Repository navigation
QuadCraft v0.5.0 — Time-owned simulation and conservative gameplay geometry
v0.5.0 — 2026-11-07
Simulation and gameplay geometry
- Give shared continuous loops fixed seconds, bounded catch-up and deterministic seeded randomness. Remove Asteroids' second wall clock and retain timer remainder across partitions.
- Resolve Asteroids' moving-target swept contacts chronologically in the physical metric, with uncapped periodic-image enumeration, lifetime boundaries and fragment remainder.
- Share ordered raw-grid obstruction traversal between Doom line of sight and hitscan; intersect physical enemy spheres and let walls win ties. Settle elapsed time before input mutation, own held mouse state explicitly and rebase clocks on respawn.
- Repair SimAnt's reversible raw-axis excavation and distinct nest occupancy. Workers/soldiers use obstacle-aware private delivery routes, replan stale paths and throttle unreachable retries.
- Add generic conservative undirected graph transport, reused by SimAnt pheromones with separate source, evaporation, cap and Float32-rounding accounting. The raw eight-neighbor terrain graph is not relabeled as the twelve-neighbor FCC graph.
- Pick Minecraft/Mahjong by actual projected footprints and reverse stable paint order, preserving raw cells and original tile identity.
Verification and documentation
- Replace shallow reporting/assertions with deterministic consumer-visible collision, time partition, routing, transport and picking regressions; keep the shared scaffold's seconds contract aligned.
- Extend the real all-game Chromium gate with late-input/respawn timing and held automatic fire when native capture is available. Record real denial and retain keyboard controls otherwise.
- Give the local stdlib HTTP server an OS-sized accept backlog for concurrent classic-script requests; keep resource failures fatal, without retrying or masking them.
- Document shared loop/field/RNG APIs, coordinate gauges/units and per-game deployment boundaries. Bump native and directory-local browser package metadata together.
Verification: 2,341 CLI checks, zero failures; all 30 Chromium scenarios passed locally. Native capture was denied on this macOS run, so its capture-only automatic-fire branch requires the native-capable CI run. Interactive native OpenGL gameplay was not exercised locally.
Coordinated release and reproducibility
| Project | Release | Exact main revision |
|---|---|---|
| QuadMath | 0.6.0 | 1e960220e2616703fc468ab3f857b21114c961be |
| symergetics | 1.3.0 | 4ab5807a37664158f11e7383f76fdd8ac6672c3e |
| QuadCraft | v0.5.0 | c0f3dc4206086e6a5167ccaaec0e28ece70b7932 |
The source-pinned Linux suite run passed all 14 selected full-profile checks, with no unexecuted checks and clean source/worktrees. The same exact-revision command passed locally; existing generated outputs were not included in source commits. Attached suite-verification.json and suite-verification-local.json retain actual commands, return codes, versions and revisions. Machine-specific artifact locations are replaced by logical aliases for publication.
- QuadMath source-specific verification: Python 3.11/3.13 tests at 100% statement/branch coverage, installed public-method consumers, Lean build/leanchecker, regenerated figures and the successful 123-page canonical manuscript.
- symergetics source-specific verification: both Python versions passed scientific coverage, standalone demos and minimally installed wheel consumers. The additional patched local scientific run passed 1,467 tests, 12 skipped, 91.07% coverage; current Matplotlib 3.11.2 comparative/dashboard consumers rendered outside checkouts.
- QuadCraft main Chromium run: all 30 games passed, with native Doom capture active and actual held-fire/late-input/respawn behavior exercised. Local macOS capture was genuinely denied; both outcomes are recorded, not mocked. The portfolio CLI reported 2,341 passing checks. Native Linux graphical compilation and Debug/Release headless checks passed; protected-main PR #6 merged normally without force/admin bypass.
- Five real implementations checked 279 coordinate cases each, 34 exact volumes, 216 FCC site comparisons, 37 rational vertex means, two fixed-width rational-ingress cases, 280 exact-metric pairs and 15 conservative Python/JavaScript graph steps. Maximum observed mass drift was
1.4210854715202004e-14.
Boundaries and migration notes
- Ordinary-mass conservative transport is distinct from symmetric-normalized heat. JavaScript uses
Number/Float64 and SimAnt stores Float32; this is not arbitrary-precision conservation. - Vertex means and uniform-solid mass centroids are different. Closed-surface moments require non-self-intersecting, mutually nonintersecting surfaces; closure alone does not certify embeddedness.
- Lean proves the documented integer quotient, metric/volume bridge and bounds, not floating simulation, gameplay, universal shell counts or BFS equivalence.
- QuadMath figure manifests now resolve relative to the configured output root; the symbolic CSV column is
V_ace_ivm, not the obsolete integer-only label. - Shared continuous-loop callbacks use seconds; intervals/browser timestamps remain milliseconds. Raw eight-neighbor SimAnt occupancy is not the physical twelve-neighbor FCC graph.
- The mega abstract accepts its declared arguments only; previously ignored extra keywords now raise. Irrational constants remain approximations.
- Interactive native OpenGL gameplay was not exercised. The local official-font PDF path was unavailable; the attached manuscript was built by the successful source-specific Linux workflow.
Assets include source-specific distributions where applicable, portable verification receipts, observed runtime evidence and SHA256SUMS. See the suite contract for units, inverses and precision boundaries.
Browser assets contain all 30 actual main-source outcomes plus initial/final screenshots.