README.md

Tests

213 tests, about five minutes (pytest). Layout:

File	Tests	Covers
unit/test_atmosphere.py	7	T-01; ISA table; offset; range rejection; CasADi/numeric agreement; Hypothesis positivity
unit/test_units.py	17	unit conversions, unknown units, T-13 (converted exactly once)
unit/test_config.py	16	register loading and validation, T-14, evidence counts, hashing
unit/test_physics.py	10	T-02, T-03, T-07, T-08, T-15, monotonicity properties, 60 s speed check
unit/test_dynamics.py	7	T-04, T-05, T-06, T-09, T-10, events, determinism
unit/test_analyses.py	15	analyses A and D, including the scenario-mass regression for D
unit/test_energy_descent.py	15	analyses E and F, steady-flight helpers, hand-calculated budget
unit/test_d_addenda.py	11	D's lumped rotor lag and the D_a_rec_curve table
unit/test_inputs.py	18	the config loader, load_inputs problems, one command end to end with the cache
unit/test_ft.py	7	FT configuration, strict settings, the z_loop split, the FT task
unit/test_ft_model.py	12	the forward controller, the TECS-lite port, the FT flags and recovery
unit/test_defaults.py	9	the ArduPilot-defaults register, resolve_default, the third case in C and FT
unit/test_view.py	15	the results viewer and workbook: tolerance, legacy folders, data-src, no network, --view
unit/test_m5.py	13	T-12 (analytic), search, ensembles, baseline controller, analysis C (including C taking Q_TRANS_DECEL from analysis D)
unit/test_m6.py	9	analysis B1 including T-11
unit/test_m7.py	23	sensitivity, run records, results writer, release gate (including the real-problem T-12 rule), the CI check, report
unit/test_register_drift.py	8	the YAML register against the physics defaults; missing register files
unit/physics_reference.py	—	the NumPy reference implementation used only by tests (T-15, T-07)

Verification IDs (proposal section 12)

The release gate reads tests by name: a test whose name starts test_tNN_ counts for T-NN.

ID	Test	Where
T-01	atmosphere at 0 and 5,000 m	test_atmosphere.py
T-02	wing limits and continuity	test_physics.py
T-03	hover power with figure of merit 1	test_physics.py
T-04	constant-deceleration limit	test_dynamics.py
T-05	coast limit	test_dynamics.py
T-06	energy conservation	test_dynamics.py
T-07	battery closed form against the iterative solve	test_physics.py
T-08	motor steady state	test_physics.py
T-09	step-size convergence against the adaptive reference	test_dynamics.py
T-10	Cartesian against flight-path-angle dynamics	test_dynamics.py
T-11	collocation replayed in the simulator	test_m6.py
T-12	optimiser repeatability from three seeds	test_m5.py tests the search machinery on an analytic cost only. The real-problem T-12 is not a unit test (a three-seed optimisation takes minutes); it is enforced by the release gate through search_spreads in the run record (test_m7.py). See analysis/README.md for the measured spreads
T-13	unit boundary	test_units.py
T-14	TBD makes a run DRAFT; report writing refused	test_config.py (and test_m7.py for the writer)
T-15	NumPy reference against the CasADi model	test_physics.py

Notes

• Tests are not under mypy (files = ["src"]), and ruff ignores the naming rules for physics symbols in tests/ (T_r, Vc, …) as in src/.
• Slow tests are the collocation ones (test_m6.py, about 70 s in total; the solves are shared through module-scoped fixtures) and the sensitivity runs (test_m7.py, about 30 s).
• Not in the suite: the benchmarks/ and regression/ folders of the proposed layout; the real-problem T-12 runs; a Sobol run on the real model.
• Property-based tests use Hypothesis (density positivity and monotonicity, thrust and drag properties, bus-voltage monotonicity, unit-conversion identities).