gnomonlearn
Docs/Historical native frontend baseline

Historical native frontend baseline

Versioned draft snapshot

RFCs, reference catalogs, and evidence reflect Spec 3b2e3e7a114fb826b1445d39a48756b8fff629cd. Counts and pending/status notes belong to that snapshot. Read the dated project update for later progress.

Source repositories are private during staging. Links marked “repository access required” need access to the linked repository.

This is a trace and gap record from the pinned draft specification. It does not establish proof or implementation conformance.

This directory preserves the original native observation bytes and protobuf source pin for framework commit 914b9ecede705e11a909ef7895f38649a8415b41. These files are immutable historical evidence; current runs write outside this directory.

audit.textproto uses the gnomon.spec.proposal.framework.v1.FrontendAudit schema from spec commit d11543c6dacbd42cdf25df2d45e5c97fb9405a33. That spec commit also contains the original audit driver and Lean export harness. Their SHA-256 values appear in the report, and were checked against its immutable Git objects when the baseline was archived. The pin’s four source hashes were checked against the framework commit’s immutable Git objects. All 35 input source hashes and sizes were checked against the unchanged proposal fixtures.

Observed outcomes were 16 native accepts, 18 native rejections, and one bounded runtime failure on rule-expressions.gnm. All three proposal-negative inputs were accepted. This is generic parser behavior, not passing language, typed AST, source-map, or compiler conformance.

The current audit driver can reproduce the old producer observations with:

python3 tools/audit_framework_frontend.py --framework-root ../gnomon \
  --pin tools/proposals/framework-frontend-history/914b9ecede705e11a909ef7895f38649a8415b41/pin.textproto \
  --output /tmp/gnomon-historical-native-audit.textproto

A reproduction may have a different driver hash or runtime-failure detail; it must retain the old source/fixture bindings and classify stalls as invocation failures. The original report is not overwritten by reproductions.