Next Gen STEP AP242 Native-ish BREP Geometric Modeling Kernel
M8/Profile-M2 validation report
Real-path results
- .NET SDK 10 detected; restore was satisfied during the clean solution build.
dotnet build Aetheris.slnx --no-incremental: pass, zero warnings, zero errors.dotnet test Aetheris.SheetMetal.Tests --no-restore: pass, 60/60 in the full run; new Profile-M2 tests also pass after rebuild.- Full solution test execution completed. All feature suites passed; Kernel.Core exposed timing-test nondeterminism under parallel load. The two failures from the parallel run passed immediately in isolation; a later full Kernel.Core run exposed a different timing assertion, which also passed in isolation.
Aetheris.FrictionLab.Testsreports no discoverable tests, as before. - Final Firmament parses and compiles with the source STEP absent: pass.
- Concept contract, Concept Struct resolution, Concept Paths, semantic constraints, Profiles, formed lowering, flattening, STEP export/reimport, DFM, and CLI inspect/compare: exercised by tests and CLI dogfood.
- Non-CTC semantic-panel fixture: pass with deterministic stable paths.
git diff --check: pass (line-ending notices only; no whitespace errors).
CLI inspection
Final authored recognition is Complete: 15 regions, seven bends, 17 cuts, nine patterns, three datums, one tab, two stepped-notch edge fragments, and 18 resolved constraints. Exact flat status is Valid; DFM is Pass after the mounting-flange profiles restored the real hole-to-edge clearance.
One warm observational run on this machine measured:
| Phase | Time |
|---|---|
| Parse total | 26.5759 ms |
| Semantic + constraint resolution | 16.0116 ms |
| Formed/profile lower | 70.3482 ms |
| Authored flat lower | 57.5023 ms |
These are diagnostic timings, not benchmarks. Constraint resolution is included in semanticResolve; exact profile construction is included in formedLower and is not separately instrumented.
Determinism
Repeated generation produced stable paths and these SHA-256 values:
| Artifact | SHA-256 |
|---|---|
ctc03-final.firmament |
3A8C4A7C93C4750FE9241B4F4071F240108966F5AB0A28F1E81ABF7ABF16CA91 |
ctc03-formed.step |
8E049EC25E0B3F3C57A1F931C4D111C51BD9D840B3B53569A7771F764567F330 |
ctc03-flat.step |
8104D7C13DF87F838DC709171E7BEABCAECFDD3A62943B6D254B68E5DC11B9B5 |
ctc03-flat.svg |
24C95460BF669BA38A78CDFF9BE6E7E1051EE56F39C8E0018249B5BAE00C5B05 |
| Flat IR deterministic hash | e3ec913e24f65d220839bbdbfa2d909dd8bdeeaed87ecc9ccfc6c1be7a1d297c |
Code quality
Verdict: acceptable bounded prototype. Generic edge-composition IR, diagnostics, stable paths, non-CTC fixture, and real export/reimport tests are cleanly localized. Debt remains: the Profile adapter is regex/block based, Sheet Metal exposes only the repeated fragment pressure needed so far, semantic and constraint timing share one bucket, and cross-edge corner ownership is not implemented. There is no CTC-specific compiler branch or generator.