Repair P08 checker-input provenance and verify evidence reuse

2026-10-14 00:24:45 +09:00
parent 9d5ec10676
commit b3d329388b
10 changed files with 500 additions and 11 deletions
@@ -1,7 +1,7 @@
 {
-  "candidate": "d1c7ce27d21a77be87081e12d20e7c8dd6ffa353",
+  "candidate": "bf559632ddd938085702529d9dd6b7c878f4ebc7",
  "original_unit_completed": false,
-  "next": "actual independent Opus review then integration/current applicable final verification",
+  "next": "actual independent Opus then integration/current applicable final verification",
  "fixed_material_limit": null,
  "file_count_limit": null,
  "original_30_88_unchanged": true,
@@ -22,11 +22,21 @@
  "independent_review_navigation": "OPUS_EVIDENCE_NAVIGATION.json",
  "prior_candidate_evidence_preserved_outside_current_build": true,
  "registered_final_regression_files_whole_verified": 5612,
-  "registered_validation": "integrity verified, deficits0, original acceptance pending",
+  "registered_validation": "bf559632 integrity verified, deficits0, original acceptance pending",
  "independent_review": {
    "state": "running",
-    "session": 14381,
+    "session": 85138,
    "model": "claude-opus-5-5",
-    "source_access": "Read/Grep/Glob frozen checkout + complete SHA-bound validation material"
-  }
+    "earlier_review": "changes requested; retained in audit-r07"
+  },
+  "repair": {
+    "state": "submitted",
+    "session": 39822,
+    "attempt": 7,
+    "cap": null,
+    "model": "gpt-6.1-sol",
+    "scope": "restore exact r04 data and implementation local test provenance; no tested source changes authorized",
+    "local_checker": "4 tests55.197s OK, actual transcript verified"
+  },
+  "reused_native_os_regression_provenance": "d1c7ce27, exact all1316 selected source/spec/input/marker equality; original receipts not relabelled"
 }
@@ -59,5 +59,9 @@
    "input": "p08-evidence/regression/runner-input.json",
    "ended": "p08-evidence/regression/runner-ended.json",
    "proof": "p08-evidence/regression/proof.json"
-  }
+  },
+  "review_candidate": "bf559632ddd938085702529d9dd6b7c878f4ebc7",
+  "actual_local_checker_run": "actual-local-checker-032c5c/stdout.log",
+  "reuse_applicability": "EXACT_EVIDENCE_REUSE_APPLICABILITY.json",
+  "repairs_for_this_review": "r04 restored byte-exact; checker4 real55.197s OK; product/native/GDScript/Python sources unchanged; earlier full source map/inputs/spec/markers exactly equal"
 }
@@ -0,0 +1,14 @@
+{
+  "run": "032c5cffc4ed468ba61ef794ac1a8d57",
+  "actual_transcript": [
+    {
+      "path": "/Users/korizatou114/Library/Application Support/agent-control/the-tree-p08-02/runs/032c5cffc4ed468ba61ef794ac1a8d57/stdout.log",
+      "sha256": "26b433e01d741d09aaa209b554e08c3f99d6cd16fccebaf24c4d4488739eb01b",
+      "snippet": "em\":{\"id\":\"item_5\",\"type\":\"command_execution\",\"command\":\"/bin/zsh -lc 'python3 -B tools/test_p08_migration_evidence.py'\",\"aggregated_output\":\"....\\n----------------------------------------------------------------------\\nRan 4 tests in 55.197s\\n\\nOK\\n\",\"exit_code\":0,\"status\":\"completed\"}}\n{\"type\":\"item.started\",\"item\":{\"id\":\"item_8\",\"type\":\"command_execution\",\"command\":\"/bin/zsh -lc \\\"python3 -B -c 'import json,pathlib,hashlib; p=pathlib.Path(\\\\\\\"docs/plans/original-completion-2026-10-02/execution/p08-02/implementation.json\\\\\\\"); d=json.loads(p.read_bytes()); paths"
+    }
+  ],
+  "candidate": "bf559632ddd938085702529d9dd6b7c878f4ebc7",
+  "test_count": 4,
+  "seconds": 55.197,
+  "formal_acceptance": false
+}
@@ -0,0 +1,29 @@
+{
+  "review": {
+    "task_id": "P08-02",
+    "commit": "d1c7ce27d21a77be87081e12d20e7c8dd6ffa353",
+    "status": "changes_requested",
+    "summary": "The evidence Root collected for the six facets is substantial, and I checked it independently. The p08-evidence validator ran check_p08_migration_evidence.py --verify-only and reported integrity \"verified\" with no deficits. The index is bound to commit d1c7ce2 with collection_error=null. There are four native ON/OFF positive and refusal pairs, each with separate writer and reader receipts. The normal-menu OS proof uses real native mouse input through the production PopupMenu id_pressed handler and covers preview, cancel, confirm and changed-source rejection. A full final regression is recorded (174 suites per Root's notes, all PASS).\n\nAt the submitted commit, though, a required implementation output no longer runs. tools/test_p08_migration_evidence.py reads R04['actual_readonly_diagnostics'] when the module loads. This commit cut ROOT_REPAIR_INPUT-r04.json down to a short summary, and that key is gone, so the checker test suite fails with KeyError before any test executes. The spec requires meaningful whole-checker positive and negative tests (duplicate actor, orphan occupancy, patient/cargo/EV loss, partial batch, overwritten original/backup, unequal cross-PID state, missing ON, wrong/stale source, fake OS signal). Those can no longer be run, which in effect disables the tests.\n\nOn top of that, implementation.json reports \"4 tests in 61.547s, PASS\" for a run at an earlier commit and presents it as the current local check. That result does not hold for the submitted sources. Because of this, I can't call the P08-02 acceptance satisfied, and the status is changes_requested.",
+    "acceptance_ids": [
+      "P08-02"
+    ],
+    "findings": [
+      {
+        "severity": "high",
+        "message": "tools/test_p08_migration_evidence.py:17-20 loads docs/plans/.../p08-02/ROOT_REPAIR_INPUT-r04.json and indexes R04['actual_readonly_diagnostics'] (plus CARGO_ENDPOINT_READONLY_DIAGNOSTIC.json and ACTUAL_WORLD_REVISION_DELTAS.json) when the module loads. This commit shortened that file to a summary with no 'actual_readonly_diagnostics' key; a Grep of the checkout finds the key only in the test file. As a result the module fails with KeyError on import, and none of the required checker positives or negatives can run at the submitted SHA. That breaks the required output and in effect disables the tests (rule: do not delete or disable tests). Fix: put the fixture data the test needs back into an allowed path, ideally a dedicated file under tests/fixtures/p08-02/, and point the test there instead of the Root handoff file. Then re-run python3 -B tools/test_p08_migration_evidence.py at the final SHA."
+      },
+      {
+        "severity": "medium",
+        "message": "implementation.json local_checks.checker_tests reports 'Recorded Astra run 7cd74f00…: 4 tests in 61.547s, PASS' and lists tools/test_p08_migration_evidence.py with its current SHA. That run happened against an earlier version of ROOT_REPAIR_INPUT-r04.json, and the test cannot run against the submitted one. The current-state test claim is therefore stale. It should either be re-recorded at the final SHA or explicitly marked as not applicable to the current sources."
+      },
+      {
+        "severity": "low",
+        "message": "tests/s1/test_p08_combined_migration.gd: when a role fails, `break` leaves only the writer/reader loop, so the remaining operations/refusal combinations still run on top of a failed state. The final marker still reports failures correctly, so this is a minor diagnostic issue rather than a false pass."
+      }
+    ]
+  },
+  "actual_record": "/Users/korizatou114/Library/Application Support/agent-control/the-tree-p08-02/runs/da13e081d6c74d55bcb70a69f6f5fbab/result.json",
+  "sha256": "84cab2ccf76fe5992ff49f79a84a62b54aa1a53b02b1cd7b482a713be9f310ed",
+  "root_import_reproduction": "KeyError actual_readonly_diagnostics; no tests executed",
+  "completion": false
+}
@@ -0,0 +1,7 @@
+{
+  "action": "run",
+  "task_id": "P08-02",
+  "reason": "P08-02 is in state repair_ready. Repair 7 is authorized, and the review gave two concrete findings. High: at d1c7ce2, importing tools/test_p08_migration_evidence.py fails with KeyError 'actual_readonly_diagnostics' because ROOT_REPAIR_INPUT-r04.json was cut down to a summary. Medium: implementation.json reports a stale '4 tests 61.547s PASS' result. The repair is limited to two documents.\n\nInstructions for Codex:\n(1) Restore docs/plans/original-completion-2026-10-02/execution/p08-02/ROOT_REPAIR_INPUT-r04.json byte-exact from the preserved_raw_utf8 field in ROOT_REPAIR_INPUT-r07.json. The original was saved at git 1545bd42. Write the bytes with no reformatting or trailing-newline change. Confirm the raw SHA256 is 23ef44a0775ca53f1f67d6d64fcc96cd33eae935478768e9ce9431796eaa12ef. Do not modify r07.\n(2) Freeze all source edits. Then run `python3 -B tools/test_p08_migration_evidence.py` once against the final files and confirm all 4 tests actually run (ast.parse is allowed too).\n(3) Rewrite implementation.json local_checks.checker_tests using only that run's real count, time and result. Record SHA256 values for the final test file, checker file and r04. Remove the old Astra 7cd74f00 result from the current-state claim, or label it clearly as history for an earlier input.\n(4) Leave unchanged: product/runtime code, native fixtures under tests/fixtures/p08-02, the GDScript suite, the checker and test Python sources, completion_marker_contracts, the 6 conditions and the 30/88 counts. Do not relocate fixtures and do not fix the low-severity `break` diagnostic.\n(5) If any test fails, report the actual output and the specific cause. Do not change tests to make them pass.\n(6) In the submitted response, state these as pending: reuse of native 8, OS and final 174 regression evidence; independent review; and integration. Root reuses earlier evidence only after confirming exact matches of source_map, input, config and environment. If the restored r04 is part of the selected source map, Root must decide whether to re-collect. The worker must not decide this or fill in any unconfirmed numbers.\n(7) Return a genuine submitted status. Do not commit, do no branch operations, and do not start Runner, native or full regression.",
+  "codex_model": "gpt-6.1-sol",
+  "codex_model_reason": "This is a narrow repair. It restores one file byte-exact from a verbatim copy with a known SHA256, runs one existing stdlib test file, and updates test provenance in one JSON file. The fix is already pinned down: the review's root cause, Root's reproduction and the exact bytes are all supplied. It needs no product, checker or fixture logic, no design judgment and no native work. The main risks are byte fidelity and honest reporting, and SHA verification and a real test run handle both. Astra would cost more time and usage for no gain in quality, so the default gpt-6.1-sol is enough."
+}
@@ -0,0 +1,32 @@
+{
+  "current_candidate": "bf559632ddd938085702529d9dd6b7c878f4ebc7",
+  "evidence_collected_commit": "d1c7ce27d21a77be87081e12d20e7c8dd6ffa353",
+  "exact_selected_source_map_equal": true,
+  "all_selected_source_files": 1316,
+  "frozen_spec_equal": true,
+  "historical_input_map_equal": true,
+  "marker_contract_equal": true,
+  "source_map_changed_paths": [],
+  "unchanged_evidence_environment": "Godot4.7.2 native macOS; exact guarded private-profile overrides and actual captured native PIDs retained",
+  "native_os_full174_receipts_relabelled": false,
+  "required_reverification": "new local Python test input r04 restored and all4 checker tests rerun, fresh registered verification and independent Opus still required",
+  "current_local_checks": {
+    "ast_parse": "passed, three Python sources",
+    "checker_tests": {
+      "command": "python3 -B tools/test_p08_migration_evidence.py",
+      "tests_run": 4,
+      "elapsed_seconds": 55.197,
+      "result": "OK",
+      "exit_code": 0,
+      "scope": "Synthetic local checker tests only; source edits frozen before this single run. Not native, OS, full regression or acceptance evidence.",
+      "final_raw_sha256": {
+        "tools/test_p08_migration_evidence.py": "fd9ef7ec63f53dfd5e04e261606c9303d01a2e6590e22a35fef2e35df1429280",
+        "tools/check_p08_migration_evidence.py": "928353233bf85436a5aaf8c4f77f3f82472d94b8ad2646cb1b030baf92b3b5dc",
+        "docs/plans/original-completion-2026-10-02/execution/p08-02/ROOT_REPAIR_INPUT-r04.json": "23ef44a0775ca53f1f67d6d64fcc96cd33eae935478768e9ce9431796eaa12ef"
+      }
+    },
+    "native_or_full_regression_run": false,
+    "evidence_reuse_pending": "Reuse of earlier native 8, OS and final 174 regression evidence is pending Root confirmation of exact source_map, input, config and environment matches. If restored r04 is in the selected source map, Root must decide whether to re-collect. Independent review and integration remain pending; no unconfirmed evidence counts asserted."
+  },
+  "acceptance": false
+}
@@ -0,0 +1,20 @@
+{
+  "candidate": "bf559632ddd938085702529d9dd6b7c878f4ebc7",
+  "passed": true,
+  "registered_receipt": "/Users/korizatou114/Library/Application Support/agent-control/the-tree-p08-02/runs/473aac189bb644448cb69156ba16b944-validation.json",
+  "receipt_sha256": "9e1d7faa9b4d6d82595eb19e3025fdce88b5a10023aa623d21be777ee7a82c70",
+  "validator_result": {
+    "integrity": "verified",
+    "deficits": [],
+    "acceptance": "pending",
+    "conditions": {
+      "mixed_construction": "pending",
+      "occupancy_change": "pending",
+      "actual_actor_positions": "pending",
+      "combined_rescue_logistics_ev": "pending",
+      "no_overlap_orphan_loss": "pending",
+      "unsolvable_preservation": "pending"
+    }
+  },
+  "original_unit_completed": false
+}