Skip to content

Next Gen STEP AP242 Native-ish BREP Geometric Modeling Kernel

docs: resolve INV-B0 blind-hole enclosed-void diagnosis

Yuechen Li445400757381
diff --git a/docs/frictionlab/FRIC-B0-chimera-pressure-test.md b/docs/frictionlab/FRIC-B0-chimera-pressure-test.md
index 207f30c9..a2f498fb 100644
--- a/docs/frictionlab/FRIC-B0-chimera-pressure-test.md
+++ b/docs/frictionlab/FRIC-B0-chimera-pressure-test.md
@@ -127,6 +127,8 @@ Zones:
- Signature: `FIRM-SCHEMA-0006` fully enclosed internal voids disallowed in process `default`.
- Effect: catches certain blind-hole outcomes before/alongside geometric family checks.
+INV-B0 follow-up (2026-04-19): `zone_b_blind_cyl` is fixture-contained (no exterior-face placement), so this rejection is expected enclosed-cavity policy behavior rather than a blind-hole misclassification.
+
## Required boundary classification synthesis
### Safe to generalize now
@@ -163,4 +165,3 @@ None. No architecture-broadening patches were applied.
- Do **not** attempt broad general CSG or arbitrary multi-family booleans.
- Do **not** remove tangent/grazing rejection boundaries.
- Do **not** expand non-box-root mixed prismatic support before a bounded representation and explicit reconstruction lane exist.
-
diff --git a/docs/frictionlab/INV-B0-blind-hole-enclosed-void-diagnosis.md b/docs/frictionlab/INV-B0-blind-hole-enclosed-void-diagnosis.md
new file mode 100644
index 00000000..8d7219f2
--- /dev/null
+++ b/docs/frictionlab/INV-B0-blind-hole-enclosed-void-diagnosis.md
@@ -0,0 +1,81 @@
+# INV-B0 — suspicious blind-hole → enclosed-void diagnosis
+
+Date: 2026-04-19 (UTC)
+
+## Outcome
+
+**Outcome A — resolved diagnosis.**
+
+The FRIC-B0 suspicious `zone_b_blind_cyl` result is a **fixture/setup bug**, not a kernel blind-hole subtract bug and not an enclosed-void checker bug.
+
+## 1) Exact suspicious case
+
+- Fixture: `testdata/firmament/frictionlab/fric-b0/chimera_a_zone_b_blind_cyl.firmament`
+- Operation: `subtract(box, cylinder)`
+- Diagnostic observed under process `default`:
+ - `[FIRM-SCHEMA-0006] ... fully enclosed internal void(s) ... not allowed for process 'default'`
+
+## 2) Fixture geometry intent vs actual geometry
+
+The fixture labels the operation as `zone_b_blind_cyl`, but it does **not** place the cylinder on an exterior face.
+
+Source values:
+
+- root box: `size[3]: [60, 40, 20]`
+- subtract tool: cylinder `radius: 4`, `height: 8`
+- no `place` block on the subtract op
+
+Execution path facts:
+
+- Booleans with `placement == null` resolve to zero translation.
+- In that case, subtract executes with the raw primitive tool body.
+- Core primitive constructors (`CreateBox`, `CreateCylinder`) are centered at the origin.
+
+Therefore, this fixture produces:
+
+- box extents: `x ±30`, `y ±20`, `z ±10`
+- cylinder extents: `z ±4`
+
+So the cylinder is fully internal, and subtracting it forms a sealed cavity (enclosed void), not an exterior-opening blind hole.
+
+## 3) Built artifact inspection
+
+Because default process rejects enclosed voids, the suspicious fixture cannot export STEP in default mode.
+To inspect actual resulting topology, the same geometry was rebuilt with a temporary additive schema (no code changes).
+
+Observed outputs:
+
+- additive rebuild succeeds and exports STEP;
+- resulting STEP uses `BREP_WITH_VOIDS` with an outer closed shell plus an oriented inner closed shell;
+- the same file fails CLI import/analyze with `Missing MANIFOLD_SOLID_BREP root entity`, confirming current analyzer only imports manifold-root forms, not `BREP_WITH_VOIDS` roots.
+
+As a baseline comparison:
+
+- `chimera_a_box_root_baseline.firmament` (through-hole) exports `MANIFOLD_SOLID_BREP` and is analyzable;
+- `Aetheris.Firmament.FrictionLab/Cases/blind-hole-mount-block/part.firmament` (explicit `place.on_face`) succeeds under default and is analyzable as a single-shell part.
+
+## 4) Boolean result vs enclosed-void checker
+
+Checker behavior is consistent with implementation:
+
+- `FirmamentSchemaEnclosedVoidValidator` runs for non-additive/default processes.
+- It calls `BrepEnclosedVoidAnalyzer.Analyze(...)` on the final body.
+- Analyzer reports enclosed voids from `body.ShellRepresentation.InnerShellIds`.
+
+Given the suspicious fixture constructs an actual inner shell, checker classification is expected and correct.
+
+## 5) Final diagnosis
+
+**Fixture/setup bug.**
+
+The FRIC-B0 `zone_b_blind_cyl` case was treated as a blind hole in reporting, but its parameters/model-space placement produce a fully enclosed cavity.
+
+No kernel blind-hole subtraction defect was required to explain the result.
+No enclosed-void checker defect was required to explain the result.
+
+## 6) Recommended follow-up
+
+1. Reclassify FRIC-B0 `zone_b_blind_cyl` from “blind-hole” to “contained cavity” in pressure-test reporting.
+2. Add a true exterior-opening blind-hole fixture variant for FRIC-B0 using explicit `place.on_face` semantics.
+3. Keep `FIRM-SCHEMA-0006` expectation for the current fixture under `default` as a guardrail.
+4. Optional tooling follow-up: extend `aetheris analyze` importer path to accept `BREP_WITH_VOIDS` roots so enclosed-void artifacts can be inspected directly via CLI.