test(integration): substrate-gate chain — composition of 7 invariant modules#425
Merged
neuron7xLab merged 1 commit intomainfrom Apr 26, 2026
Merged
Conversation
…modules
Single-file integration test proving the seven runtime-evaluable physics
invariant modules from this session produce non-contradictory verdicts
on a deterministic synthetic substrate state.
ANCHORED ingredients composed:
- core.physics.arrow_of_time (PROVENANCE_TIER = "ANCHORED")
- core.physics.anchored_substrate_gate (PROVENANCE_TIER = "ANCHORED")
- core.physics.thermodynamic_budget.bekenstein_cognitive_ceiling
(registered ANCHORED via INV-BEKENSTEIN-COGNITIVE in INVARIANTS.yaml)
EXTRAPOLATED ingredients composed (independent witnesses, do not veto):
- core.physics.cosmological_compute_bound
- core.physics.jacobson_observer_coherence
SPECULATIVE ingredient composed (does not veto anchored composite):
- core.physics.observer_bandwidth (downgraded in PR #421)
REGISTRY-ONLY (point-eval only, no state-eval):
- core.physics.simulation_falsification — exercised via
FalsificationLadder shape + reasoning_tier per signature.
Six scenarios:
1. admissible baseline — all axes pass
2. Bekenstein violation — composite inadmissible, BEKENSTEIN named
3. Arrow violation — composite inadmissible, ARROW named
4. both anchored axes fail — failure_axes == ("BEKENSTEIN", "ARROW")
(deterministic ordering per anchored_substrate_gate.py:144)
5. SPECULATIVE bandwidth violation — anchored composite still admissible
(tier separation explicit; bandwidth result reported separately)
6. same-state consistency — every runtime-evaluable axis evaluated on
baseline; tier metadata cross-checked; registry ladder shape +
reasoning_tier asserted
Negative verification: locally inverted composite logic in
anchored_substrate_gate.py (`admissible = bekenstein_holds and arrow_holds`
→ `admissible = bekenstein_holds`); scenario 3 failed with
`is_thermodynamically_admissible=True` despite `arrow_axis_holds=False`,
proving the integration test catches a compositional regression that
unit tests on individual modules cannot. Source restored before commit.
Quality gates:
pytest tests/integration/test_substrate_gate_chain.py 6/6 PASS
pytest tests/unit/physics/ + integration all green
ruff check clean (after one auto-fix for import sorting)
ruff format --check clean
black --check clean
mypy --strict tests/integration/test_substrate_gate_chain.py
clean (1 file, 0 issues)
validate_tests.py --self-check PASS 8/8
No mocked physics. No hidden randomization. No Hypothesis — deterministic
fixture only. Single substrate fixture; each scenario mutates exactly one
field except scenario 4 (intentional double-mutation, documented).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
neuron7xLab
added a commit
that referenced
this pull request
Apr 26, 2026
…stic cases (Task 1) (#426) Extends test_substrate_gate_chain.py from one comfortable baseline to deterministic boundary-adjacent regions. Each case named, mutation documented, expected verdict mapped to a specific invariant contract anchored on actual module source. Cases (10 deterministic, 1 enumeration sanity = 11 total): Bekenstein/gate (6): 1. near_zero_positive_energy — E=1e-30 J, no NaN/Inf 2. near_zero_positive_radius — R=1e-30 m, ceiling ≈ 0 3. very_large_radius_finite_energy — R=1e10 m, E=1 J, no overflow 4. bekenstein_just_below_bound — claim = ceiling × 0.999999 5. bekenstein_exactly_at_boundary — claim = ceiling (`<=` admissible) 6. bekenstein_just_above_bound — claim = ceiling × 1.000001 Arrow (2): 7. arrow_net_exactly_zero — ΔS=-1, ΔI=+1 (`>=` admissible) 8. arrow_net_just_below_zero — ΔS=-1, ΔI=+0.999999 Bandwidth (2): 9. bandwidth_gamma_equals_sigma_dot — Γ=Σ̇=5 (`>=` admissible); ANCHORED gate unaffected (tier separation) 10. bandwidth_gamma_just_above — Γ=5.01, Σ̇=5; ANCHORED gate STILL admissible Boundary semantics anchored from source: - bekenstein_axis_holds: observed_information_bits <= ceiling (anchored_substrate_gate.py) - arrow_axis_holds: net >= 0.0 (arrow_of_time.assess_arrow_of_time) - bandwidth_holds: slack_hz = bound_hz - rate_hz >= 0.0 (observer_bandwidth.assess_bandwidth_bound) Each case asserts: bekenstein verdict, arrow verdict, composite verdict. Bandwidth cases additionally assert tier separation — SPECULATIVE bandwidth violation does NOT veto ANCHORED composite. Quality gates: pytest tests/integration/test_substrate_gate_boundary_sweep.py 11/11 PASS in 0.31s pytest chain + boundary suites 17/17 PASS ruff check / format / black clean (one I001 auto-fix for import sorting) mypy --strict clean (1 file, 0 issues) No Hypothesis (deterministic boundary table per protocol §4 — property layer is post-T1 optional). No randomness. No mocks. Same fixture style as substrate-gate-chain test (PR #425). Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
neuron7xLab
added a commit
that referenced
this pull request
Apr 26, 2026
…ask 6) (#431) Custom subprocess-based mutation harness (no mutmut dependency) that proves the physics test suite catches meaningful contract regressions, not just stylistic changes. Mutants registered (all six per protocol §6): 1. anchored_ignores_arrow — composite ignores Arrow axis 2. anchored_ignores_bekenstein — composite ignores Bekenstein axis 3. failure_axes_drops_arrow — multi-failure drops ARROW 4. bandwidth_inverted — Γ ≤ Σ̇ becomes Γ < Σ̇ 5. cosmo_above_passes — claim above ceiling silently passes 6. sim_threshold_inverted — strict > becomes < in sim ladder Harness contract (tools/physics_mutation_check.py): - exact-string source mutation; pattern must appear EXACTLY ONCE (zero → SKIPPED_PATTERN_NOT_FOUND; multiple → SKIPPED_AMBIGUOUS; no silent expansion) - mutation applied → targeted pytest run → restore in finally - post-restore equality check against original source - working-tree-clean asserted via `git diff --exit-code` after run - exit codes: 0 killed, 1 survivor, 2 restore-failed, 3 pattern-not-found, 4 invocation-error - --list / --mutant NAME (repeatable) / --all / --fail-on-survivor Reproduced result on this commit: python tools/physics_mutation_check.py --all --fail-on-survivor → Killed: 6/6 (every mutant survived ZERO; targeted tests caught all six contract regressions) → git diff --exit-code: clean (no leftover mutation) This proves negative-verification at scale: each integration / unit test referenced by the harness fails when the source contract it guards is intentionally broken, then passes when source is restored. The harness is the encoded form of the negative-verification step performed manually for PR #425 (substrate-gate chain compositional regression catch). Forbidden behaviors omitted: - no persistent mutation - no mutated-file commits (verified by --exit-code check) - no broad shell magic - no deletion of user changes - no run across entire repo by default Quality gates: python tools/physics_mutation_check.py --all --fail-on-survivor 6/6 git diff --exit-code (post-run) clean ruff check / format / black clean mypy --strict clean (1 file, 0 issues) Optional mutmut adoption deferred per protocol "start custom; evaluate mutmut as optional follow-up after the custom harness is stable". Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Single-file integration test proving the seven runtime-evaluable physics invariant modules from this session produce non-contradictory verdicts on a deterministic synthetic substrate state.
Tier composition
Six scenarios
Negative verification (anchored)
Locally inverted composite logic (
admissible = bekenstein_holds and arrow_holds→admissible = bekenstein_holds); scenario 3 FAILED withis_thermodynamically_admissible=Truedespitearrow_axis_holds=False. Source restored before commit. Proves integration test catches compositional regressions unit tests cannot.Quality gate
No mocked physics. No randomization. No Hypothesis — deterministic fixture only.