproducible with explicit environment fingerprinting
|
||
|
|
- Divergence packets must be structured and field-level traceable
|
||
|
|
- Determinism gate outcomes must be binary, auditable, and policy-bound
|
||
|
|
- Max 600 lines per header
|