Repair #314: authenticate D5/C6 mutant rejection reasons without editing scientific checkers

.github/workflows/c6-factorial-moment.yml

@@ -5,6 +5,8 @@ name: C6 factorial-moment candidate checks
     paths:
       - 'frontiers/c6_factorial_moment_20260929/**'
       - '.github/workflows/c6-factorial-moment.yml'
+      - 'tools/d5_c6_replay.py'
+      - 'tests/**'
   workflow_dispatch:
 
 permissions:
@@ -18,16 +20,23 @@ jobs:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1
         with:
           persist-credentials: false
+          fetch-depth: 0
       - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97
         with:
           python-version: '3.11.16'
+      - name: Full repository regression suite in both Python modes
+        run: |
+          set -euo pipefail
+          python -B -S -m unittest discover -s tests -v
+          python -B -O -S -m unittest discover -s tests -v
       - name: Verify sources, replay both modes, reject mutants
         env:
           PYTHONDONTWRITEBYTECODE: '1'
         run: |
           set -euo pipefail
           python -B -S - <<'PY'
-          import hashlib, json, pathlib, subprocess, sys
+          import hashlib, json, pathlib
+          from tools.d5_c6_replay import replay
           root = pathlib.Path('frontiers/c6_factorial_moment_20260929').resolve()
           manifest = json.loads((root / 'SOURCE_FILES.json').read_text())['files']
           present = sorted(p.name for p in root.iterdir())
@@ -45,15 +54,7 @@ jobs:
           expected = (root / 'RESULTS.json').read_bytes()
           mutants = ['wrong-elimination', 'touching-balls', 'no-log', 'short-remainder']
           for flags in (['-B', '-S'], ['-B', '-O', '-S']):
-              out = subprocess.run([sys.executable, *flags, script],
-                                   check=True, capture_output=True, timeout=300).stdout
-              if out != expected:
-                  raise RuntimeError('output differs from RESULTS.json under ' + ' '.join(flags))
-              for m in mutants:
-                  rc = subprocess.run([sys.executable, *flags, script, '--mutant', m],
-                                      capture_output=True, timeout=300).returncode
-                  if rc != 1:
-                      raise RuntimeError('mutant ' + m + ' not rejected')
+              replay(script, expected, mutants, 'factorial', flags, timeout=300)
           print(json.dumps({'scientific_effect': 'NONE', 'mathematical_acceptance': False}))
           PY
           git diff --exit-code

.github/workflows/c6-palm-route.yml

@@ -5,6 +5,8 @@ name: C6 Palm-route checks
     paths:
       - 'frontiers/c6_palm_route_20260929/**'
       - '.github/workflows/c6-palm-route.yml'
+      - 'tools/d5_c6_replay.py'
+      - 'tests/**'
   workflow_dispatch:
 
 permissions:
@@ -18,16 +20,23 @@ jobs:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1
         with:
           persist-credentials: false
+          fetch-depth: 0
       - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97
         with:
           python-version: '3.11.16'
+      - name: Full repository regression suite in both Python modes
+        run: |
+          set -euo pipefail
+          python -B -S -m unittest discover -s tests -v
+          python -B -O -S -m unittest discover -s tests -v
       - name: Verify sources, replay both modes, reject mutants
         env:
           PYTHONDONTWRITEBYTECODE: '1'
         run: |
           set -euo pipefail
           python -B -S - <<'PY'
-          import hashlib, json, pathlib, subprocess, sys
+          import hashlib, json, pathlib
+          from tools.d5_c6_replay import replay
           root = pathlib.Path('frontiers/c6_palm_route_20260929').resolve()
           manifest = json.loads((root / 'SOURCE_FILES.json').read_text())['files']
           present = sorted(p.name for p in root.iterdir())
@@ -46,15 +55,7 @@ jobs:
           mutants = ['no-slab', 'weak-zeta', 'lambda-too-large', 'series-ratio', 'factorial-power', 'schur-upper',
                      'forget-mark-power']
           for flags in (['-B', '-S'], ['-B', '-O', '-S']):
-              out = subprocess.run([sys.executable, *flags, script],
-                                   check=True, capture_output=True, timeout=600).stdout
-              if out != expected:
-                  raise RuntimeError('output differs from RESULTS.json under ' + ' '.join(flags))
-              for m in mutants:
-                  rc = subprocess.run([sys.executable, *flags, script, '--mutant', m],
-                                      capture_output=True, timeout=600).returncode
-                  if rc != 1:
-                      raise RuntimeError('mutant ' + m + ' not rejected')
+              replay(script, expected, mutants, 'palm', flags, timeout=600)
           print(json.dumps({'scientific_effect': 'NONE', 'mathematical_acceptance': False}))
           PY
           git diff --exit-code

.github/workflows/d5-dimension-lift.yml

@@ -5,6 +5,8 @@ name: D5 dimension-lift checks
     paths:
       - 'frontiers/d5_dimension_lift_20260929/**'
       - '.github/workflows/d5-dimension-lift.yml'
+      - 'tools/d5_c6_replay.py'
+      - 'tests/**'
   workflow_dispatch:
 
 permissions:
@@ -18,16 +20,23 @@ jobs:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1
         with:
           persist-credentials: false
+          fetch-depth: 0
       - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97
         with:
           python-version: '3.11.16'
+      - name: Full repository regression suite in both Python modes
+        run: |
+          set -euo pipefail
+          python -B -S -m unittest discover -s tests -v
+          python -B -O -S -m unittest discover -s tests -v
       - name: Verify sources, replay both modes, reject mutants
         env:
           PYTHONDONTWRITEBYTECODE: '1'
         run: |
           set -euo pipefail
           python -B -S - <<'PY'
-          import hashlib, json, pathlib, subprocess, sys
+          import hashlib, json, pathlib
+          from tools.d5_c6_replay import replay
           root = pathlib.Path('frontiers/d5_dimension_lift_20260929').resolve()
           manifest = json.loads((root / 'SOURCE_FILES.json').read_text())['files']
           present = sorted(p.name for p in root.iterdir())
@@ -47,15 +56,7 @@ jobs:
                      'row-operation-third', 'euler-coefficient', 'adjugate-sign', 'one-soft-direction',
                      'flat-shell', 'weight-one-less-power']
           for flags in (['-B', '-S'], ['-B', '-O', '-S']):
-              out = subprocess.run([sys.executable, *flags, script],
-                                   check=True, capture_output=True, timeout=600).stdout
-              if out != expected:
-                  raise RuntimeError('output differs from RESULTS.json under ' + ' '.join(flags))
-              for m in mutants:
-                  rc = subprocess.run([sys.executable, *flags, script, '--mutant', m],
-                                      capture_output=True, timeout=600).returncode
-                  if rc != 1:
-                      raise RuntimeError('mutant ' + m + ' not rejected')
+              replay(script, expected, mutants, 'd5', flags, timeout=600)
           print(json.dumps({'scientific_effect': 'NONE', 'mathematical_acceptance': False}))
           PY
           git diff --exit-code

reviews/d5_c6_workflow_repair_20261005/REPLAY_SCOPE.md

@@ -0,0 +1,110 @@
+# D5/C6 named-rejection workflow repair
+
+Dylan Roy - delegated AI engineering work. Actual author: OpenAI / GPT-6 Astra
+Pro, `lm006-continuation-wrapper-repair-20261005`. Scientific effect **NONE**;
+shared-account/provider organizational-independence credit **0**. This addresses
+Math-#314, under pickup 6007256016 and contract clarification 6007291547. It does
+not modify the scientific checkers, their proof packets, manifests or outputs.
+Current execution and review dispositions belong to the PR's source-bound record.
+
+## Why an adapter is necessary
+
+The original three workflows accepted any mutant process exiting 1, including
+an unrelated RuntimeError; they also ignored baseline stderr. D5 and C6 Palm
+intentionally raise uncaught AssertionError through their own `require` function.
+Factorial instead emits JSON with an exact failed-check set. Requiring empty
+stderr and JSON directly from every unchanged original CLI would reject genuine
+D5/Palm controls. The repair translates only their authentic assertion carrier.
+
+`tools/d5_c6_replay.py` runs the full original script in a fresh child process.
+For D5/Palm it accepts only an exact AssertionError class, argument message,
+checker filename and final `(check_group, require)` traceback pair. It then emits
+an identified JSON rejection with empty stderr. The parent also requires the
+adapter's exact serialized bytes, including its single final newline. Output preceding the exception
+is not suppressed and therefore fails the parent. All other exceptions propagate
+as failures. The parent requires exit 1, empty stderr and an exact typed JSON
+record. Normal completion is not rejection.
+
+Factorial retains its original JSON output. The parent requires exactly its
+baseline object/scope/check schema, exactly one mapped false check, all remaining
+checks true, and `passed` false. Duplicate keys, nonfinite numbers, floats,
+coerced Boolean/integer values, missing/extra fields and trailing data fail.
+Every baseline must exit 0, reproduce the frozen RESULTS.json bytes and have
+empty stderr. The complete ordered mutant inventory is mandatory in both modes.
+
+The original workflow membership/size/hash checks and final tracked Git-diff
+check are retained. This is not a general sandbox or an authenticity guarantee
+against a malicious replacement of an entire reviewed repository. A process
+protocol does not prove the mathematical claim tested by a finite checker.
+
+## Authentic source cut and measured contracts
+
+Repository `d6g8k5htny-coder/Math-`, source cut
+`2f8b6f372be383d752e9dd30d38234faa977243c`.
+
+| Family | Unchanged checker path | Git blob |
+|---|---|---|
+| D5 | `frontiers/d5_dimension_lift_20260929/lift_exact_check.py` | `460efe46126bb1b7b45954a7a8e0ef2f8e93c739` |
+| Palm | `frontiers/c6_palm_route_20260929/palm_exact_check.py` | `6b0c73baeeb2ac566160e23cc405daec07eb7585` |
+| Factorial | `frontiers/c6_factorial_moment_20260929/c6_check.py` | `0adaf6382e89af8ce2537c994013d0c5e05b011d` |
+
+All three original baselines and every mutant were actually executed normally
+and under optimization: 48 native processes total. The 10 D5 and 7 Palm messages
+and originating groups agree across modes and are recorded in the helper and
+independently in the test fixtures. The 4 factorial failures are respectively
+ELIMINATION, PACKING, RADIAL and REMAINDER. The unmodified baseline identities are:
+
+| Family | Bytes | RESULTS.json SHA-256 |
+|---|---:|---|
+| D5 | 5184 | `ae9ca864ff9db1f239910fa7c349c297aa1b5ffbd9863786ca20a61a6c1814e5` |
+| Palm | 2775 | `7812de453ff1efd77ffc2431874243713615ce4fcc0e55f2beae991e24f6f6bc` |
+| Factorial | 285 | `82ef312659552ef7fd783e188961df62088764ef79f575e3b29720585f32109f` |
+
+The predecessor workflow blobs, used for red-phase controls, are respectively
+`06a9c445bc8ae7b4af78bdb1ddee41f533e3a7db`,
+`9dba863a2c1c155d8fa267da2de9cc91d8bce186`, and
+`6d3a141555d4ed98f9ab9c1e6ff36e00683282da`.
+
+## Regression and hosted execution boundary
+
+```sh
+python -B -S -m unittest discover -s tests -p test_d5_c6_workflows.py -v
+python -B -O -S -m unittest discover -s tests -p test_d5_c6_workflows.py -v
+```
+
+The suite has 16 methods. Its workflow tests use 238 real committed temporary
+Git repositories per outer mode, with hash-consistent synthetic checker/output
+fixtures. They execute the entire original or repaired verification shell step,
+including its membership/hash checks and final Git-diff command. No subprocess
+is mocked. Every one of the 21 named mutant stages is targeted separately in
+both child modes. Additional helper tests verify baseline/mutant timeout stages,
+unknown-label exits, inventory completeness and strict JSON types.
+
+The fixture scripts are not scientific evidence. They deliberately use small
+synthetic baselines where no numerical reproduction claim is being made. A
+predecessor replay can use the current test file with `WORKFLOW_SOURCE_ROOT`
+pointing to a checkout containing the original three workflow blobs; the helper
+unit tests still use the current helper while those workflow fixtures execute
+the original shell code. Expected failures demonstrate the old admission defect.
+
+All three workflows now run full repository `tests/` discovery in both outer
+Python modes, with explicit fail-fast shell settings. Full-history checkout is
+needed by the existing historical proof-reachability tests. Trigger paths cover
+the helper and `tests/**`; existing action SHAs, Python 3.11.16, original process
+timeouts, job timeouts, permissions and mathematical packet paths are unchanged.
+
+Local source copies were authenticated before execution. Complete local GitHub
+checkout was unavailable, so local full-repository discovery and original full
+packet-membership verification are not claimed. Dedicated hosted jobs must run
+the full repository tests and actual complete scientific packets at the tested
+candidate. A general green formal/downstream check alone does not establish that
+these three dedicated workflows executed. A later source/base requires its own
+applicable checks; old receipts must not be rebound.
+
+## Nonclaims
+
+No D5/C6 theorem, Gaussian regression step, continuum argument, Palm law, global
+closure, formal alignment, Lean source, scientific register or prize is changed.
+Nonauthor technical review and eligible nonauthor/nonreader integration remain
+separate from this author-side repair and its tests. The exact historical proof
+and checker bytes remain available without edits or retrospective acceptance.