Skip to content

Preserve/local progress 2026 05 06#45

Open
MrSnowNB wants to merge 304 commits into
aws-samples:mainfrom
MrSnowNB:preserve/local-progress-2026-05-06
Open

Preserve/local progress 2026 05 06#45
MrSnowNB wants to merge 304 commits into
aws-samples:mainfrom
MrSnowNB:preserve/local-progress-2026-05-06

Conversation

@MrSnowNB

@MrSnowNB MrSnowNB commented May 6, 2026

Copy link
Copy Markdown

Issue #, if available:

Description of changes:

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

MrSnowNB added 30 commits April 27, 2026 13:30
[AIFIRST] Docs consistency audit — provenance.md fix + DEMO-SPRINT-PLAN sync (T-2026-04-27-001)
…te section, clean checkboxes (T-2026-04-27-001)
Deterministic LLM-free gate for 6 COBOL->MD translations.
extract_ground_truth + extract_md_claims + gate_compare.
cp1252 fix, synthetic_paragraphs support, .gitignore artifacts.
Post-mortem: COBSWAIT inline-only program edge case resolved.
…run-logging

fix: filter Cobol-REKT scope terminator false positives + add run logging
- validation/cobol_vocab.py: single source of truth for
  COBOL_SCOPE_TERMINATORS set (replaces inline def in extract_ground_truth.py)
- validation/lint_md.py: pre-commit MD linter; exits 1 if any
  procedure_paragraph name matches a known terminator or reserved word;
  add to manifest prereqs before extract_md_claims.py
- validation/extract_md_claims.py: emits lint_warnings array in claims
  JSON listing any paragraph names that match cobol_vocab terminators
- validation/gate_compare.py: new Check 2c (invalid_paragraph_names)
  gives explicit failure category for terminator names vs generic
  hallucinated_paragraphs message

Fixes root cause of END-IF/scope-terminator paragraph contamination.
4 failing MDs to be corrected in this same PR per architectural decision.
feat(validation): lint_md pre-commit guard + cobol_vocab single source of truth
…n GT; remove GOBACK from blocklist

Two targeted fixes for the agent failure surfaced during MD cleanup:

1. extract_ground_truth.py:
   - The scope terminator filter previously only cleaned dead_code_paragraphs.
   - Cobol-REKT RC8 also places scope terminators (e.g. END-PERFORM) in the
     reachable paragraphs list for some programs (CBACT01C, CBCUS01C confirmed).
   - paragraphs_reachable now filters COBOL_SCOPE_TERMINATORS identically to
     how dead_code_paragraphs does, using the same set imported from cobol_vocab.
   - scope_terminators_suppressed now includes suppressions from both lists
     (deduped), so the gate report remains accurate.
   - Inline COBOL_SCOPE_TERMINATORS set removed; now imports from cobol_vocab.

2. cobol_vocab.py:
   - GOBACK removed from COBOL_RESERVED_NOT_PARAGRAPHS.
   - GOBACK is a legitimate COBOL paragraph name used in CardDemo programs.
   - Its presence in dead_code_not_declared_in_md (warning) is correct behavior;
     the linter blocking it was wrong and caused real paragraphs to be stripped.
fix(validation): filter scope terminators from reachable paragraphs; remove GOBACK from blocklist
MrSnowNB added 30 commits May 5, 2026 11:31
ISSUE.md was deleted in Commit A at a198dc7 (Step 2, 2026-05-04),
not deferred to Step 3. Plan-doc staleness corrected.

Per OPERATION-TIDY-PLAN.md Section 5 housekeeping.
Operation Tidy step OP-3-07/08/09. Renames pass1/ → waves/wave-1/, pass2/ → waves/wave-2/, pass3/ → waves/wave-3/ to establish consistent wave-N naming for the multi-pass proposition pipeline. All blob SHAs preserved. No content changes. Moved paths: validation/pass1/ → validation/waves/wave-1/ (43 files), validation/pass2/ → validation/waves/wave-2/ (18 files), validation/pass3/ → validation/waves/wave-3/ (8 files)
…/COCRDUPC/cfg/

Operation Tidy step OP-3-11. Moves the COCRDUPC CFG report from the ad-hoc .report subdir into the canonical structure/reports/{PROG}/cfg/ layout, matching the rekt/{PROG}.cbl.report/cfg/ nesting pattern. Blob SHA preserved (cfg-COCRDUPC.cbl.json unchanged).
Operation Tidy steps OP-3-16 (waves/README) and OP-3-NEW (rekt/README). waves/README.md documents wave-1/2/3 architecture, per-phase file patterns, naming conventions, and provenance. rekt/README.md documents rekt/ as the smojol-cli --reportDir output zone: data flow (run_rekt_all.py → rekt/ → extract_cfg_summary.py → structure/), regeneration instructions, 30-program inventory, and do-not-edit policy. OP-3-10 (rename rekt → triage) FORMALLY CANCELLED: rekt/ is the active COBOL-REKT tool output zone; renaming requires runtime code patch outside Operation Tidy scope. Documentation-only replacement via rekt/README.md is the correct resolution.
Resolves E-ERR-4 and E-ERR-5 from Commit E close-out.
… to v2.2

Resolves template/spec drift identified in Commit F planning audit.
Templates were authored for v1.0 5-gate model while 00-aifirst-protocol.md
evolved to v2.1 6-gate model. Every fresh-agent session loading v2.1
looked for G0-decompose.template.md and found nothing.

Changes:
  - Delete 5 v1.0 templates
  - Add 6 v2.1 templates with schema_version aifirst/2.1
  - Bump 00-aifirst-protocol.md to v2.2 with ## v2.2 Changelog

No behavioral changes to gate definitions, failure loop, or toolchain.
- Re-locked all 12 original programs with working-tree SHAs
  (resolves drift introduced by sync.py:208 lock bug, fixed in 04999e5)
- Regenerated gold-candidate scaffolds with fresh frontmatter SHAs
  (resolves 7 stale-scaffold warnings in doctor)
- SHA coherence verified: 12 OK / 0 DRIFT
  (manifest source_sha == git hash-object app/cbl/*.cbl for all 12)

Programs: CBACT01C, CBACT02C, CBACT03C, CBACT04C, CBCUS01C, CBSTM03A,
          CBSTM03B, CBTRN01C, COBSWAIT, COMEN01C, COSGN00C, COCRDUPC

Phase 1 complete. Ready for Phase 2 (Qwen translation of scaffolds).
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