Orchestrator O2 · 04 In-Memory E2E Test Result

doc 4 of 7 · 2026-05-20 · G4 + G6 gates

phase                : G4 — E2E test authoring + G6 — full pytest discover
outcome              : G4 PASS · G6 PASS — 359/359
production_mutation  : NONE

1. New test surface

tests/_orchestrator_o2_harness.py          : 119 LOC — shared harness
  - DOC_ID / SD_ID / SV_ID constants
  - seed_discoverer(n_candidates=12) → fully-seeded InMemoryDiscoverer
  - build_runner(tmpdir, n_candidates) → (runner, disc, store, kb)
  - write_sg1_approval(td, run_id, …) → SG_1 KB approval doc
  - write_sg2_approval(td, run_id, …) → SG_2 KB approval doc (with
                                          embedded review_decision_id)

tests/test_orchestrator_o2_e2e.py          : 12 tests
  TestKillSwitchOff (3)        — milestone tag O2, killswitch False,
                                  Mode.LIVE refused
  TestE2EHappyPath (2)         — full walk through both sovereign
                                  gates → closeout; resume idempotency
  TestApprovalRequiredOnResume (3)
                               — no-approval / wrong-gate / stale
  TestApprovalConsumedExactlyOnce (1)
                               — replay refused (design doc 03 §7)
  TestUnknownDocumentStop (1)  — STOP_DOCUMENT_UNKNOWN routing
  TestKBArtifactsWrittenForEachPhase (1)
                               — 11 phase + 2 sovereign-request docs
                                  + runs-index entry present
  TestStopDriftOnResume (1)    — source_version drift mid-pause →
                                  STOP_DRIFT

tests/test_orchestrator_o2_gate_invariants.py : 18 tests
  (catalogued in doc 03 §4)

tests/test_orchestrator_o2_phase_bodies.py : 21 tests
  TestSourcePinPhase (1), TestMarkPhase (2), TestCutplanPhase (1),
  TestBackupPhase (2), TestGrantProbePhase (2), TestCutLegAPhase (3),
  TestLegBPhase (2), TestWriteVerifyPhase (2),
  TestLifecycleEnactPhase (3), TestStructuralVerifyPhase (1),
  TestCloseoutPhase (2)

total_new_tests          : 51 (12 + 18 + 21)
total_o2_test_loc        : 627
shared_harness_loc       : 119

2. Pytest full-discover output

============================= test session starts ==============================
platform darwin -- Python 3.11.6, pytest-9.0.2, pluggy-1.6.0
rootdir: /Users/nmhuyen/iu-cutter-build/repo/iu-cutter
plugins: anyio-4.12.1
collected 359 items

tests/test_cutplan_snapshot.py ...............                           [  4%]
tests/test_cutwrite_snapshot.py ......................                   [ 10%]
tests/test_dryrun_snapshot_mark.py .....................                 [ 16%]
tests/test_idempotency.py .......                                        [ 18%]
tests/test_ledger_v2_canonical_cut.py .....................              [ 23%]
tests/test_ledger_v2_canonical_verify.py ..............................  [ 32%]
tests/test_orchestrator_o1_approval_and_batch.py ...........             [ 35%]
tests/test_orchestrator_o1_run_context.py .......                        [ 37%]
tests/test_orchestrator_o1_runner.py ...........                         [ 40%]
tests/test_orchestrator_o1_state_machine.py ..............               [ 44%]
tests/test_orchestrator_o2_e2e.py ............                           [ 47%]
tests/test_orchestrator_o2_gate_invariants.py ..................         [ 52%]
tests/test_orchestrator_o2_phase_bodies.py .....................         [ 58%]
tests/test_phase_contracts.py ...........                                [ 61%]
tests/test_prod_iu_adapter.py .....................................      [ 71%]
tests/test_prod_iu_adapter_canonical.py ...........................      [ 79%]
tests/test_real_postgres_adapter.py .................................... [ 89%]
...........                                                              [ 92%]
tests/test_security_boundaries.py ............                           [ 95%]
tests/test_state_machine.py ...............                              [100%]

============================= 359 passed in 0.36s ==============================

3. Baseline comparison

baseline_o1            : 308/308 PASS (KB v0.6-orchestrator-o1-authoring/doc 04)
o2_full_discover       : 359/359 PASS
delta_tests            : +51 (12 E2E + 18 gate-invariants + 21 phase bodies)
v0_5_regressions       : 0
runtime_total_seconds  : 0.36 s

4. Happy-path walk — invariants captured

The test_full_walk_through_two_sovereign_gates_to_closeout test exercises the complete state machine. At each checkpoint the sidecar is asserted to carry the expected pins:

after_initial_cut (stop at SG_1):
  state                           : AWAITING_CUT_AUTHORIZATION
  stop_code                       : STOP_APPROVAL_REQUIRED
  phases_passed                   : source_pin, mark, cutplan,
                                     pre_write_backup, grant_probe
  context_pins (subset asserted)  : manifest_digest, region_sha,
                                     writer_digest

after_resume_with_sg1 (stop at SG_2):
  state                           : AWAITING_LIFECYCLE_AUTHORIZATION
  stop_code                       : STOP_APPROVAL_REQUIRED
  phases_passed (added)           : cut_leg_a, structural_verify,
                                     leg_b_record, write_verify
  context_pins (added)            : change_set_id, manifest_envelope_id,
                                     verify_result_id

after_resume_with_sg2 (closeout):
  state                           : CLOSEOUT_REPORTED
  stop_code                       : None
  phases_passed                   : ALL 11
  context_pins (subset asserted)  : lifecycle_enacted=True,
                                     lifecycle_enacted_count==12,
                                     all_phase_docs_uploaded=True,
                                     sidecar_state_final=='success'

5. Forbidden-paths — assertions per design

TestKillSwitchOff.test_kill_switch_remains_off_in_o2:
  - assertFalse(__execution_enabled__)
TestKillSwitchOff.test_live_mode_refused_in_o2:
  - runner.cut(mode=Mode.LIVE) raises ProductionExecutionNotAuthorized
TestApprovalRequiredOnResume.test_resume_without_approval_at_sg1_raises:
  - runner.resume(approval_kb_id=None) raises StopApprovalRequired
TestApprovalConsumedExactlyOnce.test_same_approval_kb_id_refused_on_second_use:
  - second consume of same approval_kb_id raises ApprovalConsumedAlready
TestUnknownDocumentStop.test_unknown_document_stops_with_clean_code:
  - unknown document_id → STOP_DOCUMENT_UNKNOWN, no traceback leak
TestStopDriftOnResume.test_drift_in_source_version_id_is_detected:
  - mutating SourceVersion.id between cut and resume → STOP_DRIFT

6. Verdict

g4_outcome              : PASS (51 new tests; 12 E2E + 18 gate-invariants
                           + 21 phase bodies)
g6_outcome              : PASS (359/359 full discover; +51 tests; 0 regressions)
production_mutation     : NONE
live_db_connect         : 0