Graph invariant 10 reported Chapter 3 and Chapter 4 failures under its own
number, through a public API. `InvariantViolation` becomes
`WellFormednessViolation` carrying a two-armed `ViolationKind`: `Invariant` for
a numbered graph invariant, `Requirement` for a normative rule named by its
`req:` label. Invariant 10 keeps only reference resolution.
`check_invariants` stays comprehensive, so every broad caller keeps its rule
coverage; `check_invariant` narrows to the invariant arm and a symmetric
`check_requirement` is added. Both are projections of the aggregate. Four rules
move to their own labels, one of which this rung mints:
`req:time:tempo-segment-shape`, stating the enforced shape/end_tempo
compatibility without resolving P13-S8. `GraphInvariant` did not move: 21
variants, unchanged.
44 suites / 1606 passed / 0 failed / 0 ignored (1586 -> 1606, 43 -> 44)
clippy -D warnings clean; fmt -p epiphany-core -p epiphany-testkit clean
core_spec.pdf rebuilt, undefined references cleared
Gates 1-16 pass; evidence in spec/EVIDENCE_P13S29_EXECUTION.md. All 54 mutations
of §3 observed, none compile-only, none passing. Amendment 1 corrected M1·C6 and
M1·C7 from one observer to two after measurement contradicted the derived cell.
Six execution faults are recorded rather than left in the transcript, none a
contract defect. Three are worth naming here because no tool would have caught
them:
- Pin 10's eleventh row was read as a note, not a test, so
`reversed_aleatoric_bounds_stay_invariant_four` was never written. M11's cell
named it, and the omission surfaced only as a radius mismatch.
- `display_renders_each_arm_exactly` carried a comment stating its pinned
fixture correctly above code that built a different one. Pin 10 had foreclosed
the choice precisely because it moves radii; with the fixture repaired the
test entered M11's cell and left M17·C1's, both as pinned.
- Pin 9's `/// 10.` rider note was the one prose row left unmigrated, naming
three labels where four were required and still filing the multiplexing as
pending. Gate 16 states these outcomes have no machine observer; eight rows
were verified by their own stale phrase, and the ninth was not there.
Gate 10 rejected the ledger append over one character: the artifact read
2026-08-12 where pin 13 pins 2026-08-11. The artifact was corrected to the pin,
and the date question is flagged for the owner as a possible administrative
amendment rather than settled here.
CLAUDE.md's baseline still reads 43 suites / 1586 tests and is not in the touch
table; it follows in a separate reconciliation commit.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ps1szk2mSfgp4Cz21eVH9x
Land the two real pre-passes as canonical *derived annotations* (pure
functions of the materialized Score + profile, recomputed on
materialization, never serialized into canonical Score bytes), exposed via
`derive_annotations`:
- Spelling: a Temperley-style line-of-fifths centre-of-gravity preference
rule (key-free, deterministic), preserving authored CMN letters and only
inferring spelling for chromatic/integer input. `resolve_spelling`
layers authored overrides above the inferred default (the RespellPitch
precedence rule). `spell` now takes `&Pitch` and delegates to
`simplest_spelling`.
- Decomposition: metric greedy-aligned splitting on a 1/4096 integer grid
(barline + dyadic-boundary ties), with exact sounding->notated tuplet
conversion before gridding. Components reconstruct the event duration
(invariant 15).
- A per-event-kind eligibility `TaxonomyReport` so "ineligible" is always
explicit and counted, never silently absent.
Test infrastructure (Agent F): a 29-fixture representative corpus +
taxonomy harness (corpus.rs), the H spelling/decomposition merge gate
(prepass_harness.rs), a discrete `tests/prepass.rs` CI target, conformance
stage [7b], a dedicated CI job, and the Pass-12 batch tracker.
Review hardening folded in (nine findings):
- Guard `decompose_metric` against a zero-length measure (was a
divide-by-zero panic; now reported ungriddable).
- Resolve spelling-override priority via `Reverse` instead of negation
(was an i32::MIN overflow).
- Verify spelling *register* (octave), not just pitch class, in the gate.
- Close the decomposition under-emission gap: the unusual-outcome
taxonomy buckets are an exact per-fixture whitelist (classify_corpus
step 5b).
- Generalize `accidental_ids` to a glyph stack so authored extreme
alterations (triple-sharp+) reconstruct exactly instead of being clamped.
- Per-fixture spread checks in the non-vacuity tripwire and broad-bucket
coverage, so no single rich fixture can carry a signal (partial-stub
resistance); added a `mixed_rhythm` fixture for margin.
- Pin the integer-grid note-value math to the canonical rational helpers
via an exhaustive test; cross-reference comments.
- Replace the O(n^2) tuplet innermost-resolution scan with an id index.
fmt + clippy -D warnings clean; 199 tests pass; conformance suite green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011giSRaHCFCGm1Z2SWv6JHt
M1 — fix Agent C framework defects, tests-first:
- causal ordering: topologically order DVV edges instead of assuming HLC
alone implies causal order (false for adversarial remote envelopes);
HLC only breaks ties among ready operations.
- anomaly cutoff: quarantine from the earliest counter participating in
any violating HLC pair (suffix-minima), e.g. [100,200,50] quarantines
from counter 0, not counter 1.
- pending detection: DVV contiguous ranges use the zero-based per-replica
counter floor; first absent id in any asserted range holds the dependent
pending (vector coverage, not only dots).
- transaction snapshots: rollback removes member-generated conflicts.
- edge tests in concurrent_reduction.rs for all six audited cases.
M2 — reduce onto Agent B's real score graph:
- OperationSet::reduce_onto(&Score) -> GraphMaterialization { state, score }
mutates the real arena, voices, regions, tombstones, indexes, and
cross-cutting structures; base-free reduce() retained.
- VoiceOrigin::SystemPromoted now carries { winning_operation,
losing_operation, original_voice }; spec and Invariant 18 updated.
- graph-aware migration, forward undo, system breaks, promotion pre-pass.
- tests/graph_reduction.rs: 11 tests asserting check_invariants is clean,
plus a 64-seed order-independence sweep.
Pass-11 spec decisions resolved (blocking subset): DVV floor (P11-C7),
HLC-vs-causality, promoted-voice derivation inputs (P11-C4 / core P11-3).
Payload/Score canonical encoding remain deferred to the companion docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Levi Neuwirth