test(physics): Stage 2 hardening pins — scale domain, L2 excision, parity boundary #92

Closed

core-labs wants to merge 2 commits from test/stage2-physics-parity-hardening into docs/stage1-governance-boundary-freeze

core-labs commented

2026-07-20 20:45:18 +00:00

Owner

Summary

Stage 2 — Physics core hardening & parity completion (proof-focused; stacked on Stage 1).

Does not invent dual implementations. Pins executable evidence that:

Scale invariant 0 < k < 1 is the single _is_valid_fraction_decrease_scale used by obligation + geometric admission (PR #89 already on main; pin remains).
L2 identity helpers (_axis_projection, _mean_frame_coherence, 0.75 blend) are absent from identity.py.
GoldTether primary residual is geometric (no np.dot(psi…) authority); GoldTetherViolationError exists.
Ratifier outcomes are only ratified | demoted.
Active Cl(4,1) path smoke: geometric product, conformal embed, Gram, unitary require, modality sandwich.
Python/Rust parity boundary: dual-backend ops (geometric_product, cga_inner, versor_condition, versor_apply, embed_point) present in core-rs; rotor_power is Python-authoritative (no false PyO3 claim).
Digests: full 64-hex LE f64 SHA-256.

Files

tests/test_stage2_physics_hardening.py
core/cli_test.py (algebra suite includes Stage 2 + geometric checklist)

Test plan

uv run pytest -q tests/test_stage2_physics_hardening.py — 9 passed, 1 skipped
uv run core test --suite algebra -q — 100 passed, 51 skipped

[Verification]: Algebra suite 100 passed, 51 skipped; stage2 pins 9 passed, 1 skipped

## Summary **Stage 2 — Physics core hardening & parity completion** (proof-focused; stacked on Stage 1). Does **not** invent dual implementations. Pins executable evidence that: 1. **Scale invariant** `0 < k < 1` is the single `_is_valid_fraction_decrease_scale` used by obligation + geometric admission (PR #89 already on main; pin remains). 2. **L2 identity helpers** (`_axis_projection`, `_mean_frame_coherence`, `0.75` blend) are absent from `identity.py`. 3. **GoldTether** primary residual is geometric (no `np.dot(psi…)` authority); `GoldTetherViolationError` exists. 4. **Ratifier** outcomes are only `ratified` | `demoted`. 5. **Active Cl(4,1) path smoke**: geometric product, conformal embed, Gram, unitary require, modality sandwich. 6. **Python/Rust parity boundary**: dual-backend ops (`geometric_product`, `cga_inner`, `versor_condition`, `versor_apply`, `embed_point`) present in `core-rs`; `rotor_power` is Python-authoritative (no false PyO3 claim). 7. **Digests**: full 64-hex LE f64 SHA-256. ### Files - `tests/test_stage2_physics_hardening.py` - `core/cli_test.py` (algebra suite includes Stage 2 + geometric checklist) ## Test plan - [x] `uv run pytest -q tests/test_stage2_physics_hardening.py` — 9 passed, 1 skipped - [x] `uv run core test --suite algebra -q` — **100 passed, 51 skipped** [Verification]: Algebra suite 100 passed, 51 skipped; stage2 pins 9 passed, 1 skipped

core-labs added 1 commit 2026-07-20 20:45:19 +00:00

test(physics): Stage 2 hardening pins for scale domain, L2 excision, parity b1dc57ff22

Executable Stage 2 exit evidence: shared fraction-decrease scale domain,
identity L2 helper absence, GoldTether geometric residual path, RATIFIED|
DEMOTED-only ratifier, active Cl(4,1) product/conformal/Gram/GT/sandwich
smoke, dual-backend vs Python-authoritative parity boundary (no false
rotor_power Rust claims), and LE f64 SHA-256 digests.

[Verification]: algebra suite green (includes Stage 2 pins); stage2 9 passed 1 skipped

core-labs added 1 commit 2026-07-20 21:05:23 +00:00