RUN  v3.2.7 /tmp/circe-pr2-round3-evidence

 ↓ round2.test.ts > PR2 round 2 remaining contract paths > N1: per-proposition API also rejects a dependent whose prerequisite lacks evidence
 ↓ round2.test.ts > PR2 round 2 remaining contract paths > N2: revised binding evidence is identical at validation and dispatch
 ↓ round2.test.ts > PR2 round 2 remaining contract paths > N3: retargeting removes obsolete slot read evidence
 ↓ round2.test.ts > PR2 round 2 remaining contract paths > N4: independent target effects do not invalidate each other
 ↓ round2.test.ts > PR2 round 2 remaining contract paths > control: genuine argument read deps survive every target effect
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R1: non-executable 'query'/'inspect' never invokes
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R1: non-executable 'query'/'reference' never invokes
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R1: non-executable 'request'/'compose' never invokes
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R2: argument-only refs retain validation and frozen preconditions
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R2b: a stale argument target blocks execution
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R3: rejected prerequisite blocks dependent proposition
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R4: out-of-range relation makes whole turn uncompilable
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R5: explicit index cannot bypass a prohibition on the selected proposition
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R6: malformed sibling exclusion cannot be dropped by per-proposition API
 ↓ pr2-previous.test.ts > PR2 independent required behavior > R7: a matching frame hash cannot replace the turn frame identity
 ↓ pr2-previous.test.ts > PR2 independent required behavior > control: valid request still reaches executor exactly once
 ↓ round3.test.ts > a late concurrent reconciliation cannot regress a terminal outcome
 ↓ round3.test.ts > interrupt during an invocation prevents the remaining pending effect from starting
 ↓ round3.test.ts > ambiguous clarification cannot complete a pending request into execution
 ↓ additional.test.ts > invalid revision during validation revokes the old authorization
 ↓ additional.test.ts > an ambiguous affirmative cannot authorize a pending effect
 ↓ additional.test.ts > completed command ids reject different content just like in-flight ids
 ↓ additional.test.ts > overlapping dispatch loops invoke each frozen effect only once
 ↓ additional.test.ts > semantic ingestion snapshots classifier output before awaiting validation
 ↓ additional.test.ts > a semantic turn cannot borrow versions from a different frame
 ↓ additional.test.ts > turn exclusions cover executable slot refs, not only proposition summary refs
 ↓ additional.test.ts > synthesized batch receipts retain the idempotency key actually sent
 ↓ additional.test.ts > a late RUNNING acknowledgement cannot overwrite an observed terminal outcome
 ↓ previous.test.ts > deduplicates the same proposal while the first delivery is in flight
 ↓ previous.test.ts > a new confirm id cannot refreeze a transaction with a RUNNING effect
 ↓ previous.test.ts > a confirm for an absent transaction must not confirm a different live transaction
 ↓ previous.test.ts > a cancelled transaction cannot be resurrected by a revision addressed to its id
 ↓ previous.test.ts > caller mutation during validation cannot retarget frozen effects
 ↓ previous.test.ts > different runtimes do not reuse the same external idempotency key
 ↓ previous.test.ts > an explicitly supplied but non-covering evidence set cannot fall back to current snapshots
 ↓ previous.test.ts > semantic proposals retain the exact frame version rather than rebinding to a new world version
 ↓ previous.test.ts > an inspect/query proposition does not become an executable proposal
 ↓ previous.test.ts > ambiguous classification with a capability does not invoke through ingest
 ↓ previous.test.ts > oversized mandatory context is explicitly reported as over budget
 ↓ previous.test.ts > frames differing in coverage metadata have different exact-input hashes
 ↓ previous.test.ts > late publication cannot attach a stale neighborhood judgment to the latest generation
 ↓ previous.test.ts > an obligation event invalidates a judgment over prior knowledge
 ↓ previous.test.ts > a reused spec id rejected at a later stage cannot leave an earlier accepted value
 ↓ previous.test.ts > required arity-one slot validation also applies to revisions
 ↓ previous.test.ts > journal contains enough command identity to recognize replayed typed proposals
 ↓ previous.test.ts > clear-selection without refs clears the UI selection
 ✓ ownership.test.ts > PR2 evidence-use coverage > O1: a ref used as target and argument remains a dependency of every consuming effect 19ms
 ✓ ownership.test.ts > PR2 evidence-use coverage > O2: a declared shared read guard survives an unrelated revision 2ms
 ✓ ownership.test.ts > PR2 evidence-use coverage > O3: moving a target out of a slot does not lose evidence while the argument still uses it 2ms

 Test Files  1 passed | 5 skipped (6)
      Tests  3 passed | 46 skipped (49)
   Start at  22:43:43
   Duration  565ms (transform 338ms, setup 0ms, collect 1.43s, tests 23ms, environment 1ms, prepare 544ms)