3.5 KiB
Analysis source S2b Windows checkpoint — 2026-09-13
The original gate archive records private typed lowering at analysis source a2a28f8d23a46ba1696b6a8410d4d3b5c1808f49. This evidence-only publication does not integrate that analysis branch into production. The successful scope is lowering an ordinary Project and parsed typed units through the private harness; public source8 writing, migration, desktop workflows and release acceptance remain outside it.
| Gate | Actual corrected result |
|---|---|
| Core library listing | 142 unique tests |
| Typed parser/lowering selection | 34 passed, 0 failed |
| Raw-token selection | 1 passed, 0 failed |
| Full core library | 126 passed, 16 failed, 0 ignored; exit 101 |
| All-target core Clippy with warnings denied | Exit 0 |
All 13 new lowering tests pass in the typed and full runs; all 129 baseline test names remain. Five corrected gates bind the same 531 unchanged committed input blobs. This is oa-core --lib runtime evidence plus all-target Clippy, not all-target runtime or documentation-test acceptance. Original commands, Windows environment and executable identities remain in the receipts; executable/toolchain files are omitted.
The original source 472fe7a4339a23a3339b2b91e8d53d76442f60c4 retains its 33 passed / 1 failed, exit 101 typed result. Its positive AsKind fixture supplied an already named quantity, which the existing semantics reject. The correction changes only that test file, using an anonymous matching dimension and retaining the former input as a negative case. Original failure evidence, source review and correction review are preserved separately from the corrected pass.
The 16 full-library failures have the same names, panic locations and complete diagnostic lines as baseline b2c970b4673e062e47eadd12906d2316c4b1a7b3 (113 passed / 16 failed). The failure-cause review identifies schema8 output or migration reaching the unchanged public parser that accepts versions3–7. These failures occur before later nesting/save/recovery assertions; matching causes do not waive them into a full pass. Historical S2a production-language 7/9 remains scoped to b2, as documented in the S2a checkpoint.
The independent archive audit found no actionable archive defect after checking actual outcome membership, immutable Git input blobs, raw receipt/artifact bindings and every failure section. The preserved original manifest is SHA256 d5b2668207438e319c2ab98395e2483c016088c5446b6bc99d797a066384a62f; all 43 original archive files, including that manifest, remain byte-identical. The review is also copied unchanged. Local absolute paths inside original receipts are historical provenance; the complete portable payload is under archive/, with repository-relative paths bound by the publication inventory.
No gates were rerun for this documentation commit. Publication validation checks copied bytes, exact file membership and committed Git blob identities. Preserve access to the three identified analysis source commits when sharing this evidence; the production branch's code is not a substitute for those tested sources.