dot-iu-cutter v0.5 — Constitution Snapshot-source MARK Patch Test Result

Phase: …_patch_apply_and_CI · Nature: tests_NOT_run__blocked_upstream_by_rule_1__no_first_dryrun · Date: 2026-05-18 · doc 2 of 4

tests_run: false ; reason: apply BLOCKED at rule #1 (repo is not a git repo)
first_dryrun: NOT executed ; cut/verify: none ; production/db: none
decision_authority: GPT / User ONLY ; self_advance: PROHIBITED

---

1. Why no tests were run this phase

Patch application rule #4/#5 ("if patch applies cleanly: run tests") is only reached after a clean apply. Apply was BLOCKED at rule #1 (authoritative iu-cutter location is not a git repository — doc 1 §2/§3). The new files cutter_agent/dryrun.py and tests/test_dryrun_snapshot_mark.py were therefore not written to the repo, so the new unit suite could not be invoked there.

tests_run_this_phase: NONE
rationale:
  - rule_1 STOP_AND_REPORT is the controlling gate; tests are downstream of a
    clean apply and were not reached
  - running the suite would require writing the 2 new files into the tree — that
    is the apply action rule_1 blocks; doing it anyway = self-advance (PROHIBITED)
  - the first Constitution dry-run remains FORBIDDEN and was NOT executed
  - existing repo tests (test_idempotency / test_phase_contracts /
    test_security_boundaries / test_state_machine / test_real_postgres_adapter)
    were NOT run: they are unrelated to the unapplied entrypoint and
    test_real_postgres_adapter could touch a DB — out of scope and risk-avoided

2. Authoritative prior test evidence (KB SSOT — referenced, NOT re-executed)

The code-authoring phase already validated the exact patch artifact in an isolated, shredded /tmp scratch. Per KB SSOT (…-test-plan-and-results-2026-05-18.md and code-authoring report §1):

prior_scratch_suite (tests/test_dryrun_snapshot_mark.py, 21 tests): 21 / 21 PASS
  module_sha256: f1f42e83ca23ba0b328f79cf04a8391ac699d1b307eb1b22b52c305f2efa1422
  test_sha256:   31143968f322433cc5da62fa3ccf2a1fbe1905f461940c789a57cb0a116dc1b4
covered (per the required test list of THIS phase):
  - snapshot gate mismatch (sha / length / markers) ABORTS before parse
  - missing BEGIN/END sentinel => FailClosed
  - emits all 3 ratified levels NGUYEN_TAC / KIEN_TRUC_SECTION / DIEU
  - DIEU floor enforced (no sub-DIEU fragmentation; OD-G3)
  - status cascade excludes Điều 44 (tier_2 📋 override -> controlled_draft)
  - Điều 34 draft + 2 obsolete excluded with reason
  - no-DB import isolation (AST: stdlib-only; no db_adapter/psycopg/socket)
  - artifact-only behavior + fail-closed negatives + determinism + coverage
status: this is PRIOR evidence from the code-authoring phase. It is NOT a
  re-execution in the authoritative repo (which is impossible until apply is
  unblocked). It is cited only to show the patch artifact itself is validated;
  it does NOT satisfy this phase's "run tests after clean apply" step.

3. What a future (unblocked) test run must do

once_rule_1_resolved (separate, gated):
  - apply the 2 new files (sha-verified == f1f42e83… / 31143968…)
  - run: python3 -m unittest tests.test_dryrun_snapshot_mark   (expect 21/21 PASS)
  - py_compile cutter_agent/dryrun.py
  - local only; NO production DB; NO first dry-run; NO CUT/VERIFY
  - report PASS + git diff/status (requires the repo to be a git repo first)

4. Statement

• QG2 satisfied: no dry-run / CUT / VERIFY / production DB (nothing executed).
• QG3 satisfied vacuously: no tests run (blocked upstream); none touched a DB.
• test_result = NOT_RUN (BLOCKED upstream at rule #1); prior scratch 21/21 PASS cited as artifact validation only, not re-verified in-repo this phase.
• doc 2 of 4; STOP after 4 docs → route GPT/User. Self-advance PROHIBITED.

Companion docs: patch-apply-log (1), git-diff-status (3), patch-apply-ci-report (4).