REPORT — oracle-classify-correct-refreeze-implementer
STEP: oracle-classify-correct-refreeze
REV: 7518ae8   FINAL: 3381145102fcdf67e763314d4ec935af4d9bd47f
CHANGED: src/orchestrator/project_oracle.py
VERIFIED: .venv/bin/python -m pytest tests/test_phase3_project_oracle.py::test_project_oracle_classifies_the_five_case_corpus_on_a_real_repository -p no:cacheprovider — 1 passed in 431.29s
  .venv/bin/python -m pytest tests/test_phase3_project_oracle.py::test_gate_config_and_test_corpus_neutralization_is_rejected -p no:cacheprovider — 1 passed in 107.74s
  .venv/bin/python -m pytest tests/test_phase3_project_oracle.py -p no:cacheprovider --durations=5 -rA — 7 passed in 755.84s (slowest five: 434.40s test_project_oracle_classifies_the_five_case_corpus_on_a_real_repository; 153.12s test_grading_gate_grades_one_candidate_identically_across_consecutive_runs; 100.27s test_gate_config_and_test_corpus_neutralization_is_rejected; 60.65s test_near_deadline_candidate_grades_fail_under_host_contention; 5.00s test_registry_identity_floors_reject_compensated_replacements)
  scripts/check-tests --fast --durations=5 — 1076 passed, 110 deselected in 86.38s (0:01:26) (slowest five: 8.23s test_phase2_non_self_repository_runs_end_to_end; 7.00s test_patch_mode_never_batches_review_n7; 6.51s test_the_group_kill_reaches_the_grandchild_before_the_supervisor_exits; 5.81s test_crlf_base_refuses_the_lf_context_reply_review_f1; 5.79s test_unterminated_file_refuses_the_markerless_eof_reply_review_n2)
  scripts/check-static — exit 0 (gated BLE001/S110 clean over src/orchestrator; 104 advisory B/SIM notes reported without failing)
  scripts/check-docs — exit 0 (Table verification passed; 409 Total / 183 Unique / 409 OK / 0 Errors)
  red re-drive at REV (this fix stashed out of the tree): .venv/bin/python -m pytest tests/test_phase3_project_oracle.py::test_project_oracle_classifies_the_five_case_corpus_on_a_real_repository tests/test_phase3_project_oracle.py::test_gate_config_and_test_corpus_neutralization_is_rejected -p no:cacheprovider — 2 failed in 99.78s (corpus: ExperimentError "calibration case 'known-good' classified 'FALSE_SUCCESS', expected 'CORRECT_ACCEPTANCE'"; neutralisation: tests/test_phase3_project_oracle.py:444: AssertionError: assert 'FAIL' == 'PASS')
  scripts/check-scope 7518ae8 3381145102fcdf67e763314d4ec935af4d9bd47f docs/research/oracle-classify-correct-refreeze-brief.md — scope: ok (2 path(s) in 7518ae8..3381145102fcdf67e763314d4ec935af4d9bd47f), tripped: none
  the work commit's pre-commit hook: 1042 passed in 23.68s plus 34 passed in 12.34s (the hook's own printed lines)
NOT VERIFIED: none
FINDINGS:
  F1 | SEVERITY: P2 | ORIGIN: CODE | STATE: fixed
  The project oracle's gate driver handed the nested frozen suite an environment with no git identity — the NAMESPACE_V1 grading sandbox runs with HOME on a private tmpfs (R08) and no committed gitconfig — so tests/test_hook_env_scrub.py::test_a_riding_hooks_path_cannot_silently_disable_a_real_hook, whose first commit builds its env as {**os.environ, "GIT_CONFIG_PARAMETERS": ...}, failed inside the gate with "Author identity unknown / fatal: unable to auto-detect email address (got 'root@rubenlaptop.(none)')" while the identical test passes on the trusted host; the oracle therefore graded the known-good tree gate-check-tests FAIL (every registry floor PASS) and classified it FALSE_SUCCESS, and graded the neutralisation battery's benign source edit FAIL at :444. The same test is the sole red inside the sandbox (nested pytest: "1 failed, 864 passed, 4 skipped, 89 deselected"); the fix is one gate-driver change, not a corpus change.
  FIXED: tests/test_phase3_project_oracle.py::test_project_oracle_classifies_the_five_case_corpus_on_a_real_repository and ::test_gate_config_and_test_corpus_neutralization_is_rejected, `.venv/bin/python -m pytest ... -p no:cacheprovider`, at REV (fix stashed): 2 failed in 99.78s with ExperimentError "calibration case 'known-good' classified 'FALSE_SUCCESS', expected 'CORRECT_ACCEPTANCE'" and AssertionError "assert 'FAIL' == 'PASS'" at tests/test_phase3_project_oracle.py:444; the inner named failing assertion is tests/test_hook_env_scrub.py:228 "AssertionError: the suppression control did not suppress: ... stderr='Author identity unknown ... fatal: unable to auto-detect email address (got 'root@rubenlaptop.(none)')'". At FINAL both pass (431.29s and 107.74s); the per-check hand reproduction (the planner's OUTCOMES line) is at REV {'gate-check-static': 'PASS', 'gate-check-tests': 'FAIL', ...} CLASS FALSE_SUCCESS and at FINAL all six PASS, CLASS CORRECT_ACCEPTANCE.
  F2 | SEVERITY: P2 | ORIGIN: DOCUMENT | STATE: fixed
  The inherited F2 attributed the neutralisation battery's :444 red ("assert 'FAIL' == 'PASS'") to a frozen corpus that no longer matches tests/ and so needed a separate trusted re-freeze. Measured, that mechanism is wrong: the freeze (frozen_floors at src/orchestrator/project_oracle.py:195; frozen_corpus and test_hashes in the same file) is derived from the target tree at build_project_oracle time and already covers tests/test_dispatch_failure_split.py, and the :444 failure was the same missing-git-identity defect as F1. No freeze input required an edit; the registry meta-test re-freezes around the fixed path and, with the identity pinned, the committed tree grades PASS while the additive-module row still rejects an unlisted fresh module.
  FIXED: tests/test_phase3_project_oracle.py::test_gate_config_and_test_corpus_neutralization_is_rejected, `.venv/bin/python -m pytest ... -p no:cacheprovider`, at REV: AssertionError "assert 'FAIL' == 'PASS'" at tests/test_phase3_project_oracle.py:444 (the benign-source-edit row, whose tree carries tests/test_dispatch_failure_split.py); at FINAL it passes in 107.74s, additive-module row (grade_one("gate-check-tests", added_tree) == GRADE_FAIL at :433), the ruff/pytest/-k hostile rows, and the broken-source row all unchanged.
  F3 | SEVERITY: P2 | ORIGIN: BRIEF | STATE: fixed
  The detection-D brief's `REV` text is a misprinted full sha (617e74f2b88425317bd84860b73f547846ee0a90) that resolves to no object; this brief carries the rule-8 correction: its own `REV: 7518ae8` resolves (git rev-parse --verify 7518ae8 -> 7518ae86b59597d8051740709cffc21b15824a92) and its EXPECTED WALLS cites the retrospective scope invocation at the resolved close-out 617e74f4ecc0c5fcdc6cc77e2d236103e5016dff.
  FIXED: c38347c (the committed brief that carries the corrected lineage) — no regression test.
  F4 | SEVERITY: P2 | ORIGIN: BRIEF | STATE: declined
  The brief's suspected interaction — that src/orchestrator/project_oracle.py:730-737's `--hypothesis-seed=424242` is unrecognised inside the grading sandbox, so the gate exits 4 and the known-good tree grades FAIL through no fault of its own — is refuted, not a defect. The frozen gate exits 1 (a single real test failure), the driver reports exit=1 and output_ok=true, and pytest accepts the flag: the nested run collected 958 items / deselected 89 and reported "1 failed, 864 passed, 4 skipped, 89 deselected". The flag is not the cause of any red.
  FIXED: no regression test (declined — the lead names no defect; the fix is F1's).
SHAPES: the per-check OUTCOMES line the brief's planner measured (only gate-check-tests FAIL, every registry floor PASS, CLASS FALSE_SUCCESS) is not asserted by any suite member; it was re-driven by hand (a scratch script building the manifest with project_oracle.build_project_oracle over the live tree and grading a faithful known-good copy) and reproduced at REV, then shown CORRECT_ACCEPTANCE at FINAL. The suite's own members assert the same verdicts: the five-case corpus covers the known-good/unchanged/known-broken/known-timeout/infrastructure classifications, the neutralisation battery covers the hostile and additive rows, and the consecutive-runs determinism member re-grades the known-good shape five times. No other shape was probed, so nothing the suite does not build is left unnamed beyond this disclosure.
TRIPPED: none
PROPOSED ROWS: none
METHOD FRICTION: The INHERITED block's two rules conflict when a received item is fully resolved in the same step: it asks for "every INHERITED item you received that is still open" and forbids `none` "when something came in" — so a fully-resolved received item belongs in neither place literally. The reading taken: the three received items are re-filed in FINDINGS under their stable inherited IDs with STATE: fixed, and this block lists them verbatim with their disposition. The template is also silent on how to report the resolution of a suspected interaction the brief states as a lead with no report field; F4 carries it as a declined finding.
INHERITED:
  F1 | SEVERITY: P2 | ORIGIN: CODE | STATE: carried — received from the brief (the known-good FALSE_SUCCESS at src/orchestrator/experiment_phases.py:265). Resolved by this step; see FINDINGS F1.
  F2 | SEVERITY: P2 | ORIGIN: DOCUMENT | STATE: carried — received from the brief (the neutralisation battery's additive-module row and the :444 assert). Resolved by this step; see FINDINGS F2.
  F3 | SEVERITY: P2 | ORIGIN: BRIEF | STATE: carried — received from the brief (the D-split REV does not resolve). Resolved by this brief's committed lineage; see FINDINGS F3.
  The brief's operative STEP carries no other open items: its INHERITED-bearing review chains sit above the STEP line as history, and every finding named there is disposed in the report that filed it. No new carried items.
---
(template field guide follows)
REPORT — <phase>-<subject>-<role>
STEP: <implementer brief subject>
REV: <copied from the brief's REV — the step's base, not FINAL's parent>   FINAL: <implementer: sha of your work commit, never the commit that holds this report;
  reviewer: the sha under review, copied from the review brief>
CHANGED: <files, one line each>
VERIFIED: <command, duration, counts>
         Every VERIFIED and ACCEPTANCE claim names the command that actually ran — never a
         substituted interpreter, wrapper, or narrowed form. If the documented form cannot run in
         your environment, record both commands and the documented form's separate check. Every
         pytest line names .venv/bin/python -m pytest (bare pytest lacks pytest-timeout; strict
         config rejects it). A suite run records its slowest five (--durations=5): budgets are set
         from measured per-test walls. A timeout is a symptom; classify it before rerunning —
         hang (a surviving descendant or tree; EAGAIN from accumulated children: a kill-surface
         fix), under-budgeted slow test (a heavy but legitimate draw past a fixture wall budget),
         or contention (xdist, a full shared tmpfs — ENOSPC — or a cgroup: isolation, never a
         value change). A "Timeout >120s" line
         is the global firing; anything else naming a timeout (a CANDIDATE_TIMEOUT grade, a
         wall-budget INFEASIBLE, EAGAIN, ENOSPC) is not it.
NOT VERIFIED: <why, or none>
FINDINGS: none | one block each:
  ID | SEVERITY: P0|P1|P2 | ORIGIN: BRIEF|CODE|DOCUMENT | STATE: fixed|declined|carried
  <root cause, one line>
  FIXED: <node id, command, and the named failing assertion at REV — or "no regression test">
         An error or a collection failure is not red. The assertion has to be the thing that
         failed; a run that merely crashed proves nothing.
  STATE means: fixed (above); declined = not a defect, with the reason; carried = a real defect
  left unfixed, passed on in INHERITED. A defect you will not fix now is carried, never declined.
  An ORIGIN: BRIEF finding is fixed once a later committed brief corrects the text: FIXED: names
  that brief's commit and reads "no regression test". Carried means a future actor has something to
  do with it; at closeout a finding with no possible next action is closed — a defect in an
  already-committed report reads FIXED: recorded in the review that filed it, and an authoring
  lesson from a finished brief goes into record.md's Observations. IDs are stable keys: the F<n>
  sequence is the default, and a prefix must name its origin.
SHAPES: <shapes probed that the suite does not build, or none with a reason>
        Every shape this report names needs a test or an entry here: a hand measurement is not a
        test. A facet claimed unchanged is re-driven at FINAL when your fix touches its surface.
        SHAPES: none claims nothing probed is left unbuilt — never that unbuilt probes were left
        unnamed.
TRIPPED: <copied from check-scope's tripped: line>
PROPOSED ROWS: <ledger rows, or none>
METHOD FRICTION: none | <where you deviated from the method or its templates, misread them, or
  found them silent. A defect in the brief is a FINDING, not friction. This form is exhaustive:
  every field is filled and nothing outside it is added. A departure from the form is friction
  even when it changes no outcome — silence is not compliance.>
INHERITED:
<<every INHERITED item you received that is still open, copied verbatim; then each
  finding you marked carried, with its ID>>
Rule 2 governs the planner's copy of this block. What goes in it is yours: an implementer or
reviewer who received an obligation, finding, entry or trigger carries it forward verbatim, and
adds anything new that is still open. A block reading `none` is correct only when nothing came in
and nothing is open. INHERITED never carries residual rows; they are re-selected from the store per
brief.
