- Use fabs() for accumulated_rotation threshold (both directions)
- Reset body_index in reset_orbit_tracker()
- Remove unused compare_double() and compare_vec3()
- Rename min_time_days to min_time_seconds (eliminate conversion)
- Remove dead SECONDS_PER_DAY constant
- Add buffer overflow guards in dump_simulation_state()
- Use static const instead of #define for MIN_DISTANCE_CLAMP
- Refactor old_tests/test_inclined_orbits.cpp into 4 SCENARIOs
- Replace broken OrbitTracker period test with propagation-to-apogee check
- Rewrite TOML config to TOML 1.0 inline table syntax
- Add precalc script using sim_engine.py
- Fix orbit tracker: use fabs() for accumulated_rotation threshold
- New test_energy.cpp: energy conservation (10-day drift < 1e-12) and
zero-inclination prograde direction verification
- Tighten energy assertion: replaced qualitative < 5% with WithinAbs(0.0, 1e-12)
(actual drift ~2e-16 for 10-day simulation)
- Remove direction test from test_orbital_period.cpp (semantic separation)
- All 16 assertions across 3 test files passing
- Fix section numbering (was 1,2,3,5,4, now sequential)
- Remove stale dump_simulation_state NULL check TODO (already fixed)
- Retain OrbitalMetrics::angular_position 2D-only TODO for future session
- sim_engine.py: generic orbital mechanics simulator with RK4 + analytical
propagation, used to verify C++ test expected values
- test_orbital_period.py: precalculation script with safety timeout for
measuring Earth/Mars orbital periods and direction test values
- Remove SOI transition logic (buggy in both Python and C++ versions)
- Use dataclasses.replace() for immutable state updates
- Support single-line TOML inline tables via tomllib
- Fix orbit period detection: track accumulated signed rotation instead of
wrapped delta, completing on true 2π revolution (was < 0.05 rad)
- Rename quadrant_transitions → wrap_count (accurate semantics)
- Add accumulated_rotation field to OrbitTracker struct
- Consolidate create_orbit_tracker into create_orbit_tracker_with_min_time
- Extract reset_tracker_fields() helper to avoid duplication
- Use vec3_dot() for kinetic energy, vec3_magnitude() for speed calculations
- Define MIN_DISTANCE_CLAMP and SECONDS_PER_DAY as file-level constants
- Add NULL checks in dump_simulation_state for spacecraft/maneuver pointers
- Output summaries to tmp/summaries/ instead of alongside source files
- Add per-module timestamp caching (skip up-to-date modules)
- Add combine_summaries() to merge all into all_summaries.md
- Extract LLM provider and model to config variables
- Reorganize script: config, discovery, utilities, worker, main loop
- Remove TODO from simulation.cpp (interpolated burns now implemented)
- Update technical_reference.md: both trigger types now use sub-step
propagation to exact trigger time, not step boundary
- Update hohmann-rendezvous-quantization-fix.md: mark Option A as
resolved, remove outdated quantization fix suggestions
- Update propagation_refactor.md: resolve Issues 3 and 4, replace
interpolated time triggers TODO with code path unification TODO
- Remove make run target from Makefile (agents should not run graphical sim)
- Update AGENTS.md: clarify make test vs make test-build, add Simulation section
- Update technical_reference.md Build System: add tag/name/section filtering
- Add SCENARIO section docs with --section flag and fixture behavior notes
- Use ./build/orbit_test path, rename config_name to tag_name throughout
For e < 1e-10, compute argument of latitude instead of hardcoding
true_anomaly = 0.0. This preserves position consistency after circularization
burns.
- Move ascending node vector computation earlier so it's available for
the circular orbit case
- For well-defined ascending nodes (sin_i > 1e-6): compute argument of
latitude using the ascending node direction and h×n perpendicular
- For nearly-coplanar orbits (sin_i <= 1e-6): use atan2(y, x) directly
to avoid numerically unstable ascending node direction
Replace manual buffer management (malloc/free/snprintf) with a simple
stack-allocated TestOutput helper. Uses dump_simulation_state's 4-param
API with a single dump_state() method. Verification stays as individual
INFO() calls outside the loop.
Planning doc updates:
- Replace estimate table with actual measured values from DT sweep test
- Add 0.5s and 2.0s data points (69m and 228m respectively)
- Mark DT sweep tests as COMPLETED section
- Add analysis note about linear scaling with DT
New tests in test_maneuver_planning.cpp:
1. DT sweep: Hohmann transfer separation vs time step
- Runs full Hohmann transfer at DT = {0.1, 0.5, 1.0, 2.0, 5.0, 10.0}
- Verifies final separation matches expected ranges
- Confirms ecc < 0.01 for DT <= 1.0
2. DT sweep: quantization error is bounded by DT
- Tests 9 trigger offsets per DT value at DT = {1, 5, 10, 30}
- Verifies quantization error is always in [0, DT)
- Verifies execution time matches ceil-to-step-boundary
3. DT sweep: Hohmann arrival burn timing error
- Measures exact timing error at each DT
- Reports position error estimate (timing_error × velocity)
- Confirms timing error < DT
All 157 tests pass (240,742 assertions).
Previous: 154 tests, 240,445 assertions.
Mark Issues 1, 2, 5, 6 as RESOLVED. Mark Issue 3 as PARTIALLY RESOLVED.
Mark Issue 4 as UNRESOLVED (time-triggered quantization).
Update call chain diagram to show current merged structure.
Update call sites summary to show final 4 call sites (3 contexts).
Add remaining work section for interpolated time triggers and
parabolic/hyperbolic orbit support.
