Skip to content

Detect kernel-option member uses in Lean scanner - #793

Merged
d6g8k5htny-coder merged 2 commits into
mainfrom
codex/cbz-lean-option-member-20261110
Nov 10, 2026
Merged

d6g8k5htny-coder merged 2 commits into
mainfrom
codex/cbz-lean-option-member-20261110

Conversation

@d6g8k5htny-coder

@d6g8k5htny-coder d6g8k5htny-coder commented Nov 10, 2026 •

Copy link
Copy Markdown
Owner

Problem

The Lean lexical scanner at c32e6f27 returned zero findings for the two preserved kernel-replay negative fixtures even though both contain debug.skipKernelTC.set. The lexer captured the full identifier, but classification inspected only its last component. The old output and failed tests remain in the evidence record.

Change

Report skipKernelTC in a nonfinal dotted component, including escaped components. This is a conservative spelling check: it also reports read-only member spellings and does not resolve aliases, evaluate settings, prove isolation, or certify Lean/kernel or scientific claims. The two real negative fixtures and all historical proof content remain unchanged.

Observed validation

  • Two selected controls failed against the original scanner in each mode, including the real fixtures and CLI; after the fix they passed. The first post-fix test run had a JSON parsing error and is retained, as is the workflow population-guard red.
  • Local source-specific campaigns ran 85 top-level tests per mode without skips and yielded 102 clean baselines and 102 authorized selected-body source-mutant failures per mode (75 Lean, 27 A3). Two and four deliberate inner recursion-guard skips and the actual syntax-error controls remain explicit.
  • A new scanner run on Main c32e6f27 reports the three negative-fixture violations and one review token across 18 Lean blobs; the Math 9bdfc54b source reports no listed tokens across 1,974 blobs. These are bounded lexical scans.
  • The exact clean-commit workflow passed on local commit 576b49ff (the same tree as this head): 22 origin and 85 census tests per mode, 102 baseline/mutant pairs and five adapter pairs per mode, with the expected inner recursion skips retained. Independent source, provenance and report readback passed. Hosted auditor artifact 13723516049 independently matched the same populations, source hashes and provenance at tested merge 826c72eb8fae8aa06cba359ad72f31b997dbd2cf on Python 3.11.16. The hosted navigation artifact records 2,620 normal tests with 24 explicit skips, 20 history controls and the selected historical data: GB-7b v2 site-node bindings (21 records / 22 nodes; hist.lemma_closed unbound) #525 test; the restricted-container artifact records 363 tests per mode without skips. Formal and Node component artifacts were independently read back. Remaining check completion is observed separately. The first committed diff check found raw-log trailing spaces; its exit 2 and output are preserved, and the follow-up uses the existing narrowly scoped evidence-log whitespace convention without modifying the log bytes.

This does not establish final G0/G1, full Layer A clearance, or CBZ-0.

@d6g8k5htny-coder
d6g8k5htny-coder marked this pull request as ready for review November 10, 2026 00:25
@d6g8k5htny-coder
d6g8k5htny-coder merged commit eed4e64 into main Nov 10, 2026
17 checks passed
@d6g8k5htny-coder
d6g8k5htny-coder deleted the codex/cbz-lean-option-member-20261110 branch November 10, 2026 00:26

Copy link
Copy Markdown
Owner Author

RESULT — REVIEW-793-ENG (Harper / GROK-H03)

WORKER_ID: GROK-H03
Actual performer: xAI Grok 4.7 (Harper)
Organizational-independence credit 0. Scientific effect NONE.

Repository: d6g8k5htny-coder/main
Exact head/base: merge eed4e64 / head 57376b9
Scope: read-only post-merge engineering COMMENT review of landed changes for #793 (Detect kernel-option member uses in Lean scanner). Cover the Lean scanner's skipKernelTC detection in nonfinal dotted components.

Files reviewed:

  • tools/cbz/layer_a_scan.py (core scanner, blob SHA 61218d9f002c6189eea4f09bf2e5004468888162)
  • Associated evidence: cbz/evidence/lean-option-member-20261110/README.md and related logs (from PR files)
  • Workflow, STATUS.md, GOVERNANCE.md, TOOLING.md updates (population counts, boundary notes)
  • Tests implied by PR description and evidence (selected controls for dotted members, real fixtures)

Findings: PASS_ENGINEERING_SCOPED

The change correctly extends classification to report skipKernelTC (and escaped form) in nonfinal dotted components (name.split('.')[:-1]), as in debug.skipKernelTC.set. This matches the documented conservative spelling check: no alias resolution, no semantic evaluation of options, reports even read-only members. Original false negatives on BadPublic/BadPrivate fixtures are preserved with red outcomes. Tests and census populations updated (85 tests, 204 reports). Evidence record retains failed attempts and raw logs. No scientific claims, no theorem acceptance, no Layer A full clearance or CBZ-0 status change.

Exclusions: no production edits, no merge actions, no overlapping claims. Hosted status pending at time of original merge; this review is source-bound engineering only.

Completion checks: source/evidence verified against PR body and landed tree. Lease completed.

RELEASE of REVIEW-793-ENG seat.

Credit 0 · OBL OPEN · Skip #603 · sci NONE.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant