Skip to content

Independently audit BIT COPY support propagation at the retained caller - #4206

Merged
Mister-Meeseeks merged 3 commits into
CrocSwap:mainfrom
ToikY:codex/bit-copy-cap-intersection-audit
Oct 21, 2026
Merged

Mister-Meeseeks merged 3 commits into
CrocSwap:mainfrom
ToikY:codex/bit-copy-cap-intersection-audit

Conversation

@ToikY

@ToikY ToikY commented Oct 20, 2026

Copy link
Copy Markdown
Contributor

PR4200's seven-rank COPY floor uses a support-propagation implication: a branch with private-frame intersection L can affect target p only if L lies in its final cap. This answers its explicit independent-review handoff with a separately written paired-execution proof and exact caller-premise checks.

The proof isolates branch-injected influence while retaining arbitrary correlated public dirt and the same fixed trace. Every dependency must preserve L⊆U ⇒ L⊆W; upward public MOVEs and common-frame ADDs suffice. This is a sufficient local condition, not a necessity claim for every whole trace with dead/cancelling edges. Linearity/unchanged-q and private closure are separated from the causal lemma. Downward-MOVE and unequal-frame-ADD controls deliberately leave the predecessor model; they are not predecessor bugs.

Evidence:

  • Exact original 38,927-record frame trace, 636 rational span ranks, 111 cap equalities and 6,470 orthogonality checks. Public suffix: 1,699 records = 760 upward MOVEs + 939 same-frame ADDs, with unchanged nine sources and no COPY.
  • Separate finite closure: 283 states / 1,780 transitions; code-disjoint oracle: 34,208 symbolic states / 317,824 transitions / 1,632 closed walks, arbitrary dirty columns and cancellation controls.
  • Independent Fraction cross-check of 636 ranks / 2,904 MOVE differences / 111 caps. Six normal/strict-optimized registered commands pass locally and from exact committed Git-free bytes. Source-byte and declared-source ancestry checks pass.
python3.12 -B scripts/research_package.py --package research/bit-copy-support-audit --check-only
python3.12 -B scripts/research_package.py --package research/bit-copy-support-audit
python3.12 -B scripts/check_research_commit.py --package research/bit-copy-support-audit --check-source-history

Canonical base 12fa8c5c844be21bfefb38a19503fbd867a11244, priority 7. Handoff-first assignment coverage-law-change, explicitly changed from the generic complex ceiling card to this named BIT proof audit; claim. Pinned predecessor PR4200 ddfc3133cbdf548e163c867d5671f8f01dd5d5ae; 320 open/draft bodies and current hub/canonical thread checked, with broader closed/code/Discussion coverage partial. New PR4204 paired endpoints and PR4205 rational FFT are distinct; Ross Richardson's active mixed-total construction is excluded. Minimal untouched exports are retained in this PR's own source commit 2c2b29438773f12b3bb5903723c3bb16f0dcd583, preserving cold availability and original attribution.

This does not rerun R/K or recipient/projected-recipient ranks, all 660 scalar endpoint columns, full root closures, physical MOVE semantics, complete invoice, precision or the outer theorem. The original seven-rank floor/projection argument remains Ross Richardson's PR4200 work. Source credit: Ali Romman, Sagar Thakrar, DreamingOfClouds/eumemic/PR589/PR555 and retained NOTICE lineage; historical direct-decoder floor credit remains Russell Seligmann/PR757 and Cody Harris/PR761 following Joel Pulikkan/GamingPuzzled/PR620.

Prepared for ToikY with substantial OpenAI Codex assistance, including separate proof, oracle and review agents; no human/formal-kernel acceptance. Apache-2.0. No new supplier or κ, merge or selected-frontier change. Maintainer edits disabled. Concrete open next questions and remaining obligations are in README/TASK.json.

@rcfc-jreed

Copy link
Copy Markdown
Contributor

Independent scoped audit of the BIT COPY support-propagation package at e77cea5e: packaging seal defect; scientific finite layer otherwise consistent

Code-disjoint reimplementation; does not import package source_check.py / independent_oracle.py / frame_rank_cross.py. Package bytes from head e77cea5e7998b20fc7126019bd8368d7989fdb33 via GitHub contents API. CLAIM.

Defect (S1): PACKAGE-REVIEW.json file_sha256 is stale for exactly three root-binding files after the pending “final root package byte check” noted in its own finding resolution:

file sealed in PACKAGE-REVIEW actual package bytes
README.md b2c612df…4720bc 1dfc8efd…a9b3f8
SOURCES.json 675bed1c…584b7f 13134769…07db117
VERIFY.json e9370cc0…2e4fee61 2f13e167…4893881

Scientific cores listed in the same seal table still match. REVIEW-FILE-MAP current equalities, SOURCE-PINS, oracle provenance, RESULT/source_check binding, and count cross-links are consistent.

What passes (24 of 25):

  1. Pins. Base 12fa8c5c…; predecessor ddfc3133… (Prove the BIT COPY floor for closed snapshot paths #4200). Status PASS_INDEPENDENT_CAUSAL_SUPPORT_AUDIT_NO_NEW_KAPPA; new_kappa_claimed=false.
  2. Source pins / code. Five exports match bytes; RESULT word/geometry/code_sha256 bind to packaged inputs and source_check.py.
  3. Trace / frame. 38927 records = 36005 ADD + 2904 MOVE + 9 COPY + 9 ERASE; frame-rank census sums to 636; 111 rank-7 caps; postscatter 1699.
  4. Toy / oracle. Closure 283/1780; oracle 34208 states / 1632 walks / 317824 transitions; provenance normal≡optimized.
  5. Reviews. INDEPENDENT-REVIEW outstanding empty; verification counts match RESULT/ORACLE/FRAME; REVIEW-FILE-MAP equalities hold.

Not covered: K/R coefficients; decoded scalar endpoint; physical MOVE semantics; live re-execution of oracle/source checkers.

This is a finite replay of committed receipts, not a formal proof. Suggested fix: refresh PACKAGE-REVIEW.file_sha256 (or drop those three entries) for the post-binding README/SOURCES/VERIFY bytes.

check_4206.py
#!/usr/bin/env python3
"""Independent finite controls for PR4206 (BIT COPY support propagation audit).

Recomputes source pins, review-map byte equalities, frame-rank census sums,
oracle/source count cross-links, and PACKAGE-REVIEW seals from committed
receipts. Does not import package source_check.py / independent_oracle.py /
frame_rank_cross.py.

Package bytes from head e77cea5e7998b20fc7126019bd8368d7989fdb33 via GitHub API.

James Reed (@rcfc-jreed), with substantial AI assistance (Claude, in Cursor).
Apache-2.0.
"""
import argparse, hashlib, json, sys
from pathlib import Path

HEAD = "e77cea5e7998b20fc7126019bd8368d7989fdb33"
BASE = "12fa8c5c844be21bfefb38a19503fbd867a11244"
PRED = "ddfc3133cbdf548e163c867d5671f8f01dd5d5ae"
results = []


def check(name, ok, detail=""):
    results.append((name, bool(ok), detail))
    print(("PASS " if ok else "FAIL ") + name + ((": " + str(detail)) if detail else ""), flush=True)
    return bool(ok)


def load(pkg, name):
    return json.loads((pkg / name).read_text(encoding="utf-8"))


def sha(path: Path):
    return hashlib.sha256(path.read_bytes()).hexdigest()


def run(pkg: Path):
    sources = load(pkg, "SOURCES.json")
    task = load(pkg, "TASK.json")
    pins = load(pkg, "SOURCE-PINS.json")
    result = load(pkg, "RESULT.json")
    oracle = load(pkg, "ORACLE-RESULT.json")
    oprov = load(pkg, "ORACLE-PROVENANCE.json")
    frc = load(pkg, "FRAME-RANK-CROSS.json")
    primary = load(pkg, "PRIMARY-REVIEW.json")
    package = load(pkg, "PACKAGE-REVIEW.json")
    irev = load(pkg, "INDEPENDENT-REVIEW.json")
    rfm = load(pkg, "REVIEW-FILE-MAP.json")
    verify = load(pkg, "VERIFY.json")

    check("I1 SOURCES/TASK/PACKAGE base_commit == 12fa8c5c…",
          sources["base_commit"] == BASE == task["base_commit"] == package["base_commit"])
    check("I2 predecessor / SOURCE-PINS head == ddfc3133… (PR4200)",
          result["predecessor_head"] == PRED == pins["source_head"]
          == task["predecessor_head"] == oprov["source_head"] == irev["predecessor_head"])
    check("I3 RESULT status PASS_INDEPENDENT_CAUSAL_SUPPORT_AUDIT_NO_NEW_KAPPA; not formal kernel",
          result["status"] == "PASS_INDEPENDENT_CAUSAL_SUPPORT_AUDIT_NO_NEW_KAPPA"
          and result["not_a_formal_kernel_proof"] is True)
    check("I4 PRIMARY: new_kappa_claimed False; formal_kernel/human_review False",
          primary["new_kappa_claimed"] is False
          and primary["formal_kernel_verification_claimed"] is False
          and primary["human_review_claimed"] is False)
    check("I5 VERIFY ready; TASK outcome completed",
          verify["status"] == "ready" and task["outcome"] == "completed")

    # Source pins
    bad_pin = []
    for e in pins["exports"]:
        p = pkg / e["path"]
        if sha(p) != e["sha256"] or p.stat().st_size != e["bytes"]:
            bad_pin.append(e["path"])
    check("P1 SOURCE-PINS exports match package bytes/sha256/size (5 files)",
          not bad_pin and len(pins["exports"]) == 5, bad_pin)

    # RESULT / PRIMARY inputs
    trace = result["actual_source_trace"]
    check("P2 RESULT word/geometry sha256 match pinned source/*.json.gz",
          trace["word_sha256"] == sha(pkg / "source/WORD.json.gz")
          == pins["exports"][0]["sha256"]
          and trace["geometry_sha256"] == sha(pkg / "source/GEOMETRY.json.gz")
          == pins["exports"][1]["sha256"])
    check("P3 RESULT.code_sha256 == source_check.py bytes",
          result["code_sha256"] == sha(pkg / "source_check.py")
          == primary["file_sha256"]["source_check.py"])
    check("P4 PRIMARY actual_input_sha256 matches source WORD/GEOMETRY",
          primary["actual_input_sha256"]["WORD.json.gz"] == sha(pkg / "source/WORD.json.gz")
          and primary["actual_input_sha256"]["GEOMETRY.json.gz"] == sha(pkg / "source/GEOMETRY.json.gz"))

    # Frame ranks / trace counts
    ops = trace["operation_counts"]
    check("T1 operation_counts ADD+COPY+ERASE+MOVE == records 38927",
          ops["ADD"] + ops["COPY"] + ops["ERASE"] + ops["MOVE"] == trace["records"] == 38927
          and ops == {"ADD": 36005, "COPY": 9, "ERASE": 9, "MOVE": 2904})
    check("T2 FRAME-RANK-CROSS: census sums to checked_frame_ranks 636; MOVE 2904; caps 111",
          sum(frc["rank_census"].values()) == frc["checked_frame_ranks"] == 636
          == len(trace["exact_rational_frame_ranks"])
          and frc["checked_MOVE_events"] == ops["MOVE"] == 2904
          and frc["checked_final_cap_equalities"] == trace["rank_seven_final_caps"] == 111
          and frc["status"] == "PASS"
          and frc["inputs"]["WORD"] == trace["word_sha256"]
          and frc["inputs"]["GEOMETRY"] == trace["geometry_sha256"])
    check("T3 postscatter records 1699 = 760+939; interval [24069,25768]",
          trace["postscatter_records"] == 1699
          == int(trace["postscatter_counts"]["0"]) + int(trace["postscatter_counts"]["1"])
          and trace["postscatter_interval"] == [24069, 25768])

    # Toy closure
    toy = result["finite_toy_closure"]
    check("T4 finite_toy_closure totals match closure sums (283 states / 1780 edges)",
          toy["total_states"] == sum(c["reachable_states"] for c in toy["closures"]) == 283
          and toy["total_edges"] == sum(c["edges"] for c in toy["closures"]) == 1780)

    # Oracle
    check("O1 ORACLE-RESULT PASS; 34208 states / 1632 walks / 317824 transitions",
          oracle["status"] == "PASS"
          and oracle["total_symbolic_states"] == 34208
          and oracle["total_closed_private_walks"] == 1632
          and oracle["total_transition_checks"] == 317824
          and oracle["ambient_dimension"] == 3
          and oracle["coefficient_field"] == "F2")
    check("O2 ORACLE-PROVENANCE PASS; normal/optimized byte-identical; source_pr 4200",
          oprov["status"] == "PASS"
          and oprov["normal_optimized_results_byte_identical"] is True
          and oprov["source_pr"] == 4200
          and oprov["source_main_commit"] == BASE)
    omap = {
        "NOTES.txt": "ORACLE-NOTES.txt",
        "RESULT.json": "ORACLE-RESULT.json",
        "RESULT-O.json": "ORACLE-RESULT.json",
        "independent_oracle.py": "independent_oracle.py",
    }
    bad_o = [k for k, rel in omap.items() if sha(pkg / rel) != oprov["file_sha256"][k]]
    check("O3 ORACLE-PROVENANCE file_sha256 matches packaged oracle artifacts", not bad_o, bad_o)

    # Independent review cross-link
    v = irev["verification"]
    check("R1 INDEPENDENT-REVIEW PASS; outstanding_findings empty",
          irev["status"] == "PASS_SCOPED_AGENT_REVIEW_AFTER_SCOPE_CORRECTION"
          and irev["outstanding_findings"] == [])
    check("R2 INDEPENDENT-REVIEW verification counts match RESULT/ORACLE/FRAME",
          v["oracle_symbolic_states"] == 34208
          and v["oracle_closed_walks"] == 1632
          and v["oracle_transitions"] == 317824
          and v["source_MOVE_events"] == 2904
          and v["source_ADD_events"] == 36005
          and v["source_frame_ranks"] == 636
          and v["source_cap_equalities"] == 111
          and v["source_records"] == 38927
          and v["tiny_closure_states"] == 283
          and v["tiny_closure_transitions"] == 1780)

    # REVIEW-FILE-MAP current equalities
    bad_eq = []
    for eq in rfm["current_byte_equalities"]:
        path = eq["package_path"]
        if not path or path.startswith("("):
            continue
        p = pkg / path
        if not p.exists() or sha(p) != eq["sha256"]:
            bad_eq.append(path)
    check("M1 REVIEW-FILE-MAP current_byte_equalities all match package bytes",
          not bad_eq, bad_eq)

    # PACKAGE-REVIEW seals — expected defect on README/SOURCES/VERIFY after root binding
    bad_pkg = []
    for rel, dig in package["file_sha256"].items():
        if sha(pkg / rel) != dig:
            bad_pkg.append(rel)
    check("S1 PACKAGE-REVIEW.file_sha256 matches all listed package files",
          not bad_pkg, bad_pkg)
    # Document the acknowledged pending root binding if S1 fails
    if bad_pkg:
        check("S1a mismatched seals are exactly README.md / SOURCES.json / VERIFY.json "
              "(scientific receipts still sealed; PACKAGE-REVIEW finding noted pending root bind)",
              set(bad_pkg) == {"README.md", "SOURCES.json", "VERIFY.json"})
        sci = [rel for rel in package["file_sha256"]
               if rel not in bad_pkg
               and not rel.endswith(".md") or rel in (
                   "PROOF.md", "RESULT.json", "ORACLE-RESULT.json", "ORACLE-PROVENANCE.json",
                   "PRIMARY-REVIEW.json", "INDEPENDENT-REVIEW.json", "SOURCE-PINS.json",
                   "independent_oracle.py", "source_check.py",
                   "source/WORD.json.gz", "source/GEOMETRY.json.gz",
                   "source/PR4200-PROOF.md.txt", "source/NOTICE.md.txt",
                   "source/RELATED-WORK.md.txt", "ORACLE-NOTES.txt", "TASK.json")]
        # simpler: all non-mismatched scientific cores match
        core = ["RESULT.json", "PROOF.md", "source_check.py", "independent_oracle.py",
                "ORACLE-RESULT.json", "ORACLE-PROVENANCE.json", "ORACLE-NOTES.txt",
                "PRIMARY-REVIEW.json", "INDEPENDENT-REVIEW.json", "SOURCE-PINS.json",
                "source/WORD.json.gz", "source/GEOMETRY.json.gz"]
        check("S1b scientific core seals in PACKAGE-REVIEW still match",
              all(sha(pkg / rel) == package["file_sha256"][rel] for rel in core))

    check("NEG1 claiming new κ contradicts PRIMARY.new_kappa_claimed / RESULT status tag",
          primary["new_kappa_claimed"] is False
          and "NO_NEW_KAPPA" in result["status"])
    check("NEG2 K/R / full endpoint explicitly out of FRAME-RANK-CROSS scope_excludes",
          "K/R coefficients" in frc["scope_excludes"]
          and "decoded scalar endpoint" in frc["scope_excludes"])

    failed = [name for name, ok, _ in results if not ok]
    print(f"\n{len(results) - len(failed)}/{len(results)} PASS", flush=True)
    if failed:
        print("FAILED:", ", ".join(failed), flush=True)
        return 1
    return 0


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--pkg", required=True)
    ap.add_argument("--expect-head", default=HEAD)
    args = ap.parse_args()
    check("HEAD pin recorded (API fetch)", args.expect_head == HEAD, args.expect_head)
    return run(Path(args.pkg))


if __name__ == "__main__":
    sys.exit(main())

James Reed (@rcfc-jreed), with substantial AI assistance (Claude, in Cursor).

@ToikY

ToikY commented Oct 20, 2026

Copy link
Copy Markdown
Contributor Author

Confirmed: the three README/SOURCES/VERIFY seals in PACKAGE-REVIEW.json were stale. Commit 304cfbd1910cf7041e6e33ee7db9b2045e74d9f6 corrects that single metadata file, preserves the complete original table and chronology in historical fields, and explicitly records the later final-root binding.

All20 current seals match; the other23 package files, including every scientific program/proof/input/receipt, are unchanged. Six registered commands pass from this exact committed Git-free head, all24 files remain unchanged, and source history passes. Hosted run39451598579 passed at this exact new head; decoded replay-job118729027097 logs contain all six registered checks and the unchanged-input step passed.

James Reed’s independent scoped check identified this packaging defect. Original scientific and predecessor credit remain unchanged, and the initial frozen head/archive are retained. This correction is not a new research result, supplier or κ claim.

@rcfc-jreed

Copy link
Copy Markdown
Contributor

Independent scoped audit of #4206 BIT COPY support propagation at 304cfbd1: 13/13

Code-disjoint checks; forks of source_check.py, independent_oracle.py, and frame_rank_cross.py (with sealed RESULT.json receipt). Head 304cfbd1910cf7041e6e33ee7db9b2045e74d9f6. CLAIM.

No packaging defect in SOURCES scope. All 7 digests match (package-local pins at HEAD).

What passes: sealed RESULT / ORACLE-RESULT / FRAME-RANK-CROSS PASS statuses; fresh forks regenerate matching PASS tokens for all three VERIFY entrypoints.

Not covered here: full scalar endpoint/K/R replay (explicitly out of package scope); Lean; new κ; merge/close.

Finite pin/seal/frame-rank replay ≠ formal proof.

James Reed (@rcfc-jreed), with substantial AI assistance (Claude, in Cursor).

@Mister-Meeseeks
Mister-Meeseeks merged commit e51227b into CrocSwap:main Oct 21, 2026
2 checks passed
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.

3 participants