Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions .gitattributes
Original file line number Diff line number Diff line change
Expand Up @@ -6,3 +6,5 @@ cbz/evidence/receipt-composition-20261029/*.log whitespace=-blank-at-eol
cbz/evidence/verifier-audit-db642-20261029/**/*.log whitespace=-blank-at-eol,-blank-at-eof
cbz/evidence/receipt-source-custody-20261029/**/*.log whitespace=-blank-at-eol,-blank-at-eof
cbz/evidence/receipt-source-custody-20261029/red/dirty-input.patch whitespace=-blank-at-eol,-blank-at-eof
# Preserve exact failure-log bytes, including their historical trailing spaces.
cbz/evidence/lean-option-member-20261110/*.log whitespace=-blank-at-eol
4 changes: 2 additions & 2 deletions .github/workflows/cbz-auditor-origin.yml
Original file line number Diff line number Diff line change
Expand Up @@ -60,12 +60,12 @@ jobs:
if CBZ_AUDITOR_EVIDENCE_DIR="$evidence/$mode/census" python "${flags[@]}" -m unittest tests.test_cbz_layer_a_scan tests.test_cbz_auditor_hardening tests.test_cbz_layer_a_scan_mutants tests.test_cbz_a3_float_census -v > "$evidence/$mode/census.log" 2>&1; then status=0; else status=$?; failed=1; fi
printf '%s\n' "$status" > "$evidence/$mode/census.exit" || failed=1
grep -Eq '^Ran 22 tests in ' "$evidence/$mode/origin.log" || failed=1
grep -Eq '^Ran 83 tests in ' "$evidence/$mode/census.log" || failed=1
grep -Eq '^Ran 85 tests in ' "$evidence/$mode/census.log" || failed=1
grep -Fxq 'OK' "$evidence/$mode/origin.log" || failed=1
grep -Fxq 'OK' "$evidence/$mode/census.log" || failed=1
if grep -Fq ' ... skipped ' "$evidence/$mode/origin.log" "$evidence/$mode/census.log"; then failed=1; fi
[ "$(find "$evidence/$mode/adapter" -name '*.json' -type f | wc -l)" -eq 10 ] || failed=1
[ "$(find "$evidence/$mode/census" -maxdepth 1 -name '*.json' -type f | wc -l)" -eq 202 ] || failed=1
[ "$(find "$evidence/$mode/census" -maxdepth 1 -name '*.json' -type f | wc -l)" -eq 204 ] || failed=1
done
sha256sum -c "$evidence/source-sha256.txt" > "$evidence/source-readback.log" 2>&1 || failed=1
exit "$failed"
Expand Down
10 changes: 10 additions & 0 deletions BASELINE.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,15 @@
# CBZ baseline ledger

## Lean option-member candidate — 10 November 2026

The previous Main scanner missed `debug.skipKernelTC.set` in both retained
negative kernel fixtures. The candidate reports the option name even when a
dotted member follows it; red observations and the first failed test-harness
attempt remain preserved. The resulting bounded source census and its limits
are in [STATUS](STATUS.md#lean-option-member-scanner-correction--10-november-2026).
Exact-head hosted qualification remains required. This is not final G0/G1,
a proof of fixture isolation, or a CBZ result.

## Auditor-origin candidate — 9 November 2026

The observed published cut has Main `05f2289478a0ee9492adc1eb501466e0f66866df`,
Expand Down
20 changes: 20 additions & 0 deletions DECISIONS.md
Original file line number Diff line number Diff line change
Expand Up @@ -196,3 +196,23 @@ Append-only. Historical source outcomes remain in their original records.
Preserve the red output in the evidence record and fail the job on either
metadata-write error; test both guard removals in each Python mode. This
controls evidence completeness, not mathematical truth or scientific status.

## 2026-11-10 — retain Lean option-member false negatives

- Preserve the real `BadPublic.lean` and `BadPrivate.lean` fixture results at
Main `c32e6f27d318e25708939105285366a7a0daf971`: the scoped scanner returned
zero findings despite `debug.skipKernelTC.set` in each. Record their native
paths, original scanner blob and red outcomes in
`cbz/evidence/lean-option-member-20261110/`. This finding changes no
mathematical statement, bound, attribution, or historical kernel result.
- Conservatively report `skipKernelTC` in nonfinal dotted identifier components
as well as terminal ones. Keep alias resolution and semantic option state
outside the scanner's claimed scope. Retain the initial failed post-fix test
parser and the workflow-population red results. The source-specific local
campaign ran 85 tests and 102 selected baseline/mutant pairs per mode, with
the documented full-child inner skips preserved; hosted exact-head
qualification remains required. Do not award full Layer A or final G0/G1.
- Workspace #63 integrated the earlier Main pointer after fresh sequential
default-ref comparisons and normal required checks. Any later Main merge
requires a new Workspace pin comparison; no atomic cross-repository snapshot
or scientific acceptance is inferred.
10 changes: 10 additions & 0 deletions GOVERNANCE.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,15 @@
# CBZ-0 governance

## Lean kernel-option spelling boundary — 10 November 2026

The Lean lexical scanner reports `skipKernelTC` as an A2 violation even when
it is a nonfinal dotted identifier component, including an escaped component.
This deliberately conservative rule also reports read-only member names and
does not resolve aliases, evaluate settings, or certify kernel safety. Run and
verify it using the registered [auditor commands](TOOLING.md#auditor-origin-controls);
retain negative fixtures and actual errors. The pre-correction false negatives
are recorded in [the evidence](cbz/evidence/lean-option-member-20261110/README.md).

## Auditor assertion-origin boundary — 9 November 2026

A mutation kill requires a clean designated baseline and an actual failure
Expand Down
10 changes: 10 additions & 0 deletions PROOF_INDEX.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,15 @@
# CBZ proof and quarantine index

## Lean option-member finding — 10 November 2026

The retained `BadPublic.lean` and `BadPrivate.lean` fixtures deliberately
construct malformed declarations using `debug.skipKernelTC.set`. A prior
lexical scan missed both because the option was not the last dotted name
component. The source correction and its [adverse evidence](cbz/evidence/lean-option-member-20261110/README.md)
change no theorem or proof disposition. These fixtures are negative test
data, not closed proofs; lexical detection does not prove transitive isolation
or replace a kernel and axiom audit.

## Auditor-origin boundary — 9 November 2026

The Main and Workspace engineering changes and the auditor-grader repair do
Expand Down
28 changes: 28 additions & 0 deletions STATUS.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,33 @@
# CBZ-0 execution status

## Lean option-member scanner correction — 10 November 2026

At published Main `c32e6f27d318e25708939105285366a7a0daf971`, the scoped
Lean scanner returned no findings and exit 0 on the two retained
`BadPublic.lean` and `BadPrivate.lean` negative fixtures, although each uses
`debug.skipKernelTC.set`. Its previous Main census therefore missed those two
fixture-local A2 signals; that observation and the failed regression controls
remain in [the evidence record](cbz/evidence/lean-option-member-20261110/README.md).
The fixtures are negative test data, not closed mathematical modules, and no
transitive isolation or federation-wide axiom clearance is inferred.

This candidate reports `skipKernelTC` in any dotted identifier component,
including escaped components, so a method suffix cannot hide the option name.
It conservatively reports uses such as `.get` as well as `.set`; it does not
resolve aliases or determine whether a use enables the option. Main source at
`c32e6f27` then yields three violations in the retained negative fixtures and
one `unsafe` review token across 18 Lean blobs; the 1,974-blob Math source
`9bdfc54b` yields no listed tokens. These are lexical observations, not Lean
compilation, kernel/axiom proofs, scientific acceptance, or full Layer A audit.
The candidate requires exact-head hosted testing and readback before admission.

Workspace #63 merged as `ca321af7df4b63a832b0a13c5929eab99c84a25a`, pinning
Main `c32e6f27` and Math `9bdfc54b`; its hosted contract ran 254 tests per
mode without skips and its pin job checked eight current pins and four
canonical inputs. Those observations were sequential, not a final G7 cut. A
later Main merge requires another guarded pin refresh. No CBZ-0 or terminal
Partial is declared.

## Auditor-origin repair and published pin cut — 9 November 2026

Main #791 merged as `05f2289478a0ee9492adc1eb501466e0f66866df`; its hosted
Expand Down
9 changes: 8 additions & 1 deletion TOOLING.md
Original file line number Diff line number Diff line change
Expand Up @@ -28,12 +28,19 @@ exercise clean subjects, actual selected-body failures, direct and nested
subject failures, changed source, missing provenance, forged text and
failed evidence writes. Five adapter source mutations have designated
clean-baseline and actual-failure controls; the two existing source campaigns
contain 74 Lean and 27 A3 mutations. The dedicated workflow preserves original
contain 75 Lean and 27 A3 mutations. The dedicated workflow preserves original
child reports and source snapshots in both modes. Its actual shell guards are
also exercised with synthetic child outcomes and seven guard-removal controls. Actual exact-head hosted
execution and independent readback are required before qualification; historical
pilot observations do not establish final G0, G1 or scientific acceptance.

The Lean scanner also treats a `skipKernelTC` component with a following dotted
member as an A2 signal, including escaped components. This is conservative
spelling detection, not alias resolution or a determination that a setting was
enabled. Two actual negative fixtures exposed the earlier miss; the original
and repaired outcomes are preserved in
[the evidence record](cbz/evidence/lean-option-member-20261110/README.md).


**State: component registry; the final federation toolchain is incomplete.**

Expand Down
71 changes: 71 additions & 0 deletions cbz/evidence/lean-option-member-20261110/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,71 @@
# Lean kernel-option member detection: bounded adverse record

At Main commit `c32e6f27d318e25708939105285366a7a0daf971`, scanner blob
`0d53463de621b821519b71559db2834eef3a8d94` (SHA-256
`bf797f33d5df7a8e695f62e6a2d70dca05ec99956ac5c33aa0347164eac48410`)
returned exit 0 with zero findings on two tracked negative fixture files:
`kernel_replay/tests/fixtures/BadPublic.lean` and `BadPrivate.lean`. Both
contain the live token `debug.skipKernelTC.set` at line 13, column 41. Their
respective SHA-256 values are
`9b636915c4c868091d72b6c72337b6c255cab8ede30b8b394b43389e1fdd95d5`
and `fa95b63ae8037158cdc643a6542ed28dc1adde86a529a535cfc0c713033ec429`.
The exact original scanner is recoverable from that native Git blob; the
normal and optimized command outputs and exit statuses are retained here.

The root cause is that the lexer emits the entire dotted identifier, while
classification checked only its last component, `set`. The added control
exercises real fixture bytes, CLI exit and finding counts, literal and escaped
option components, a `.get` member, comments, strings, a different identifier,
and an escaped identifier containing a dot. Two selected controls failed in
both modes against the original scanner (`lean-option-member-red-*.log`),
with six assertion failures each. The first post-fix runs detected the
fixtures but errored because the new test tried to parse the CLI's findings
and summary as one JSON document; those errors are preserved in
`lean-option-member-green-*.log`. The test was corrected to parse the final
summary line, and both selected tests then passed in both modes
(`lean-option-member-green2-*.log`).

The scanner source SHA-256 after the fix is
`dd8ee1c3dcf1f263c77238d35d639ed000bdc1dd7f382512219ac281d80e4a17`.
Its conservative component check reports even read-only member spelling; it
does not resolve aliases, inspect generated Lean declarations, decide whether
an option is enabled, or establish full Layer A coverage. The historical
fixtures and their statements are untouched and remain negative test data;
this lexical check does not prove their transitive isolation.

Using that working scanner, Git-tree scans of Main `c32e6f27` report three
fixture violations and one `unsafe` review finding in 18 Lean blobs in each
mode; the report SHA-256 is
`53f6ed8b22ba5e2d1051d0b9a65f3304f9f2e2305c17c691108ab8b39e9823e9`.
Math `9bdfc54b76173b57912503bfcf901a5f9d43b09c` reports no listed tokens
in 1,974 Lean blobs in each mode; report SHA-256
`f8bc30f2df38fcf1b301b774c4c555df3540829348a1e97f1536a76dbb05100f`.
These byte-bound scans do not compile Lean or prove axiom, kernel, isolation,
scientific, or whole-federation status.

The local source-specific census campaign ran 85 top-level tests in each
Python 3.12.14 mode without top-level skips. Its independently read-back
reports contain 102 clean selected baselines and 102 authorized selected-body
mutant failures per mode: 75 Lean and 27 Python census mutants. The new mutant
removes the nonfinal option-component check; its selected test has both a
clean baseline and an actual authorized failure. The retained full-child
records include the explicitly checked two and four recursion-guard skips and
two deliberate syntax errors in each mode. Readback logs and top-level outputs
are included; raw child reports and source snapshots are retained separately
in the supplemental evidence archive. This is scoped test evidence, not final
G0 or G1.

The workflow's expected population is now 85 tests and 204 census reports per
mode. Its synthetic guard fixture first failed with the previous counts
(`lean-option-member-workflow-red-*.log`); after the count update, both workflow
guard tests passed in each mode (`lean-option-workflow-green-*.log`). Those
shell-guard tests use synthetic child outcomes; they are not additional real
scanner mutation executions. Exact clean-commit hosted qualification and
artifact readback remain required.

The first staged `git diff --check` reported trailing spaces in the captured
raw unittest failures. The command sequence still created commit `fd8b9e5c`;
that observed failed check and exit 2 are retained as `diff-check-first.*`.
The follow-up applies the repository's existing evidence-log whitespace
attribute narrowly to this directory, preserving the original log bytes rather
than editing the recorded failures. The complete diff must be checked again.
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
2
12 changes: 12 additions & 0 deletions cbz/evidence/lean-option-member-20261110/diff-check-first.log
Original file line number Diff line number Diff line change
@@ -0,0 +1,12 @@
cbz/evidence/lean-option-member-20261110/lean-option-member-red-normal.log:1: trailing whitespace.
+test_kernel_bypass_option_members_are_not_lost_in_dotted_names (tests.test_cbz_layer_a_scan.ViolationTokens.test_kernel_bypass_option_members_are_not_lost_in_dotted_names) ...
cbz/evidence/lean-option-member-20261110/lean-option-member-red-normal.log:5: trailing whitespace.
+test_real_kernel_replay_negative_fixtures_fail_cli_scan (tests.test_cbz_layer_a_scan.ViolationTokens.test_real_kernel_replay_negative_fixtures_fail_cli_scan) ...
cbz/evidence/lean-option-member-20261110/lean-option-member-red-optimized.log:1: trailing whitespace.
+test_kernel_bypass_option_members_are_not_lost_in_dotted_names (tests.test_cbz_layer_a_scan.ViolationTokens.test_kernel_bypass_option_members_are_not_lost_in_dotted_names) ...
cbz/evidence/lean-option-member-20261110/lean-option-member-red-optimized.log:5: trailing whitespace.
+test_real_kernel_replay_negative_fixtures_fail_cli_scan (tests.test_cbz_layer_a_scan.ViolationTokens.test_real_kernel_replay_negative_fixtures_fail_cli_scan) ...
cbz/evidence/lean-option-member-20261110/lean-option-member-workflow-red-normal.log:9: trailing whitespace.
+AssertionError: 1 != 0 :
cbz/evidence/lean-option-member-20261110/lean-option-member-workflow-red-optimized.log:9: trailing whitespace.
+AssertionError: 1 != 0 :
Loading
Loading