TNFR Testing Guide

This page owns local validation instructions. pyproject.toml defines tools and dependencies; the workflow guide describes CI, whose executable commands belong to the linked workflow YAML. Expected mathematics and contracts follow AGENTS.md and the source being tested. A passing test establishes its asserted behavior and domain, not a general physical theorem.

Run the repository tests

Run commands from the repository root with the interpreter of a separate environment. Install the editable project and the dependency groups used by the main CI test job:

python -m pip install -e ".[test,numpy,yaml,orjson]"
python -m pytest

test is the compatibility alias for test-all; NumPy is already a core dependency. There are no dev or all extras. Smaller extras such as test-unit install only their declared tools and may not support collection of the whole repository. Aggregate extras reference the smaller groups so their dependency bounds have one owner; the compatibility alias does not maintain another dependency list. The maintained benchmark instruments are standalone CLIs, so these groups do not install pytest-benchmark. The test-performance extra remains available as an explicit optional plugin for downstream tests; it does not define a benchmark-directory test suite.

The default testpaths in pyproject.toml selects the routine engine gate: nodal execution, operator/grammar contracts, numerical admission, public diagnostics, caches, CLI and SDK. It includes the complete operator and SDK directories and the listed production-field owners. The broad historical research/certificate directories are not part of every routine run. pythonpath = ["src"] imports the working tree, and addopts = "-m 'not slow'" excludes marked slow cases. No custom collection plugin or second test registry is involved. The test dependency requires pytest 7.4 or newer, supporting the standard testpaths wildcard configuration.

Use the explicit root path for the complete retained inventory, or a research module/directory when its owner changes:

python -m pytest tests --collect-only -q
python -m pytest tests -q
python -m pytest tests/physics/test_phase_alignment_metric.py -q
python -m pytest tests/mathematics -q

Explicit paths take precedence over testpaths. Research tests remain executable evidence for retained conditional results; they have not been declared obsolete merely because they are outside the routine gate. Add a new production-field regression to testpaths when it belongs in that gate. A routine pass is not a claim that every mathematical campaign was replayed.

Choose affected mechanisms through the theory-to-execution map, then inspect their source and tests. That map links shared implementations, representative controls and theorem owners; this guide does not maintain a second inventory of research results or individual test cases.

Shared evidence maintenance

The evidence workflow separates maintained artifact mechanics, model-specific audits and immutable archived programs. For changes to their shared owners, select exact-record, archive and restoration controls separately from scientific producers:

python -m pytest tests/research/test_artifact_io.py tests/research/test_frozen_source.py tests/scripts/test_restore_frozen_source.py tests/mathematics/test_validated_taylor_arithmetic.py tests/mathematics/test_validated_box_taylor.py tests/physics/test_sine_flow_kernel.py -q

Exercise duplicate/nonfinite JSON, exact tags before scalar comparison, unsafe or duplicate archive names, expanded-byte limits, altered hashes and refusal to replace an existing record. Restoration controls use disposable synthetic Git repositories and read-only inspection of the committed freeze. Verify the full pinned base, admitted supplemental files and destination rejection; do not create a real reserved workspace or invoke an evaluator during regression.

Retained Taylor reconstruction checks primitive shape/source admission and Horner/remainder/endpoint arithmetic. It does not prove stored derivative enclosures or replace a consumer's source, law, event or strict Picard checks. The shared sine field still needs independent edge-sum and jet controls for its layouts and held coefficients. Select changed consumers and existing read-only evidence audits in addition to these shared tests.

Reuse expensive parsed records and archive bytes through module-scoped fixtures; copy a fixture before a mutation test. Pure test helpers must not import other test modules' autouse fixtures. Keep no-execution guards scoped and reversible, so an audit cannot leak patched producers or subprocess calls into unrelated tests. Shared arithmetic must not collapse a producer and its independent mathematical expectation into the same implementation. Current-code regression does not update frozen evidence or authorize the pending research response.

Select research checks by contract

Select the changed mathematical owner and its consuming APIs through the theory-to-execution map. The map owns individual module/test links; proofs own model-specific hypotheses, constants and frozen preparations. This guide groups the obligations needed to choose coverage, rather than repeating each research result.

For the constitutive foundation audit, select the countermodels and complete-law controls whose premises are being reassessed. Check pressure selection, capacity and support freedom, auxiliary variational premises, phase closure, structural clock and event occurrence separately; no one channel's success supplies the other laws.

python -m pytest tests/physics/test_pressure_constitutive_scope.py tests/physics/test_constitutive_capacity_scope.py tests/physics/test_constitutive_support_scope.py tests/physics/test_constitutive_variational_scope.py tests/physics/test_phase_exchange_constitutive_freedom.py tests/physics/test_structural_clock_scope.py tests/physics/test_selection_birth_closure.py tests/physics/test_relational_sine_comparison.py tests/research/test_sine_constitutive_robustness.py -q

These existing controls preserve scoped alternatives and known nonselection results. They do not evaluate a new alternative at the currently acquired objects or close a new research gate. Formation, response and storage evidence remain associated with their own complete laws. Read retained evidence without replaying its producers; test each new constitutive consequence independently when its mathematical owner changes.

For claimed NFR representations and collective actions, also select tests/physics/test_epi_phase_representation_scope.py and tests/physics/test_operator_quotient.py; the routine gate covers registered operator and grammar contracts. A new constitutive comparison must test its actual full field, balance, matching order and law-specific derivative/source bounds. Existing sine controls do not certify a different current, even when target storage and the full tangent agree. Keep local-onset evidence separate from finite-horizon discrimination and from acquisition under the new law.

For class-mediated collective response, select both the primitive/report controls and independent full-support algebra. Rebuild the 27-node support, its (3,4,3) central degrees, common early derivatives and the first class-dependent receiver term. Retain paired probe/unprobed histories, all source errors, the complete phase-blind postcontact control, exact event work and actual conserved-mean shifts. Reflection and absent-probe controls must not invent a distinction that the interface hides. A tangent coefficient alone cannot replace a finite nonlinear remainder, fresh formation handoff or full-family identity.

python -m pytest tests/physics/test_sine_class_mediation.py tests/physics/test_sine_class_mediation_algebra.py tests/physics/test_sine_class_mediation_evidence.py -q

Before the protocol freeze, these controls must not evaluate an admitted reserved source. After a retained outcome exists, audit its primitives and arithmetic separately from report wiring; keep old formation/contact evidence unchanged and do not rerun frozen producers for unrelated changes. The read-only saved-evidence suite checks the archive/protocol association, preserved prospective proof, original source and endpoint budgets, exact interval inflation, work and identity margins, and all declared stopping predicates. It neither calls the assessor nor executes its archived producer.

For acquired-mediator effective memory, also select the memory reader/bound controls and independent full-support algebra. Check the 36-visible/18-hidden partition, rational spatial factors with separate ideal scalar enclosures, class-independent zero-lag kernel and class-dependent first derivative. Retain the hidden initial source; test its cancellation only in a paired tangent difference with the same full initialization. Verify the nonlinear residual bound against both probe and baseline, including exact zero budgets and failed sufficient separation without erasing available finite bounds.

python -m pytest tests/physics/test_sine_class_memory.py tests/physics/test_sine_class_memory_algebra.py tests/physics/test_sine_class_mediation_evidence.py -q

The stationary-map controls must keep visible mobility and distinguish linear paired cancellation from the nonlinear comparator's uncertain visible sources. They establish no invariant midpoint section or fast limit. The same-charge hidden perturbation checks the exact visible acceleration difference; the proof separately supplies existence within the finite formation image, without a measured reachable radius. Changes to shared response coefficients must preserve the frozen source/report arithmetic. These controls need no new acquisition or reserved campaign.

For two-probe nonlinear superposition, rebuild four continuations of one complete initial state with a common final readout. The delayed-only control carries its unprobed state to the second event. Check exact tangent cancellation, full-law reflection/sign equivariance, the carried bridge-gap curvature identity and the ideal cubic-amplitude, fourth-joint-time coefficient. Arbitrary source residuals must remain admitted rather than being projected onto a symmetric subset.

python -m pytest tests/physics/test_sine_class_superposition.py tests/physics/test_sine_class_superposition_algebra.py tests/physics/test_sine_class_mediation_evidence.py -q

Keep scalar-statistic cancellation distinct from intersection of all four endpoint record sets; neither proves continuous-history equivalence or that every record is shared. Test equality at each sufficient noise threshold, failed bounds without a fabricated positive response, and signed/zero impulses and boundary event times. Independent work and identity checks retain the actual preevent state, both conserved-mean changes and the separate radius/storage guards. These are conditional arithmetic and implementation controls, not a new acquired response or permission to replay the frozen class-mediation assessment.

For the finite nonlinear protocol, test the fixed degree-32 heat polynomial as an analytic coefficient, including rational history cancellation, actual-degree normalization, three cosine channels and the contraction tail. The full-law remainder must retain the odd internal corrections and the source bound valid on the longer horizon. Use independent coefficient or algebra controls; an unvalidated trajectory cannot replace the claimed enclosure.

python -m pytest tests/physics/test_sine_class_nonlinear_protocol.py tests/physics/test_sine_class_nonlinear_protocol_algebra.py tests/physics/test_sine_class_superposition.py tests/physics/test_sine_class_mediation_evidence.py -q

Distinguish true sign, recorded sign above 4*delta and disjoint nonlinear/tangent four-record sets above 8*delta, including equality and unavailable-sign cases. Check exact rational endpoints separately from outward display intervals, signed impulses and exact null histories. The new horizon admission must not enlarge the old superposition API's domain. Reused work/identity ledgers retain both preevent states and independent policy guards. These controls acquire no source or reserved response and must not rerun a frozen producer.

For organization-dependent nonlinear interaction, test exact common-channel cancellation before transcendental enclosure, the mediator-channel heat tail, two complete nonlinear remainders and independent source errors between classes. Keep the eight-reading recorded contrast budget separate from the sixteen-error comparison against an independently noisy zero-contrast alternative. Exercise exact nulls, strict boundaries, unavailable signs and independent work/identity guards.

python -m pytest tests/physics/test_sine_class_nonlinear_organization.py tests/physics/test_sine_class_nonlinear_organization_algebra.py tests/physics/test_sine_class_nonlinear_protocol.py tests/physics/test_sine_class_nonlinear_protocol_algebra.py tests/physics/test_sine_class_mediation_evidence.py -q

The unchanged design must retain its unresolved full-law verdict despite the signed heat contribution. Check that the independent nonlinear-error estimate still straddles zero with source and sensor errors removed; this proves a limitation of that estimate, not actual class equivalence or overlap of the original reading vectors. Independent edge algebra and heat quadrature are coefficient controls, not complete sine trajectories. The tests acquire no source and replay no reserved response. SDK wiring uses test_reduced_port_sdk_wiring_does_not_evaluate_research separately.

For the independent class comparison protocol, use unrelated polynomial controls for the two-source composition. Both sources must be admitted before either runs, and the first incomplete child must stop later execution. Reconstruct complete law, source, event carry, step arithmetic, counts and observations before consuming a child report. A saved band or status cannot replace that evidence; derivative generation remains an explicit premise.

python -m pytest tests/physics/test_sine_class_comparison_readout.py tests/physics/test_sine_class_comparison_protocol_algebra.py tests/physics/test_sine_class_readout.py tests/physics/test_sine_class_four_history_readout_algebra.py tests/research/test_frozen_source.py tests/scripts/test_restore_frozen_source.py -q

Keep nominal reference covers separate from the original correlated acquired families and their Cartesian outer covers. Check that source transport is added once, while the cubic prediction never narrows a forward interval. Test both open-prediction consistency checks, the transported width postcondition and the distinct strict eight/sixteen-error margins. A declared step/order policy does not guarantee the requested numerical width. Preserve unavailable and first failure evidence; these tests execute no selected reference or reserved response.

The comparison freeze has a separate read-only association and synthetic-helper gate: python -m pytest tests/physics/test_sine_class_comparison_freeze.py -q. Reuse the shared source inspector for archive/base checks. Admit both source recipes, fixed predictions, version policy and prospective proof prefix; compile only the named pure helpers for serialization/error-retention controls. Never execute an archived main or producer to validate the freeze.

The first retained comparison has its own read-only response gate: python -m pytest tests/physics/test_sine_class_comparison_evidence.py -q. Parse the large response once, reconstruct both complete branch trees with the shared evidence reader, then independently rebuild the cross-class contrast, source transfer and all observation criteria. Retained derivatives and Picard generation remain execution premises. This gate must not rerun the producer, regenerate a coefficient or replace missing response evidence.

For repeated joined interaction, check the full joined degree metric, its exact spectral bounds, the distinction between Euclidean and degree-weighted common-origin projections, and the preserved means of all eight branches. Rebuild recurrent work and storage from the new relative source family; it is not the original formation image. Test exact return bounds below the interval grid and the independent tangent residual allowance, including the common-mean counterexample and strict noise boundary. Reuse the comparison audit's module fixture to associate the nominal interval only after reconstructing its retained full response.

python -m pytest tests/physics/test_sine_class_repeated_interaction.py tests/physics/test_sine_lyapunov.py tests/physics/test_sine_class_comparison_evidence.py -q

These are conditional arithmetic and read-only evidence checks. They neither simulate the repeated schedule nor replay acquisition or a frozen producer. The quadratic tangent return and the nonlinear acute trapping proof retain different premises despite sharing the modified-energy kernel.

For the nonlinear collective interface, check the six-port/hidden-state partition, retained source term, reflection-odd quadratic feedback, and central pressure observers independently of the coefficient producer. The finite input family and its approximation error do not establish identity or work admission for arbitrary schedules. Exercise separate nominal amplitude tails and nonlinear initialization defects without assigning nominal parity to actual residuals. Simultaneous port jumps retain their quadratic cross terms.

python -m pytest tests/physics/test_sine_class_collective_interface.py tests/physics/test_sine_class_cubic_evidence.py tests/physics/test_sine_class_repeated_interaction.py -q

The cubic evidence module reconstructs the supplied comparison premise once before applying the interface certificate. It must not regenerate coefficients or a nonlinear response. The modified unit-delay pressure helper also requires the amplitude, superposition and repeated-interaction contract tests; it does not merge their distinct initial storage, contact or mean hypotheses.

For causal component composition, check both complete boundary rows against independent edge equations with the actual joined degrees. Exercise component charge exchange, internal and bridge storage, simultaneous signed event work, hidden initialization and primitive full-lift residual admission. The explicit cubic residual includes hidden quadratic feedback, nonlinear source interactions and omitted bridge terms.

python -m pytest tests/physics/test_sine_class_interface_composition.py tests/physics/test_sine_class_collective_interface.py tests/physics/test_sine_class_port_readout_evidence.py -q

The retained-response audit reuses its module fixture and reconstructed gap to check that this substitution preserves the existing memory distinction. It subtracts the new approximation defect once; actual-source transport and both recording errors are already present. No coefficient or response is regenerated. A tolerance for one impulse is not a uniform two-channel input-map certificate, and output closeness alone proves neither feedback existence nor exact storage, identity or work properties for an arbitrary substitute.

For distinct-neighbor nonadditivity, derive the donor/receiver mixed onset from independent incident-edge algebra. Reconstruct all normalized rows used by the finite phase and cubic-current bounds; the mediator's internal edges contribute alongside its two contacts. Check exact agreement with the collective descriptor while preventing the onset calculation from constructing unconsumed coordinate-memory partitions. Retain hidden quadratic recoupling, cubic phase feedback, amplitude tails and the four nonlinear initialization defects. The exact common linear source cancels only because all four histories start from the same complete state.

python -m pytest tests/physics/test_sine_class_neighbor_nonadditivity.py tests/physics/test_sine_class_collective_interface.py tests/physics/test_sine_class_superposition_algebra.py tests/physics/test_sine_class_interface_composition.py -q

Check the exact zero-input identities and preserve the scope of nominal opposite-input symmetry. Exercise the earlier nonlinear observation budget, strict independent-comparator threshold and separate work/identity admission. The reused event ledger receives an explicit pressure bound for the commuting distinct-port jumps; its old delayed donor default is not applicable. These controls check the implementation of the conditional analytic distinction using static algebra and rational envelopes. They neither evaluate selected time coefficients or full-law responses nor establish the mathematical proof or exclusion of all pairwise dynamical laws.

For the independent distinct-neighbor protocol, check optional event/readout selector admission, legacy defaults, complete state carry and retained arithmetic. Independent polynomial fixtures must exercise simultaneous and delayed distinct-port events, observed-node jumps, shared-prefix cancellation and first-failure behavior. Research controls rebuild the signed absolute-phase cover and the four nonlinear source defects; they keep numerical width, closed prediction overlap and four/eight-reading error decisions separate.

python -m pytest tests/physics/test_sine_class_readout.py tests/physics/test_sine_class_comparison_readout.py tests/research/test_sine_class_neighbor_forward.py tests/research/test_sine_class_neighbor_forward_evaluator.py tests/research/test_frozen_source.py tests/scripts/test_restore_frozen_source.py -q

The new research policy tests use the shared no-regeneration guard. Evaluator wiring uses unrelated temporary records and mocked scientific entry points; its attempt, exception, export-failure and no-retry controls do not execute the selected four-history response. Existing frozen default reports remain subject to their read-only reconstruction audits. A successful implementation test is not an independent full-law evaluation or a proof of source acquisition.

After freezing, run the separate read-only association gate: python -m pytest tests/research/test_sine_class_neighbor_forward_freeze.py -q. It pins archive, protocol, receipt and Git base, then checks primitive policy and dormant evaluator structure under the no-regeneration guard. The evaluator wiring test admits pinned archive/member bytes before compiling only function definitions; it cannot silently fall back to a mutable build script. Freeze inspection never invokes the reserved experiment or authorizes a retry.

For nonlinear excess storage and loss, run the response-free mathematical gate:

python -m pytest tests/physics/test_sine_class_storage_excess.py tests/physics/test_sine_class_neighbor_nonadditivity.py tests/physics/test_sine_class_collective_interface.py tests/physics/test_sine_class_interface_composition.py -q

Check full/tangent balances using their own original-coordinate states, simultaneous event work, nonzero ordinary tangent cross-energy, exact onset and static adjoint cones. Rebuild quartic, amplitude-tail and common-source errors independently, including the zero-input source defect. Exercise exact nulls, unresolved bounds, primitive admission and the separately supplied coordinate-to-storage observation allowance. The shared no-regeneration guard must block response and time-coefficient execution. A finite static sign is an analytic prediction, not retained storage-response evidence.

For the prepared passive-loss observer, check the complete full/tangent fields, original phase units, carried passive loss, simultaneous events, global work limits and first-failure evidence.

python -m pytest tests/physics/test_sine_class_storage_readout.py tests/physics/test_sine_class_storage_excess.py tests/mathematics/test_validated_taylor_arithmetic.py -q

Synthetic polynomial fields test wiring and arithmetic reconstruction; an unrelated small full-law reference checks the actual fields. Guards reject the selected storage word and historical scientific producers. Re-admit source, target, law and every consumed Taylor step before checking cached losses. Derivative and strict Picard generation remain execution premises. These controls prepare an instrument, not a frozen storage comparison or an independent evaluation of the selected storage prediction.

The first retained neighbor response has a separate read-only gate: python -m pytest tests/research/test_sine_class_neighbor_forward_evidence.py -q. Admit exact attempt, protocol, archive and sole outcome-member bytes before decoding. Reuse one module-scoped report reconstruction to check all 64 full54 steps and distinct-port events. Rebuild source, width, prediction and recorded separation decisions independently; cached endpoints/verdicts are checked, never used as their own premises. Selective corruptions exercise selector, carry and Taylor-arithmetic rejection. Shared guards block all scientific regeneration; derivative and strict Picard generation remain retained execution premises rather than authenticated by this arithmetic audit.

For the changed-input collective prediction, check grounded-path kernel coefficients, exact causal convolution weights, hidden quadratic feedback and independent linear initialization. Compare low orders with independently assembled edge equations; do not execute the old full-state coefficient producer as a substitute for this causal implementation. The new single mediator impulse needs degree-four work and its own mean shift.

python -m pytest tests/physics/test_sine_class_port_prediction.py tests/research/test_sine_class_collective_protocol.py tests/physics/test_sine_class_port_prediction_evidence.py -q

The evidence audit reconstructs retained endpoints and all consumed error allowances without rerunning the selected prediction or a complete-law response. Time-series error, interval rounding, physical source width, amplitude omission and recording errors remain separate. Success against the grounded tangent alternative does not prove nonlinear necessity, class discrimination or a reserved forward comparison.

For the independent changed-input full-law protocol, select the complete 54-coordinate producer, its retained-evidence reader and the research policy together. Check both primitive sources before any field execution, absolute phase lifts, the single mediator form event, all-coordinate carry and the global first-failure attempt budget. Rebuild retained Taylor endpoints before consuming readings; stored derivative and strict Picard generation remain separate execution premises.

python -m pytest tests/physics/test_sine_class_port_readout.py tests/research/test_sine_class_collective_forward.py tests/research/test_sine_class_collective_forward_freeze.py tests/research/test_sine_class_collective_forward_evaluator.py tests/mathematics/test_validated_taylor_arithmetic.py -q

The policy controls reconstruct the already retained causal prediction without regenerating its coefficients. They independently check target-plus-residual source association, one actual-family source allowance, closed nominal prediction overlap, the declared agreement tolerance and strict two-error recorded separation. Amplitude and nonlinear-initialization errors belong to the inspected reduced prediction, not the full-law forward enclosure.

This selection admits the producer and prospective freeze; it does not execute the selected forward response. Freeze controls inspect source, protocol and attempt boundaries. The evaluator wiring controls admit pinned archive/member bytes before compiling function definitions and exercise only temporary records with mocked scientific calls and Git inspection. They can be selected independently of the freeze audit. They check preflight rejection, exclusive attempts and retained projection/export failures without executing the selected producer or restoring its workspace. Numerical completion, resolution, prediction consistency and discrimination remain separate verdicts; an earlier passing prediction cannot narrow or substitute for an independent forward interval.

For the retained first full-law comparison, select the read-only response audit:

python -m pytest tests/physics/test_sine_class_port_readout_evidence.py -q

The audit reads the canonical response ZIP and preserves its exact uncompressed record and attempt association. It re-admits both absolute full-state sources, the form event, complete law and fixed schedule, then reconstructs retained Taylor endpoints before applying the source allowance once. Prediction overlap, the declared resolution tolerance and noisy-comparator separation are checked independently. Producer and coefficient regeneration remain prohibited; stored derivative enclosures and Picard generation remain explicit execution premises. This finite comparison does not establish physical identification or nonlinear necessity for the changed-input observation.

The forward-policy, forward-freeze and retained-prediction audits share the lazy execution guard. Each module opts in separately: scientific generators and their imported aliases, acquisition and restoration are blocked until fixture teardown. Subprocesses are prohibited except for the freeze audit's exact pinned read-only Git commands. Independent mathematical assertions and artifact hashes remain with each audit owner. The guard controls check lazy imports, alias coverage and restoration after failure or nested use:

python -m pytest tests/research/test_sine_evidence_helpers.py -q

For common-scale amplitude feasibility, check whole-interval coefficient homogeneity and higher-amplitude bounds, the complex-domain boundary, the fixed source/noise allowances and the separate work and identity verdicts. The supplied base coefficient is a conditional premise of the public calculator; finite scalar admission alone does not establish its association with the model.

python -m pytest tests/physics/test_sine_class_amplitude_feasibility.py tests/physics/test_sine_class_amplitude_feasibility_algebra.py tests/physics/test_sine_class_amplitude_feasibility_evidence.py tests/physics/test_sine_class_cubic_evidence.py -q

Independent algebra checks cover carried amplitude levels, weighted heat symmetry and the donor Laplacian in the exact jump work. Reconstruct the retained coefficient endpoints and tails before applying the conditional calculator; do not regenerate coefficients or a nonlinear response. Upper-scale work and radius bounds cover the interval by monotonicity, whereas conserved means require their own interval bounds. If the shared event ledger changes, also select the superposition, nonlinear protocol, organization, cubic and spatial contract owners. Preserve their default conservative bounds and frozen evidence.

For the spatial class observation, test the two-node observation, receiver-odd locality, complete quadratic phase feedback and the heat-only class-blind control. Exact low-order jets check the first class-sensitive time coefficient independently of the finite assessment. The higher-amplitude bound already compares classes; sum its three nonzero-history contributions only once.

python -m pytest tests/physics/test_sine_class_spatial_observation.py tests/physics/test_sine_class_spatial_observation_algebra.py tests/physics/test_sine_class_contrast.py -q

Preserve arbitrary source residuals and sixteen independent nodal reading errors; the separately noisy null comparison needs thirty-two. Reuse the full-coordinate coefficient owner instead of a second solver. Previously retained coordinates are prior information, not unseen responses. Test the original eight-reading decisions and cubic/heat contracts when their shared projection or decision helpers change.

The compact spatial assessment references the preceding complete coefficient bundle. Its read-only audit reuses the full coefficient-evidence checks and reconstructs the new two-node quadratic projection, higher-order bound and source/noise decisions. Select both evidence owners together:

python -m pytest tests/physics/test_sine_class_spatial_evidence.py tests/physics/test_sine_class_cubic_evidence.py -q

The reference prevents duplicating complete series; it does not authenticate their generation. No audit may execute a coefficient or response producer.

For the complete cubic-amplitude response, check all three full-coordinate variation levels, the gamma scalings, oriented-edge signs, quadratic odd-mode feedback and exact jump ancestry. Time-polynomial truncation and the higher-amplitude Cauchy remainder need independent controls. Fixed order 64 uses exponential-tail starts 65, 64 and 63, with each endpoint enclosure carried into its suffix. This analytic coefficient recurrence does not change the full-flow integrator's limits.

python -m pytest tests/physics/test_sine_class_cubic_response.py tests/physics/test_sine_class_cubic_response_algebra.py tests/physics/test_sine_class_contrast.py tests/physics/test_sine_class_nonlinear_organization.py tests/physics/test_sine_class_nonlinear_organization_algebra.py -q

Use module fixtures for the complete finite coefficient; admission and decision controls reuse it or use explicitly synthetic coefficients. Nominal amplitude parity cannot remove arbitrary actual-source errors. An unavailable complex-amplitude domain must withhold the full-response bound, while exact event-pairing identities remain separate. A true sign can coexist with a certified scalar noise-cancellation witness; neither implies that the complete recorded vectors overlap. Preserve the original heat method's unresolved outcome after shared decision refactoring. These controls evaluate no acquired source or reserved nonlinear response.

The retained coefficient evidence has a separate read-only audit:

python -m pytest tests/physics/test_sine_class_cubic_evidence.py -q

It verifies bounded archive/source associations and re-admits the primitive policy and all retained intervals. Rebuild the three-level event carry, time-polynomial endpoints and tails, class subtraction and complete source/noise/work decisions. Coefficient generation remains the archived execution premise; the audit neither regenerates the coefficients nor replays a nonlinear producer. Preserve the recorded packaging failure before the numerical attempt and the subsequent complete source archive.

For the four-history observation producer, select its full-state/event controls and the shared source-box kernel. Use unrelated sources: check both rows from independent edge sums, stationary states, exact donor-only jumps, all 54 carried coordinates, shared-prefix ancestry and the delayed-only unprobed control. Zero delay, final-time jumps and zero duration must preserve their declared events.

python -m pytest tests/physics/test_sine_class_readout.py tests/physics/test_sine_class_four_history_readout_algebra.py tests/mathematics/test_validated_box_taylor.py tests/physics/test_sine_two_port_readout.py tests/physics/test_sine_aperture_readout.py -q

The independent field/source/prefix controls rebuild both nodal rows and retained first derivatives, distinguish an acquired subset from its Cartesian cover, and reconstruct suffix receiver increments under nonzero source width. Raw endpoint subtraction is a separate outer bound. Exercise failure inside a segment, budget exhaustion before an event and explicit unattempted branches. All-four availability must not be inferred from a completed prefix or an individual reading. Domain guards establish smooth flow, not acquisition or identity. Interval widths contain source and numerical enclosure effects; no sensor error is supplied by the producer. Do not evaluate the prospective class-two source to tune this admission; retain every earlier frozen source and response unchanged.

For the matched four-history freeze, audit immutable protocol/archive/receipt associations and the canonical source cover from primitive arithmetic. Check the acquired-family assumptions separately from its Cartesian cover, fixed event ancestry, the unique-step budget and the isolated forward argument list. Read-only controls must reject altered bytes, undeclared inputs and any response or attempt claim at the freeze stage; they must not call acquisition, prediction, the reserved producer or an archived worker. Test open-band overlap/containment and strict zero/four-error/eight-error classification using unrelated synthetic intervals, including equality and unavailable evidence. Hash agreement establishes association, not authentication or the truth of retained derivative enclosures.

python -m pytest tests/physics/test_sine_class_readout_freeze.py tests/physics/test_sine_class_readout_protocol_algebra.py -q

The independent protocol arithmetic checks the fixed branch budget, source cover and strict observation criteria without executing a reserved flow. The freeze audit checks the archived execution boundary and immutable byte associations; it is not a retained-response audit.

For the retained first outcome, select the read-only response audit:

python -m pytest tests/physics/test_sine_class_readout_evidence.py -q

Its module fixture reads the canonical response ZIP once, verifies the unchanged outer and uncompressed bytes, and checks the pinned source/protocol/attempt association. It reconstructs every retained Taylor increment and endpoint, full-state handoffs, exact events, raw endpoint and shared-prefix mixed bounds, and all five declared predicates. Guards prevent producer, field, jet, Picard and subprocess execution. Retained derivative enclosures and their generation remain numerical premises; arithmetic reconstruction is not independent derivative validation. The source cover does not erase the acquired family's correlation requirements.

For a physical-source admission, test deductions from the declared source law separately from manufacturer specifications and measured responses. The digital-PLL audit uses test_dpll_physical_admission.py to derive the detector from exact XOR overlap, construct distinct delayed histories with identical current state, and check the scoped current-state and phase-map obstructions. These algebraic controls neither establish experimental reachability of their supplied history family nor calibrate an instrument. They require no response download, fitting or frozen replay.

python -m pytest tests/research/test_dpll_physical_admission.py -q

Shared admission changes can cross the routine/research selection boundary. The default gate includes native relational execution through tests/test_*.py and the listed conservation-diagnostic tests. It does not select test_sine_admission.py, the tests/physics/test_relational_sine_*.py family, or the regional-transfer research modules. For a change to the common stored-coefficient boundary, exercise both native execution and its sine adapter explicitly:

python -m pytest tests/test_relational_exchange_execution.py tests/test_relational_regular_execution.py tests/physics/test_sine_admission.py -q

Then select the affected consumers from the theory-to-execution map. For a chained report, cover the producer's primitive admission and the downstream calculation: sampling smoothness to sample budgets, cycle modes to gains, visible evidence to hidden-state/capacity inference and prior forecasts, or captured states to regional currents and supplied increments. An SDK pass checks its adapters and projection; it does not replace those mathematical controls. A shared primitive-source change also needs the relative-pattern and forecast-endpoint association controls, including unchanged held capacities and partial validated horizons.

Changed contract	Required independent controls
Complete law, state and coordinates	Differentiate both full nodal rows; retain support, degree normalization, capacity, clock and inputs. Compare native Arg, baseline sine and alternative-mobility laws only under their actual premises. Test moving references, common origins, zero coordinates and same-observation/different-response counterexamples.
Primitive and numerical admission	Reject malformed support, law identifiers, authoritative aliases, nonfinite values, Boolean physical scalars, invalid radii and missing consumed coordinates before arithmetic. Rebuild consumed geometry and gradients; cached flags or displayed intervals cannot widen exact admission. Preserve each consumer's zero-capacity/loss and work-limit domain.
Algebra, derivatives and storage	Use independent edge sums, high-precision evaluations, exact matrix identities or analytic solutions. Check signed work and loss, all field directions, endpoint events and strict boundaries. Symbolic irrational targets, rational probes and rounded graph states provide different evidence.
Relative state, uncertainty and memory	Retain hidden initialization, every environmental node, original port degrees, actual observation times and each member's conserved means. Distinguish unknown common origins, independent nodal residuals and correlated relative coordinates. Inference compatibility is not existence, identifiability or a replayed response.
Forecasts and continuous enclosures	Exercise strict Picard inclusion, Taylor remainders, whole-time tubes, endpoint chains and held-parameter semantics against separate analytic fixtures. Retain partial-horizon status and actual validated time. Samples, Euler chords and held-pressure integration cannot replace a coupled continuous enclosure.
Geometry, equilibrium and recovery	Check exact circular reconstruction, full sine-current cancellation, weighted gaps and Hessian inertia with independent full-node equations. Geometric feasibility is not equilibrium. Local recovery, all-sector capture and conservative trapping have different initial sets and loss premises. Failed sufficient inequalities remain unavailable.
Formation, preparation and budgets	Separate identity absent initially from a supplied target. Keep the original form cost, phase uncertainty and complete law in positive and negative controls. Distinguish scalar storage, directional loss, transient barrier passage, maintained endpoints and accumulated port work; an instantaneous sign is not an integrated supply.
Symmetry and composition	Test whole-support/state/capacity equivariance, cycle orientation, relabeling and genuine quotient reconstruction. Ordinary reflection, combined sign/reflection and set-invariant uncertainty boxes have different consequences. Retain external attachments, relative zero modes and noncommuting weighted operators.
Collective observations and structural events	Check all competitors, mutuality, unresolved ties and state/support admission separately. Compare quotient and inherited rates with the full fine field. Hypothetical bridges, cuts, AL proposals and resets retain their work budgets and read-only scope; conservation or eligible contact does not prove occurrence.
Resonance, pulse and recurrence	Retain the selected port/readout, full tangent pencil and actual coefficient domain. Distinguish gain peaks from complex poles, exact periodicity from orbital stability, and family recurrence from a chosen-state verdict. Independent quadrature or static variation controls do not prove a nonlinear infinite-time response.
SDK, CLI and evidence projection	Check delegation, one-source capture, immutable observations, unavailable values, supported node labels and exact fraction export. Direct report schemas and the generic SDK envelope are separate contracts. Malformed reports and source changes must not authenticate themselves through serialization.

For local-response geometry inference, select the inverse owner and SDK projection through the theory-to-execution map. Use independent full-node edge sums and graph-distance cancellations to check the noncritical affine family and its remote-current remainder. Synthetic observation bands test monotone exclusion, calibration/noise propagation and unresolved arithmetic; they are not reserved responses. Keep the nominal-family parameter separate from the actual arc mean under phase residuals. An enclosing interval cannot serve as an existence or non-identifiability witness. Retained dipole and capture evidence can be audited without rerunning their frozen producers.

For successive-input geometry/gain inference, also select test_sine_two_pulse_inference.py. Check the complete carried state, the factored leading determinant, degenerate and unresolved rank, and all three reading-error coefficients. Positive factors can certify rank even when their determinant product rounds across zero; test factor-cancelled inversion and unresolved individual factors separately, including horizons below the product's resolution floor. A shared middle reading must retain its incidence coefficients; for this inverse's opposite-sign rows the marginal sensor bounds coincide with the conservative separate-increment bounds. Test exact large-offset cancellation and distinguish interval width from sensor error. Synthetic leading responses exercise the inverse but are not reserved complete-flow observations; necessary marginal intervals prove neither joint realizability nor exact full-state identifiability.

For the bounded observation-clock adapter, also select test_sine_clock_inference.py and the shared joint-inverse suite. Check constant-rate reduction, observation-time versus structural-time duration, maximum-duration error admission, effective gain J=G*rho, exact positive-prior projections and the shared three-reading error map. Include rational scale extremes and rejection at invalid clock/horizon domains. The nested auxiliary envelope must not be treated as an actual maximum-clock trajectory. Keep the exact all-row-rate/clock equivalence distinct from an ideal-source curvature control disproving exact gain/clock equivalence. Neither synthetic control is a reserved response or uniform noisy clock-identification theorem.

For four-reading finite-curvature inference, also select test_sine_curvature_inference.py with the clock and joint-inverse suites. Independently check the raw finite-difference coefficients, original-source and third-derivative errors, positive divisions, necessary clock/gain projection and unchanged actual-angle meaning. All four reading pairs require admission, including a malformed half-window value absent from the coarse child. Preserve one middle-reading history and exact offset cancellation; curvature is not a new independent sensor error. Test unavailable refinement with a retained coarse report, strict exclusions, touching constraints and informative as well as nonimproving outer bounds. Exact local information results for an ideal fixed-source subfamily do not imply noisy global identification.

For clock drift, select test_sine_clock_drift_inference.py with the curvature and clock suites. Check exact equal-exposure ambiguity, positive global rate and derivative-bound admission, endpoint matching, individual reading-transfer bounds and the unchanged sensor-error allowance. Include zero drift and singleton-prior reduction, all four primitive pairs, necessary mean-rate/gain projections and unavailable reference arithmetic. Use an independent varying-clock full-state control for finite transfer; this is an implementation check, not a reserved response or proof of instantaneous-profile recovery. Do not treat the widened reference readings as physical observations or a new independently sampled noise channel.

For finite-aperture observation, select test_sine_aperture_inference.py with the clock-drift and curvature suites. Reconstruct the fixed boxcar moments and all four virtual-reading rows independently; test constant, linear and quadratic exactness, cubic remainders and the separate post-event second-window error. Keep all four primitive averaged pairs, their original sensor-error map, clock transfer and numerical widths distinct. The curvature child must receive zero additional sensor error. Check source-gated bounds, zero-drift aperture error, necessary marginal projection and method abstention. Independent full-state integration of the averaging law is an implementation control, not a reserved response or a calibration of the sensor kernel. Point-sample exposure equivalence alone must not be reused as equality of interval averages.

For the response-free resolution budget, select test_sine_aperture_budget.py with the aperture inverse and SDK suites. Independently reconstruct moment row norms, separate sensor/enclosure/clock/reconstruction errors, strict source and quotient guards, and conditional width targets. Exercise exact boundary equality, large and subgrid rational inputs, and failed sufficient budgets without inventing unavailable width certificates. Rebuild the noise-overlap threshold from the two complete-history remainder bounds; check that only additive sensor error admits that existential witness. Numerical halfwidth alone must not trigger it, and an inadmissible whole-ball sufficient guard must not erase the valid zero-residual witness. Guard these controls against producer, inverse and frozen-worker execution. No selected readings or reserved responses are required for this theorem.

When changing the independent full-state response generator, select the direct source-box Taylor suite and two-port readout suite. The former checks independent linear, coupled oscillator and nonlinear solutions, uncertain-baseline cancellation, thirty-six and sixty-four coordinates, domain failures and the fixed Picard budget. The latter rebuilds both fast-clock sine rows and the actual support, retains all state coordinates and tests complete-flow enclosures on unrelated synthetic preparations. Its numerical cross-check is not a validated response or a reserved experiment. Changes to the common Picard or jet owners also need their existing comparison and retained-metric consumer suites.

For the validated finite-aperture producer, also select test_sine_aperture_readout.py. Check affine endpoint-rate positivity separately from the smooth extended field, both nodal clock-scaled rows, the observation-time integral row and complete 38-coordinate carry. Verify all four exact aperture widths, phase-only events and cumulative integral increments without resets or unrelated endpoint subtraction. Independent stationary, exact-integral and coupled numerical controls must not substitute the inverse's response approximation for the forward law. Exercise a later-window failure and retain its successful prefix, failed tube and absent complete result. These controls admit an implementation; they do not execute a reserved response or calibrate an averaging sensor.

python -m pytest tests/mathematics/test_validated_box_taylor.py tests/physics/test_sine_two_port_readout.py tests/physics/test_sine_two_port_inference.py tests/physics/test_sine_two_pulse_inference.py -q
python -m pytest tests/physics/test_sine_clock_inference.py tests/physics/test_sine_two_pulse_inference.py tests/sdk/test_relational_reports.py -q
python -m pytest tests/physics/test_sine_curvature_inference.py tests/physics/test_sine_clock_inference.py tests/physics/test_sine_two_pulse_inference.py -q
python -m pytest tests/physics/test_sine_clock_drift_inference.py tests/physics/test_sine_curvature_inference.py tests/physics/test_sine_clock_inference.py tests/sdk/test_relational_reports.py -q
python -m pytest tests/physics/test_sine_aperture_inference.py tests/physics/test_sine_clock_drift_inference.py tests/physics/test_sine_curvature_inference.py tests/sdk/test_relational_reports.py -q
python -m pytest tests/physics/test_sine_aperture_budget.py tests/physics/test_sine_aperture_inference.py tests/sdk/test_relational_reports.py -q
python -m pytest tests/physics/test_sine_aperture_readout.py tests/mathematics/test_validated_box_taylor.py tests/physics/test_sine_two_port_readout.py tests/sdk/test_relational_reports.py -q
python -m pytest tests/physics/test_sine_formed_evidence.py tests/physics/test_sine_curvature_evidence.py tests/physics/test_sine_clock_drift_evidence.py tests/physics/test_sine_aperture_evidence.py -q

The reserved inference protocol separates the full response producer from the inverse. Verify the actual inverse keyword allowlist: no hidden state, true angle, exact sensor gain, realized error or producer remainder may enter. Retain the two scalar readings and baseline correlation, independent calibration readings, numerical interval width and per-reading error as distinct evidence. Run retained-record audits after a first evaluation; tests must not regenerate that reserved response or silently retune a failed budget.

The reserved two-input protocol uses one held but uncalibrated gain per trajectory and three readings. The retained-record suite checks both complete thirty-six-coordinate Taylor certificates, exact endpoint/event carry, the one middle reading, public-only inverse packets, original angle/gain coverage, marginal widths and the fixed controls. Rebuild consumed arithmetic from saved primitives; cached success flags are not the evidence. These audits execute neither frozen producer nor inverse. Preserve a false optional whole_window_acute_certified flag without treating it as a failed inference criterion or a proved trajectory event.

The retained four-reading protocol and result add a passive half-time observation and one held unknown clock per case. Select the same shared Taylor/readout suites for the zero-jump continuation; check all three segments and both complete endpoint handoffs, distinguishing observed from structural segment times. The dedicated read-only audit checks the retained four-reading record and public ten-key packets without replaying its producer or inverse. Rebuild finite-curvature, source, coverage, width and fixed-control criteria from primitives, including available coarse children under false clock/gain priors. One half-time reading and its error must be reused consistently; a separate fitted curvature is not evidence. Preserve first failure, numerical abstention and optional acute flags.

The reserved nonconstant-clock result is audited by test_sine_clock_drift_evidence.py. Rebuild the primitive source, linear profile exposures, held sensor and all nine full-state certificates, then the recorded differences and necessary inverse bounds. Check the prospective sign/error margins, both endpoint handoffs per history, exact eleven-key packets, independent public transfer allowances, width/coverage and every fixed stopping predicate. The companion uses an exact equal-exposure association to the positive history; it is not a duplicate response. Reconstruct its derivative/rate admission and zero integral correction without executing an archived helper, producer or inverse. Saved derivative enclosures remain a premise, and hash/attempt receipts do not independently authenticate chronology or physical acquisition.

The retained finite-aperture result is checked by test_sine_aperture_evidence.py. Rebuild all twelve 38-coordinate step certificates, affine-clock provenance, full endpoint carry, passive cumulative integral and exact observed-width normalization. Reconstruct the raw averages, held sensor/errors, eleven-key public packets, moment/error transfer, nested inverse and every frozen stop. Check the signed-average/curvature budgets independently of the saved verdict. Only the 36 nodal coordinates share the exact first-window exposure equality; the rate and cumulative integral need not match the reference. No point-sample companion association transfers automatically to these averages. Namespace imports of shared audit helpers must not import their autouse fixtures into another module. Keep producer/inverse/worker guards and frozen-byte checks; stored derivative enclosures remain premises, with independently assembled first-rate consistency and whole-tube inclusion checks.

Boundaries that need explicit regression coverage

• Chained evidence is rebuilt from its premises. Change primitive inputs independently of cached fields, and check that downstream bounds are recomputed or unsupported declarations reject. Retain valid stored coefficients without renormalizing them. Test stale favorable flags, zeroed derivative bounds and inconsistent state/support associations with independent expected results. Invalid source declarations must fail before a solver runs or independent one-shot samples are consumed; a supplied signed endpoint increment must not be replaced by an evaluated form rate.
• Shared admission does not mean shared theorem domains. Specialized sine-cycle recovery retains its larger support domain (including the 51-node control); general phase geometry has separate work caps. The validated comparison and retained-metric Taylor owners keep their 24-coordinate limit, with boundary refusal and independent controls. Hidden-capacity sine forecasts use 2*n+1 coordinates. The separate direct source-box Taylor path admits at most 64 coordinates, without changing that comparison cap or introducing a Jacobian propagation. Test both policies and the full 36-coordinate two-port source; dimension admission alone certifies no response.
• Metric uncertainty remains correlated. Compare retained-ball propagation with independent analytic rotating, contracting and nonlinear flows, including boundary points outside coordinate axes. Check original-clock growth bounds, both directions, every domain margin and partial coverage. Include tiny local errors whose squares fall below the interval grid; normalized norm arithmetic must not create an artificial square-root precision floor. A coordinate projection is not the ball consumed by the next step, and a manufactured solver control is not a formation experiment.
• Exact and represented geometry remain distinct. A rounded pi, a small field residual or overlapping intervals cannot establish an exact antipodal state, equilibrium or symmetry. Test zero resultants, branch limits, unresolved denominators and strict equality boundaries explicitly. An admitted native proposal chord is not a continuous trajectory proof. Exact Fraction input to a graph may pass through binary64 capture; distinguish that source from an explicitly supplied rational preparation. An unrecognized symbolic cancellation with a zero-containing interval must remain unavailable, not become either equality or a counterexample.
• Preparation and endpoint sets retain their provenance. A tighter correlated source bound cannot replace a forecast's larger Cartesian endpoint enclosure. Partial forecasts keep their validated time and original requested-horizon status. Full-state capture must include remaining form storage, every sector face and the actual phase uncertainty; favorable phase geometry alone is insufficient.
• Reduction retains information and scale. Memory/second-order identities keep initial form, the weighted constant mode and moving history. Slow-phase controls retain the fast transient, memberwise references, original form remainder and preparation cost. Test zero horizon, tiny positive feedback, large growth and monotone exponential tails without silently clipping time. For collective contact observations, independently vary internal deviations at fixed means and verify the complete nodal response. Declared contact lifts can change a collective energy split while full storage is unchanged; test its signed remainder and exact local derivatives without treating them as a finite-horizon prediction. Exact-family membership, finite-width maintenance and acquisition into that neighborhood need separate controls. For reduced component ports, independently verify projection/lift identities, orbit multiplicities, the nonlinear bridge and its changed degree weights. Include odd initial errors and nonlinearly generated discarded modes; linear parity invariance is not an exact nonlinear quotient. Transfer the unchanged component law to the held-out receiver and compare total prediction error against the final outward full-response lower bound, retaining preparation and readout errors. Prove actual joined identity from the full law, not from the surrogate's stability or coordinate count. For network assembly, derive central mobility from every live contact degree and check exact charge/storage identities against the fine graph. Include a multiply connected port that rejects unchanged one-contact normalization, cycles with consistent component origins, and disconnected states whose instantaneous rows do not establish global capture. A uniform whole-state approximation bound must retain original preparation errors and generated odd modes; it is not a receiver-contrast or sensor budget. For all-time tracking, independently admit both invariant charts and the actual degree-weighted spectral gap. Preserve initial-error transients, generated odd modes and both conserved global mean floors; convergence to a target alone is not cross-flow contraction. Test separate form and phase resolution margins, including a valid envelope that resolves only one channel. Failure of a sufficient resolution bound is not trajectory error evidence, and a phase pass must not hide a failed natural form scale. A sharper form comparison must preserve operator order in its heat convolution, include the initial phase-to-form contribution and nonlinear bridge variation, and strictly admit both parity feedback gains. Exercise noncommuting operators and dyadic gain boundaries independently. Keep the previous partial artifact unchanged; a new method is separate evidence, not permission to enlarge the old allowance or replace its verdict. A two-port equilibrium admission must reconstruct every fine edge and cycle period, distinguish balanced circulation from nonzero nodal rates, and retain the matched-class zero-current control. Enclosed root existence and strict acute margins, not small residuals, justify the equilibrium; local recovery does not establish capture of a supplied initial family. For a storage handoff obstruction, reconstruct the complete boundary witness and its integer periods independently of the equilibrium root. Check the source lower bound across relative origins, the nonnegative full-form contribution and the radian phase-error allowance. Exercise the exact strict-margin boundary and reject Boolean or nonfinite radii. A lower-storage boundary witness excludes the scalar certificate, not capture by the full dynamics; current-source checks must leave frozen equilibrium evidence untouched. A finite two-port transit certificate must retain both initial form and phase errors, actual degrees and conserved means, every consumed clock factor and the full-state reconstruction. Verify its reference derivative independently on all eighteen nodes. Its acute-chart comparison needs a closed first-exit argument, not an assumed reference domain. Test actual endpoint changes against each member's own initial short gaps, with the initial error counted as well as the endpoint error. Whole-window acuteness and finite directional change do not establish eventual capture. Keep nonsymmetric error families and explicit unavailable margins in coverage. A longer capture gate must distinguish a validated nominal gradient reference from the full uncertain form-phase family. Check the folded metric and full reconstruction, whole-time tube admission, exact completed horizon and retained numerical radius. Every reference target, rate and chart prerequisite must be rebuilt from primitives. The analytic final interval continues the same full flow and needs its own original-form bound; a small phase distance alone is insufficient. Keep incomplete prefixes and failed capture margins unavailable, and preserve the first fixed-budget response even when its sufficient criteria fail. A supplied probe must retain the actual pre-event form and phase residuals, the exact jump, its storage work and the changed conserved-mean leaf. Reconstruct the receiver observation on each support: the common rule uses receiver degree masses twenty and eighteen, not identical coefficient vectors. Check the unjoined invariant directly, independent readout errors, complete-law finite response remainders and strict post-event trapping. An endpoint-ball assessor cannot certify the source's acquisition; test the separate finite source handoff without consuming cached verdicts. Keep a pure-heat transmission countermodel so the response is not reported as winding-specific evidence or selection of the sine constitutive law. An interior phase dipole needs the identical local coefficient vector in both models, with scalar observation errors distinguished from per-node errors. Verify its exact local sine response before bounding both full propagators; matching local adjacency does not prove finite response equality. Preserve the correlated target-angle contrast and every nonlinear remainder term. A degree-metric warmup must remove each connected component's own means, retain the combined control norm and prove strict original-form and phase budgets through the shared modified-energy kernels. Its starting source is already trapped; two norm caps alone cannot grant acquisition or trapping. Test missing warmup and target prerequisites so endpoint candidates never become delivered response, work or identity evidence. The phase-blind control needs its own complete rows and original-source heat bound: zero causal phase-pulse response does not erase raw background relaxation. Preserve both sine-law storage margins, signed jump work and all four scalar readout errors.
• All-time and symmetry verdicts have different scope. Fixed-budget consensus tests need an independent mixed-Lyapunov derivative from both full rows, zero-budget and failed-premise controls. Reflection excludes winding only at nonantipodal observations; it proves neither antipodal avoidance nor consensus. Equal-budget sources, tiny exact asymmetries and combined sign/reflection provide discriminating controls.
• Classification is not a numerical case survey. Exact factorized critical sets, full-support Hessian congruences and mode counts need independent witnesses and invalid-domain controls. Do not replace their completeness proofs by enumerating thousands of trajectories or transfer a positive-loss convergence result to zero loss or frozen capacities.

Current checks and retained evidence

Distinguish three kinds of validation when a research owner changes:

• Current-source contracts: exercise the shared engine, SDK or CLI, including numerical admission, unsupported domains, read-only behavior and atomicity.
• Mathematical controls: check independent identities, counterexamples or exact coefficient probes under their stated hypotheses. Rational probes do not replace the actual irrational constants in an ideal theorem.
• Retained evidence audits: validate frozen inputs, source fingerprints, saved intervals/checkpoints and original verdicts without rerunning a producer. A current-source regression is separate from authenticating historical evidence.

For example, the composition and memory owners identify their static controls and retained finite responses. The relational admission owner identifies recovery, formation and capture obligations. A successful current snapshot or a passing subset must not rewrite a stopped or negative historical verdict. Check the selected fixture before claiming that a run has no evolution.

The benchmark guide owns producer invocation, freezing and artifact lifecycle. Reuse the appropriate saved evidence or shared fixture; do not regenerate completed responses for unrelated changes. A new reserved prediction needs its own declared inputs and prospective protocol.

The formed C9 bundle audit checks the committed artifact hashes, archived source inventories, protocol associations and the distinct reduced-model stopping rules without calling an assessor or producer: the two-component error fraction uses its final outward receiver-gap lower endpoint, while network composition uses an absolute whole-window full-state allowance plus the exact degree/charge/storage control. Neither rule substitutes for its saved actual-family identity and supplied-work obligations. The all-time relaxation audit separately retains the phase and form resolution decisions, including any qualified partial outcome; it must not reinterpret a failed sufficient form margin as an observed tracking failure. The separate ordered-heat form certificate preserves that earlier report body as its baseline and applies the same primitive inputs and channel allowances. Its audit checks both new strict margins without relabeling the original phase_only result or its false joint stopping rule. The two-port compatibility bundle has a different static criterion: primary and matched-control equilibrium admission, full cycle/current consistency, strict acute and Hessian bounds, and nonzero versus zero interface current. It certifies neither a source trajectory nor capture after an attachment. The separate two-port handoff obstruction is an exact analytic result with current-source boundary-witness controls in the handoff suite, not another frozen experiment. Select it together with the compatibility suite when changing their shared owner. Its failed scalar admission does not revise the compatibility bundle or assert a failed trajectory. The two-port capture bundle retains all 4,096 successful nominal-reference steps separately from its analytic complete-state handoff. Its audit must preserve the original thirty-six-coordinate preparation, clocks, root prerequisites, strict tube margins, endpoint radius and final storage margin. Read the saved first response and archived source; do not rerun its producer to inspect the result or substitute the folded reference for the full family. The supplied two-port probe has a separate conditional endpoint suite, response and maintenance controls. Its source-evidence audit must match the original capture primitives and rebuild every consumed handoff bound. Artifact hashes and rational chain consistency associate the retained record with its declared source; they do not independently revalidate omitted Taylor calculations or authenticate execution. The probe protocol freezes its own assessment without replaying the capture producer or altering its saved response. Its original attempt failed during auxiliary control export; the retained response comes from a separately archived deterministic export-recovery recomputation with unchanged scientific inputs and runtime. Check the original archive, failure record, recovery wrapper and explicit evaluation history separately. A successful retained recovery must not overwrite the failure or be labeled as a successful first attempt. Select the source-handoff suite with the conditional endpoint suite when changing this acquisition-to-probe chain. Exercise altered primitive laws, lost metric correlations, malformed or incomplete evidence and cached verdict changes independently; a cached flag must neither grant nor remove an otherwise justified handoff. The research audit's direct schema and explicit retained-execution premise are separate from the generic SDK projection of the probe report. The interior-dipole suite adds finite warmup, exact local response, full-flow remainders, phase-blind background exclusion and separate work/identity admissions. Select it with the modified-energy kernel suite and the source-handoff suite for changes to that continuation. The initial capture evidence remains fixed; a later analytic warmup neither replays its validated reference nor authenticates its retained execution. The saved dipole response passes its thirteen conditions in the first frozen assessment. Audit its original protocol/source association, source handoffs, warmup, finite contrast, heat exclusion and both work/identity channels separately from the earlier form-probe export recovery.

python -m pytest tests/physics/test_sine_formed_evidence.py -q

Missing committed artifacts fail this check. Historical source need not match the current implementation; current-source regression remains separate. The original pair/probe records still lack an evaluation-time source snapshot. Content consistency neither authenticates chronology nor proves the mathematics.

Optional retained-record audits skip explicitly when local evidence is absent; they must not recreate a producer or count missing evidence as a passed response. Synthetic intervals, step records and stubbed producers test consumer logic, not historical execution. Preserve original source archives, failures and inconclusive verdicts; corrections require separately identified evidence. Forecast readers recompute their consumed admission and Picard inclusion, but do not replay the full Taylor proof or authenticate record production. Test that boundary explicitly: partial coverage remains partial, and failure to certify a requested duration or margin is not an exclusion of actual identity. Retrospective channel integration from saved whole-time tubes must preserve the original verdict and receive its own analysis provenance.

Local execution is serial by default. To use the main CI scheduling policy, run python -m pytest -n 2 --dist loadfile; this keeps the same not slow selection and assertions. Each file stays on one worker to reuse its module fixtures. Workers isolate Python state and prepare session fixtures independently. Pytest-benchmark does not collect timing measurements under xdist; performance measurements need a separate serial run.

Choose an affected directory, module or test for bounded validation:

python -m pytest tests/core_physics -q
python -m pytest tests/operators/test_u3_hard_invariant.py -q
python -m pytest tests/sdk -q
python -m pytest tests/cli -q
python -m pytest tests/sdk --collect-only -q

To include all retained slow tests, use python -m pytest tests -o addopts="". To select only marked slow tests, use python -m pytest -m slow with those paths. Inspect python -m pytest --markers and collection before treating a marker-selected run as coverage: a registered marker can select no tests.

Standalone scripts do not inherit pytest's source-path configuration. Use the editable installation above or explicitly set the shell's PYTHONPATH to src so a previously installed release cannot replace the working source.

Organization

tests/ contains top-level modules and subject directories. Search for the affected API rather than maintaining another inventory of individual tests. Core nodal behavior is in core_physics/, operators in operators/, specialized certificates in physics/ and public network usage in sdk/. CLI integration checks the same study recipe through Python and the module entry point, including malformed input, diagnostic availability, output replacement and logging isolation. Catalog checks do not execute a study. conftest.py and utils.py own shared helpers. The core scope map identifies the engine tests that replace retired self-contained illustrations.

Consolidation removes work, not just collected test identifiers. Test each invalid boundary at its shared owner, then retain distinct consumer-wiring, rollback and representation cases instead of repeating the entire Cartesian product at every wrapper. Reuse an expensive producer only when its results are read without mutation; keep independent numerical oracles and negative controls. Do not turn removed parameter grids into hidden loops or replace meaningful assertions with finiteness checks. Operator labels alone do not justify energy-sign or conservation assertions with arbitrary tolerances.

The separately packaged arithmetic applications have their own test paths:

python -m pytest applications/factorization-lab/tests -q
python -m pytest applications/factorization-lab/benchmarks/test_benchmark_suite.py -q
python -m pytest applications/primality-test/tests -q

Run these as separate invocations: their local import roots differ from the core suite. The root default selection does not include them. See the corresponding factorization guide and primality guide for application setup and scope.

Research producers under benchmarks/ have declared entry points and provenance requirements; a default pytest run does not implicitly cover them. Do not regenerate retained evidence for an unrelated change.

For examples with run_protocol and build_report, reuse the real report through a module fixture for numerical and scope checks. The shared example helper checks main separately with a sentinel protocol and synthetic report, without rerunning a scientific producer.

Reusable independent oracles belong in test support modules rather than in another collected test suite. The pair oracle helper shares fine support, exact jets and high-precision reductions without importing production certificates or pytest fixtures. Each consuming suite retains its own complete-law assumptions, preparation budgets and module-scoped reports. When optimizing a shared certificate kernel, compare complete exact projections on representative available and unavailable cases as well as these independent controls; timing differences alone cannot establish unchanged evidence.

Runtime facade tests own the two cold import orders for the P2/REMESH example families; those checks still use fresh processes and resolve the actual public APIs.

Cold subprocesses that test this checkout explicitly request the source_tree_environment fixture from conftest.py and pass it as subprocess.run(..., env=source_tree_environment). It copies the process environment and prepends pytest's configured source paths to PYTHONPATH, without changing the parent's environment. Pytest's pythonpath setting only updates its own interpreter; a fresh Python process can otherwise import a different installed package. Keep the fixture opt-in: tests of installed-package discovery or intentional import isolation retain their declared environments. Working directories, timeouts and optimized-mode flags remain test-specific.

The Makefile target make test uses the routine engine selection; make test-all includes the retained research suites, still excluding slow. make dev-test adds coverage over src, including compatibility shims. Research producers use the individually documented entry points in the benchmark guide, not a directory-wide pytest campaign. make validate checks importability, references, documentation integrity and SDK tests.

Structural regression evidence

Assert the contract of the changed path. Record representation, graph/support, coefficients, primitive triad, input history, step size and operator sequence as relevant. An operator jump and a continuous solver step need different expected results. For a full grammar word, verify initiation, closure and contextual admission; distinguish fragments explicitly.

Do not assume all operators increase coherence or that SHA freezes every later EPI update. Its capacity attenuation, pure-EPI diffusion and a multichannel trajectory have different scopes. Inspect the operator contracts and grammar scope before asserting an invariant. Initialization may set fixture state directly; test subsequent evolution through the API under review.

When stochastic behavior changes, control the RNG and record its seed, initial state and execution order. A seed alone does not fix timestamps, external data or every backend. When relevant, compare available tetrad fields with estimator provenance and unavailable-field reasons. Diagnostic scores are not acceptance thresholds unless that specific policy is being tested.

Backends and optional dependencies

tests/conftest.py accepts --math-backend and TNFR_TEST_MATH_BACKEND. The command-line option takes precedence; the selected value sets TNFR_MATH_BACKEND and clears the backend cache before collection.

python -m pytest tests/mathematics/test_backends.py --math-backend=numpy -q -rs

Install compute-jax or compute-torch before requesting those optional adapters. Explicit per-test backend requests are not replaced by the session setting. Review skip reasons with -rs and report adapters actually exercised. NumPy is required by the shared conftest; a missing installation is not evidence of successful NumPy-free execution.

Code quality and documentation checks

Install tools needed for the check; options are configured in pyproject.toml:

python -m pip install -e ".[dev-minimal,test-quality,typecheck]"
python -m black --check src/tnfr
python -m flake8 src
python -m pydocstyle src/tnfr
python -m mypy src/tnfr
python -m pyright src/tnfr

Pass affected files for a focused edit where supported. Black and isort use 88 columns; pydocstyle uses NumPy conventions. CI's advisory checks are listed in the workflow guide. make format modifies files; it is not a read-only validation step.

The lazy physics namespace has one maintained export map and a generated typing facade. After changing that map, run python scripts/generate_physics_stub.py --write; validate it with python scripts/generate_physics_stub.py --check. The generator reads the source AST without importing TNFR, and the facade tests check typed exports, runtime object identity and cold import boundaries.

For shared relational report export, select tests/sdk/test_relational_reports.py. Preserve every registered owner-managed type, actual-class and subclass admission, native label validation, exact fractions and unavailable fields. Cold-process controls must show that export does not import unrelated research owners. Use guarded zero-horizon reports or synthetic wiring for producer-backed types; an SDK export test must not regenerate a selected scientific response.

Optional pre-commit setup requires installing pre-commit separately before pre-commit install. .pre-commit-config.yaml configures Black, isort, pydocstyle and a local Bandit-command guard. That guard uses Bash, which must be available on Windows. The hooks do not perform a code review; CI's formatting gate checks all tracked files selected by those hooks.

For Markdown-only changes, use the relevant reference check. For site or documentation-infrastructure changes, also run integrity and build checks:

python scripts/verify_internal_references.py --ci
python scripts/check_documentation.py
python -m pip install -e ".[docs]"
python scripts/prepare_docs.py
python -m mkdocs build --strict

The reference checker accepts --dirs followed by files or directories to bound the check. The staging script copies repository owners into build/docs-source/; MkDocs renders that generated source into site/. Edit the repository owners, never either generated tree. The documentation integrity check also validates the theory catalog, its generated navigation, operator contracts, grammar roles and glossary cards/index. Regenerate views with python scripts/check_documentation.py --write-generated after changing their source declarations, then rerun the read-only check. It executes the Python block in the README's Quick start section and compares its captured output with that section's declared output. Publication metadata uses the same strict JSON reader as the engine, including duplicate-key rejection.

Security checks

Follow SECURITY.md for reporting and trust boundaries:

python -m pip install -e ".[security]"
python -m pip_audit
python -m bandit -r src -c bandit.yaml

The dependency audit covers installed packages, including extras actually installed; it does not automatically cover every optional dependency. Bandit's configured exception is recorded in bandit.yaml. A clean report does not guarantee that dependencies or code contain no vulnerabilities.

Validation workflow and reporting

1. Identify the affected contract and existing owners. Reproduce a suspected bug before the fix when feasible; distinguish a new counterexample from a measured baseline.
2. Add meaningful regression coverage when behavior or an important boundary changes. A small documentation correction need not add numerical tests.
3. Run affected checks. Broaden to dependents, optional backends or the default suite when changed scope or a failure justifies it. Do not rerun unrelated expensive research producers as a routine checklist item.
4. Report actual commands, interpreter and relevant dependency versions, pass/fail/skip counts, warnings and untested scope. Do not label a failure pre-existing without baseline evidence.

The structural_rng fixture supplies NumPy's generator with seed zero. structural_tolerances supplies atol=1e-12 and rtol=1e-10; these are not exact-theorem eligibility gates. Choose scale-appropriate numerical tolerances and keep them distinct from represented exact equality. The autouse cleanup resets selected global state; restore additional state your test changes.

Coverage is feedback, not proof of a physical invariant. With test dependencies installed, a selected run can generate a report:

python -m pytest tests/sdk --cov=tnfr --cov-report=term-missing

The current project configuration does not enforce a coverage percentage. The Python 3.11 CI job instead uses --cov=src to retain the whole source-directory scope, with -n 2 --dist loadfile. Pytest-cov combines the workers' data and produces terminal and XML reports. Do not wrap distributed pytest with coverage run: that alone measures the controller instead of combining worker execution. Coverage of Python subprocesses launched inside tests is a separate configuration and is not enabled here.

Retained C6 campaign reproduction

The union/history report tests that reconstruct the retained B54/B55 lineage are marked slow, including tests whose shared fixtures perform that work. During the 2026-09-19 release check, one union-report fixture consumed more than 20 minutes of CPU time. This observed cost is not a runtime ceiling. Lightweight lineage and output-overwrite rejection tests remain available when their research paths are selected; the finite-state union oracles run independently of the retained campaign. C6 is outside the routine engine gate.

Select the expensive report checks explicitly:

python -m pytest -m slow tests/physics/test_c6_winding_union_report.py tests/physics/test_c6_winding_history_report.py -q

For the independent small-state controls:

python -m pytest tests/physics/test_c6_carried_return_unions.py -q

This classification changes selection by cost, not assertions or scientific scope. Report the expensive campaign as unrun when omitted, and retain failures from earlier attempts. A passing reduced suite does not certify the historical report or close C6 global stability.