Skip to content

WIP: finish mathematical implementation and verification infrastructure (certify in successor PR) - #113

Draft
HMarcusWH wants to merge 597 commits into
mainfrom
research/final-mathematical-closure
Draft

HMarcusWH wants to merge 597 commits into
mainfrom
research/final-mathematical-closure

Conversation

@HMarcusWH

@HMarcusWH HMarcusWH commented Oct 21, 2026 •

Copy link
Copy Markdown
Owner

Revised scope — implementation and verification infrastructure

PR #113 completes the promised mathematics, implementation, historical/security compatibility and working verification/release tools. Full independent mathematical certification and publication happen in a successor PR (temporary label: CERT).

This replaces the earlier single-PR completion sequence. It does not relax the mathematical requirements or claim that an unfinished proof is certified. The original 45 acceptance cases, 11 mathematical claims, historical T/A2 checks, baseline 259-declaration inventory, accepted release bytes and protected core/pins remain authoritative.

Implementation scope for #113

  • Reconcile each promised theorem to its exact Lean type and assumptions; mark implementation complete / partial / missing independently of acceptance.
  • Repair source contracts and provenance wiring, including original 21-producer binding, 76 additional controls, compiler trigger candidates and preseal status.
  • Build a separate infrastructure acceptance gate, successor-PR identity contract, authenticated reviewer-input transport and integrated release dry-run machinery.
  • Reproduce historical T and A2 alongside their successors, then obtain genuine independent security/engineering review before activating a modified protected workflow profile.
  • Reproduce the source-faithful draft paper, freeze the final source, run all required engineering/Lean/historical contexts, and request a separate owner merge decision.

Deferred to successor CERT (not waived)

  • Capture exact-Hcert original 4+9+8 producing jobs, 259 baseline declaration receipts plus approved additions, 45 original positive and 45 semantic negative controls, 11 positive/negative claim pairs and 76 source-bound owner controls.
  • Obtain authenticated independent 45/45 and 11/11 mathematical acceptance, exact merge-source checking, final-paper scholarly review, clean reproducibility and independent preseal/postseal verification.
  • Publish only the exact independently checked package after separate explicit owner authorization.

Approval/status boundaries

Branch: research/final-mathematical-closure. Original base: 17c315f7df7abf457eb94d393bd526d46da6f6f4. The actual current head must be obtained from GitHub; old SHA-labelled results are historical only.

Successor numbering: GitHub #114 is already a closed repair PR merged into this branch. The certification PR will be opened after a real authorized #113 merge, and GitHub's actual new number will be used.

Operational handoff: docs/operations/pr113_implementation_scope_and_certification_handoff.md (the split's 13 implementation steps and 11 certification steps).

Do not merge or tag without a separate owner decision.

…s to authentic producer jobs without v2 self-approval
…ped test receipts

Preserve frozen mathematics and historical CI policy. Add early research-Python syntax regression, reject unowned candidate receipts, and permit claim-only positive/negative tests without manufacturing approvals.
… profile

Audit the exact former-provider Git delta while preserving the original A2 historical verifier and four kernel-checked Lean proof blobs. Explicitly fail closed on changed workflow/security/test providers, dependency pins, source shadows, missing imports or missing independent approval. No historical gate is relaxed or replaced.
… acceptance

Fail closed when executable verification, research release, Windows integrity or either application contract fails or is missing. Add source-bound negative regression tests. Historical T verifier and approval records remain untouched.
…audit

Permit only two explicitly located formatting-only blank lines after the original A2 proof imports. Any change to import content, theorem provider, or other root bytes still fails closed. Versioned successor remains unapproved.
…iagnostics

Preativation FINAL-01 to FINAL-04 implementation candidate.
Preserve original 25 workflows and historical T/A2 authorities.
No workflow activation, evidence approval, merge or release.
… CI failure

Capture the adversarial renderer inside its unittest and assert the expected
failure identity without emitting a misleading external failure marker.
…ted regression failures

R6-01: validate both installed historical_runtime_v3 call sites while retaining
original preactivation archive/T/A2 and historical v2 fragment assertions.
R6-02: emit bounded method-level active30 failure diagnostics, secure redaction,
source-bound failure receipts, timeout and skip/error negative regressions.

Preativation engineering-only; no workflow activation or security approval.
… lifecycle

Export one nonauthorizing v3 receipt path for all phases; the staged Research consumer reauthenticates genuine original T/A2 for #113 and inherited M113 for descendants. Reject ambiguous detached JSON and borrowed evidence. Preserve historical A2 executor and workflow installation status. Add synthetic four-phase/negative integration tests and repin the staged test Git blob.
Add synthetic four-phase GITHUB_ENV producer/consumer coverage, reject NaN and overlarge floats in original historical events, and record an explicit noncircular two-stage cutover runbook without granting security, merge or publication authority.
…utant bytes

Repair v2 current A2 intake to require source_1 through source_4, full/focused command sets, and only the declared one-step semantic source mutations. Add real consumer negative controls for missing probes, unrelated modified Lean source, parser, import and timeout failures.
Scope original A2 compiler expectations to immutable introducer source. Validate active30's phase-aware required consumer against exact staged Git blob and full 25/30 workflow profile. Replace obsolete v2-only staged assertions; preserve negative no-skip and original A2 historical source identities.
Update the unapproved migration manifest's single stale Git object reference to the revised phase-aware A2 test. Preserve original A2 and all other historical/approval contract fields. Existing original and atomic cutover source guards now enforce the same staged identity.
…light

Ensure the phase-neutral four-event Research consumer, historical Git source identity, strict external event JSON and semantic A2 compiler receipt validation are exercised before expensive full research compilation. No historical verifier or protected workflow changes.
…xtures

Bind staged A2 identity to committed Git bytes and unapproved R3 manifest rather than obsolete blob literals. Run complete four-mutant/four-source diagnostic fixtures through the real v2 receipt validator, retaining negative parser/import/resource controls. Add the boundary suite to exact-head fast QA. Preserve historic T/A2 and 25 active workflows.
Replace assertion-only M113 child checks with explicit runtime validation of twelve exact-H successful check identities, original independent security review, historical-only CI scope and producer/reviewer IDs. Add optimized-Python synthetic rejection controls to fast preflight. No historical source or active workflow alteration.
…ow cutover

Select the native v1 or v2 compiler only after the complete exact-Git source-profile audit. Preserve the original 40 declarations and four semantic controls, reject source drift and never fall back after failure. Add 13 routing regressions, register them in the 53-suite preflight, and consolidate the source/evidence/approval/activation sequence required for merge readiness. Historical verifiers, active workflows, reviewer authority and merge state are unchanged.

Copy link
Copy Markdown
Owner Author

Consolidated merge-completion plan and source repair

Pushed 19d2f29d36bd426e19298a2e59b175affb928df0 on the existing PR branch. The full ordered plan is docs/operations/PR113_MERGE_COMPLETION_2026-10-28.md.

New defect repaired: the mandatory current-source A2 Lean test always selected the v1/25-workflow compiler. It would reject the installed 30-workflow configuration. It now selects v1 or v2 only after the complete exact-Git workflow-profile audit. Both native compilers and historical A2 remain unchanged. There is no fallback after an audit or compiler failure.

Verification: 13 routing tests passed locally in normal Python and the same 13 passed under -O. These use synthetic profile/module inputs plus real temporary Git-head tests; they are not Lean compilation, historical replay, or independent review. Git blob identities of all five pushed files match the locally checked bytes. The fast preflight now lists 53 suites. New exact-head GitHub CI is running; full-suite success is not asserted.

Remaining merge dependencies: genuine preactivation historical evidence; real owner-authorized independent reviewer roster; independent source-security review and separate installation authorization; atomic ten-path cutover; genuine final-head T/A2 producer artifacts and artifact-bound independent approval; all twelve required contexts and applicable downstream Research steps; effective main protection and separate owner merge decision.

This comment is an engineering handoff, not an approval. No protected workflow was activated, no reviewer identity fabricated, and no merge or publication performed. CERT retains independent mathematical/scholarly acceptance.

…02 compilation

Replace the mismatched current-head historical integration with mandatory fresh historical and current-source compiler legs. Execute the immutable original gate in a clean detached historical checkout, share only verified pinned dependencies, prohibit historical project-build reuse, validate complete receipts and retain failure diagnostics. Preserve the registered P02 method's current-source coverage and the standalone native successor test. Reconcile exact integration pins and staged/active source guards together; add 13 real-Git protocol regressions and the 54th fast preflight suite.

No installed workflow, original executor, Lean proof, dependency pin, archive verifier, reviewer authority or merge setting is changed. Research green remains actual-CI evidence, not historical migration approval or merge authorization.

Copy link
Copy Markdown
Owner Author

Research source-target repair deployed — 073df6e1ccc855676123ded41859f1a75f55d5c4

The previous exact-head Research run had 1,896 tests and one failure: the immutable historical A2 executor was invoked against current source whose twelve known historical providers intentionally differ. The repair is now committed as one coordinated nine-path change.

Upstream and downstream coverage

  • The registered LocalizedCompilerIntegration.test_actual_final_source_and_semantic_mutations method creates a fresh detached checkout at original A2 f71a76032b169aac7bfe0694aa715339a94d4e05 and executes that checkout's original gate/support in a separate process.
  • Only clean, exact-pinned dependency packages are reused after equality of all three committed build/dependency configuration files. Historical project .lake/build must start absent; current project proof objects are not copied.
  • The SAME method then executes the audited native CURRENT-source compiler. This preserves P-02's acceptance-map meaning; historical compilation alone would not satisfy its registered current-source compiler route. The standalone current-source compiler test remains mandatory too.
  • Both receipts must satisfy the complete 40-declaration/four-semantic-mutation validators, source/tree identity, artifact/log hashes and unchanged-source requirements. Any failure remains nonzero; raw receipts/logs are printed before temporary evidence cleanup.
  • Exact current adapter pins and staged25/active30 source guards were reconciled together. Original historical integration bytes remain pinned at their immutable Git revisions.

Verification actually performed

  • 13 focused protocol tests passed normally and the same 13 under optimized Python.
  • These use real temporary Git/worktrees and dependency source identities but SYNTHETIC compiler/receipt adapters. They are not actual historical Lean execution, reviewer approval or migration evidence.
  • All eight changed/added Python sources parsed; all nine uploaded Git blob identities matched the locally checked bytes.
  • Fast preflight now registers 54 distinct suites. A full local repository/Lean run was NOT performed.
  • New exact-head Research run: https://github.com/HMarcusWH/PermanssonLean/actions/runs/40626382833 — running at the time of this comment. No green Research result is claimed yet.
  • New frozen-source-preservation and release-export-integrity jobs have passed. The independent archive-restoration gate remains red at its original historical tooling boundary.

No installed workflow, original A2 executor, Lean proof, dependency pin, acceptance map, archive verifier, reviewer policy or merge setting changed. The existing C10, D0, default-build, import-coverage and source-firewall steps still must actually execute successfully after regressions pass. Research success must not be relabelled as full historical migration, security approval or merge readiness.

Implementation note: docs/operations/PR113_RESEARCH_CI_SCOPE_REPAIR_2026-10-28.md.

…pport blob

Keep both revisions exactly pinned, including historical localized_prefix_support b07bc7d and current 93462e9; preserve the unchanged gate binary and original integration source. Recreate real synthetic two-revision support drift in 16 replay tests, add negative historical/current substitution and an authentic Git source-object pin test. No workflow, accepted evidence, Lean theorem or security approval changed.
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.

2 participants