F379 trivial scalar catch object - final verification Date: 2026-08-12 UTC LLVM 18 build: cmake --build build -j2 PASS; libsymcc.so rebuilt and linked LLVM 17 compatibility build: cmake --build build-llvm17 -j2 PASS; libsymcc.so rebuilt and linked Complete LLVM lit: cmake --build build --target check -j2 PASS; 237 passed, 1 unsupported, 0 failed, 238 discovered, 212.56 s Targeted Python regression: scalar catch + lifecycle + cleanup + typed exception suites PASS; 30 passed, 27 subtests passed, 0 failed, 3.00 s Capability-closed Python gate: PASS; 868 passed, 229 subtests passed, 137.12 s collected=868; skipped=0; xfailed=0; xpassed=0; deselected=0 missing-nodeids=0; unexpected-nodeids=0 canonical nodeid SHA-256: 352426efcf9c42a6f06ce2affd9e0ac2d66779db14c48d02100af1714b96fcd7 Generated artifact replay checks: scalar-catch-object.json: values 7,42; scalar capability and value op present gep-reject.json: rejected unsupported exception-object use store-reject.json: rejected unsupported exception-object use atomic-reject.json: rejected atomic scalar object load pointer-materialization-reject.json: rejected ptrtoint object materialization Static and visual gates: ruff: PASS, 0 diagnostics py_compile: PASS SVG XML + rendered PNG: PASS, PNG 1800x1100 RGB Review findings retained in the record: Review round 1 repaired the test capability fixture and exposed a production scalar-to-typed capability dependency. Review round 2 closed all-use-set admission and added GEP/store/atomic fail-closed cases. Review round 3 checked phase, frame, exact payload width, fresh-executor restore, and evidence claim boundaries. All complete gates were rerun after the fixes. Claim boundary: These results establish a direct read-only scalar projection from the active caught payload. They do not establish an exception arena, object addresses, layouts, destructors, RTTI inheritance, complete C++ ABI behavior, benchmark throughput, coverage, or vulnerability-discovery improvement.