colitur.git/test/test_precedence_ef.ml, branch v0.1.0

test(ef): commemoration, transfer, and privilege coverage for the Major Litanies

2026-08-13T18:15:42Z

test_precedence_ef.ml: a privilege_cases row for RG109(f), deliberately rank Class4 so it cannot pass via the pre-existing "of a I-class day" branch by accident. Three disposition_cases rows: the two RG80 trigger shapes (Transfer, not Commemorate) and a synthetic third row proving the winner-slug guard is specific to the two named trigger days, not "any I-class winner". Two transfer_target unit tests, both using a pathological always-blocking occupant to prove no RG96 search runs at all for this candidate -- a much stronger witness than a cooperative occupant, which could pass by accident. test_golden.ml: three new pinned tests through the real data pipeline -- the ordinary case (2026-04-25, matches the oracle's own M5 date), the four Sunday-displacement years in 2005-2050 (2010, 2021, 2027, 2032, RG111(b)), and the Easter-Monday transfer shape (2011, complementing the Easter-Sunday shape already pinned by the existing 1666/2038 Easter-extreme tests, now widened to include the transfer and its target). test_rite_ef.ml: the existing "no transfer lands inside the Easter octave" property gets a named, cited exemption for the Major Litanies (RG80's own deliberate exception), plus a positive companion test asserting the exemption is not vacuous (exactly two landings in 2005-2050, both Easter+2). That companion test was written but never registered in its own suite list -- caught while mutation-testing the transfer (a mutation that should have failed it passed silently); fixed by registering it. Mutation-tested, each of the three parts separately (recorded in the task's own report): moving the data entity off 25 April fails 7 tests (golden + oracle only -- zero unit tests, since they build synthetic candidates); corrupting the transfer's winner-slug guard fails 6 (the two disposition_cases rows, 3 golden pins, and the newly-registered transfer-count test); flipping the privilege to Ordinary fails 9 (privilege_cases, the disposition SYNTHETIC row, the oracle's count pin, 5 golden pins, and the exhaustive Validate property sweep).

fix(ef): RG 112(d) excludes a same-BVM commemoration (fix round 1, F1)

2026-08-13T15:09:27Z

The BVM Saturday Office is itself "de B. Maria Virg."; RG 112(d) (Caput XVI, "De Commemorationibus") excludes another commemoration whose own oration invokes the SAME BVM's intercession. our-lady-of-mt-carmel (16 July, Commemoration_only) was wrongly admitted as an ordinary commemoration alongside the Office on every 16-July-Saturday -- its own collect explicitly invokes "her" (eius) patronage, and both photographic scans print a rubric anticipating exactly this collision (an either/or of Mass texts, not a joint commemoration). Precedence_ef gains marian_slugs (a closed, hand-verified list of sanctoral slugs that are themselves feasts of the BVM) and is_bvm_office (true for subject = Bvm OR a marian_slugs member -- the disjunction is load-bearing: the Office is identified only by subject, since its own slug is deliberately reused from the ordinary ferial fallback; Mt Carmel is identified only by the list, since its own subject tag is Saint, not Bvm). disposition gains a new branch, checked before the pre-existing Commemoration_only catch-all, since the one live witness is Commemoration_only and that branch's own "always Commemorate" had no exception for this shape. Checked exhaustively against every Marian-named sanctoral entry: only Commemoration_only-status entries can ever reach this branch live (a Feast-status one always wins outright instead of losing). Of the two such entries in the data, only Mt Carmel is reachable -- our-lady-of-ransom (24 September) is provably unreachable: 24 September falling on a Saturday forces 1 September to a Thursday, which by construction makes 24 September the September Ember Saturday, Class2, every time, not a sampled coincidence. Re-measured against the full pre-change binary: the original 75,853-day blast radius is unchanged in total, splitting into 74,633 colour-only + 1,220 colour+comms (16-July Saturdays). Mutation-tested: exactly 3 new tests redden, nothing else.

test(ef): reach RG 91 entry 27 through the real pipeline, end-to-end

2026-08-13T13:57:14Z

Entry 27's band branch has been unreachable for the right reason since it was written: precedence_ef.ml's own comment on it acknowledged the row existed, but the one test exercising it built a hand-typed candidate, never a real Temporal_ef.temporal output, and nothing built the office itself. Closes that gap on both sides: - The band-table row now sources its candidate from Temporal_ef.temporal directly (of_temporal), the same review-finding-3 discipline the Ember/vigil rows already use, so a future regression in the office's own construction fails this row too, not only the unit-level test. - Two new resolve-level end-to-end tests, using the real band/ disposition/admit triple (not stubs): the office winning a real otherwise-unoccupied Saturday and admitting an ordinary commemoration alongside it (1 August 2026, Holy Machabees -- the task brief's own worked oracle example), and the office genuinely losing to a real competing feast and being omitted under RG 26, with no confusion from the one live data witness that also carries subject Bvm (12 September, Most Holy Name of Mary) -- proving RG 26's rank-keyed omission fires before any subject-keyed rule ever could. - Two golden pins mirroring the same two dates, checked by hand against `colitur day 2026` before being typed in, following this file's own "a golden test blesses whatever it is given, verified independently" discipline. The existing 2038-03-06 pin (RG 91 entry 24 outranking a Class4 candidate) is corrected in passing: that date is itself a Saturday, so the displaced candidate the pin's own omitted-list check names is now precisely entry 27's own shape, not entry 28's bare catch-all -- the win/lose outcome the pin proves is unchanged, only its own more precise citation. Mutation-tested (recorded in the task report): reverting is_bvm_saturday to a constant false reddens the dedicated unit test, the new end-to-end resolve test, and the 2026-08-01 golden pin, while leaving the 2026-09-12 "loses to a real feast" pin untouched, exactly as expected since that date's outcome never depended on the fix.

test,docs: RG110(c)'s adjacency is unobservable by construction

2026-08-13T10:24:19Z

The re-review asked for a third commemoration in the shape-(c) row, so that "spliced immediately after its trigger" could be told apart from "appended at the end" -- it had proved an append-at-end variant passed the whole suite green, the pair alone being too short to distinguish them. Attempting it establishes something better than the test requested. On this day-shape a third commemoration is rubrically impossible. Shape (c) requires the trigger to lose the day while still being admitted, and a Class2 trigger can only lose to a I-class day (RG 111(a): one privileged commemoration) or a II-class Sunday (RG 111(b): one, "scilicet de festo II classis"). Both caps are one. The third candidate is now offered and correctly refused, which is the assertion worth having. So RG 110(c)'s "ante omnes alias commemorationes" half has no reachable witness in the EF's rubrical space, and the re-review's own finding that shape (c) never co-occurs with a third commemoration across 8417 years is a consequence of the rubric rather than an accident of the data. The append-at-end variant stays indistinguishable from the splice on every input the engine can construct. Recorded rather than papered over with a synthetic row proving a shape no rubric permits. Also drops Record from the "nothing compares commemoration order" note -- Record.t has no commemorations field at all -- and names, at privilege_of's own (a) branch, that the Holy Name's Sunday privilege currently rests on a substring match rather than on RG 17's closing paragraph, which is its real warrant.

fix round 1: RG 110(c)'s ordering was inverted; M19 lacked an identity gate

2026-08-13T10:01:25Z

RG 110(c), word for word: "quoties vero oratio unius Apostoli addenda est ad modum commemorationis, huic orationi additur altera immediate, ante omnes alias commemorationes." "Huic orationi" (dative, "to this oration") refers to the trigger's own oration; "additur altera" means the companion is added to it, i.e. follows it. "Ante omnes alias commemorationes" governs the pair's position relative to other, unrelated commemorations, not the pair's own internal order. The previous build prepended the companion uniformly in both RG 110 shapes, citing this same clause for both -- correct for (a)/(b) (the day's own oration, not a list member, has nothing to order against), backwards for (c), where the trigger is itself a list member and the companion belongs after it. Fixed in rg110_additions: shape (a)/(b) still prepends; shape (c) now splices the companion in immediately after its own trigger. Re-pinned the unit test that had asserted the wrong order, and corrected both rows' comments to cite the clause that actually governs each shape. No layer in this codebase compares commemoration order: the lectio differential never compares commemorations, the oracle sorts both sides into a multiset, and the golden tests' own describe function sorts its commemorations field too. Recorded as a new, permanent blind spot in CLAUDE.md and the register, not merely fixed around -- the admit_cases unit table is the only place order is asserted at all. Re-measured the full blast radius against the same pre-change binary: same 14,627 differing days, same four shapes and counts, confirming the fix changed only the content of the 593 shape-(c) rows and nothing else. Mutation-tested: reverting the splice back to a uniform prepend reddens exactly the shape-(c) unit test; a new golden pin for a real shape-(c) day (2004-02-22) stays green under the same mutation, living proof that golden pins cannot see this axis, not merely a claim. Separately, M19's own predicate checked a date and a diff shape but not which candidate -- proved insufficient by fabricating a second, bogus companion on 30 June and confirming the whole suite stayed green. Fixed by requiring the admitted commemoration to be exactly commemoration-of-st-peter, mirroring C16's own identity guard. Added a golden pin for 30 June, the only assertion of that entry's identity outside the allow-list. Also: corrected a false claim in the task report that the electronic transcription carries no corroboration of the Holy Name fallback -- its own table of contents does, at a line not checked on the first pass. Added two further citations, both re-verified against the scan: RG 17's own closing paragraph, settling the Sunday shape's Sunday- exclusion at the rubric level rather than only in the propers; RG 124(c), which assigns red to 30 June by paragraph number rather than by the generic apostle/martyr default. Noted the fragility of deriving the Sunday shape's RG 109(a) privilege from a slug-substring match rather than from a citation. Cosmetic: a stray blank line and an indentation inconsistency.

precedence(ef): RG 110, the inseparable Peter/Paul commemoration

2026-08-13T09:19:19Z

RG 110 (Caput XIV): "In Officio et Missa S. Petri semper fit commemoratio S. Pauli, et vicissim... pro unica habeantur" -- in the Office and Mass of either Peter or Paul, a commemoration of the other is always made, uncapped by RG 111's own admission count. Both photographic scans, full text including sub-clause (c) (a separate trigger this register previously had not transcribed: the same rule also fires when one Apostle's oration is admitted merely as an ordinary/privileged commemoration, not only when it is the day's own office). Three real pairs in the 1962 calendar, all confirmed on both scans: 25 January (conversion-of-st-paul + peter), 22 February (chair-of-st-peter + paul), 30 June (in-commemoratione-sancti-pauli-apostoli + a new commemoration-of-st-peter). The first two already existed as ordinary sanctoral candidates but were wired into RG 111's ordinary, capped admission contest instead of this rule's own uncapped one -- the defect this closes. 30 June had no companion at all: a genuine gap in lectio's own source data (confirmed directly against tridentine-calendar.ini) and in missalemeum's own oracle output for both fixture years, not only a colitur bootstrap miss, so colitur is now uniquely correct on this date among the three engines compared. Added via data/ef/adjustments.sexp's own Add directive, the architecture's existing extension point for exactly this shape of gap, since data/ef/sanctoral.sexp is generated and lectio must not be touched. Precedence_ef.admit gained rg110_additions/rg110_companion_slug, layered on after admit's own four ordinarily-capped branches decide their normal result: checks both the observed candidate and every already-admitted candidate against the three-pair lookup and prepends any companion found, uncapped and additional, never competing for RG 111's own slot. Measured full blast radius (1583-9999, pre-change binary from git archive at 59fbda8 vs the fixed binary, 3,074,246 days each side, 14,627 differing days, all four predicted shapes, zero unclassified): 3,619 Holy Name fallback days (previous commit), 3,533 days Paul was wrongly excluded from Chair of St Peter's own day by a competing privileged feria, 593 days Chair of St Peter itself was admitted only as an ordinary commemoration with Paul entirely absent, and 6,882 days the new 30 June companion is added. The 3,533+593 figure supersedes an earlier, narrower 852-day estimate that measured only the delta between two older code commits, not the rule's full scope. The lectio differential needed no change: it does not compare commemorations at all, and this fix never touches season, slug, rank, or colour. The missalemeum oracle allow-list needed three changes: M12 removed (22 February 2027 no longer diverges in commemoration count), M15's date list gained 2027-02-22 (now reaching identity comparison for the first time, where it lands on that entry's own pre-existing, unrelated limit -- one of the two commemorations is temporal-origin and unnamed), and a new M19 for the 30 June gap. Mutation-tested, both halves separately: disabling the admit-level addition reddens the two new unit tests, the oracle's own M12-shaped checks, and an existing golden pin (updated to expect the Paul companion); removing only the 30 June data entry reddens solely the oracle's new M19 count, confirming the code and data halves are independently load-bearing.

temporal(ef): Holy Name of Jesus, RG 17(a) -- fallback + Lord subject

2026-08-13T08:40:40Z

RG 17(a): "festum Ss.mi Nominis Iesu, celebrandum dominica quae occurrit a die 2 ad 5 ianuarii (secus die 2 ianuarii)" -- both photographic scans, confirmed on the calendarium's own January table and the Mass propers' own heading. Two defects closed together, since they interact through RG 91 entry 14/RG 16(a): 1. The Sunday shape (2-5 January) now tags subject = Lord, matching RG 91 entry 14 ("Festa Domini II classis") -- previously subject = Temporal, the same defect shape Holy Family had before it. 2. The fallback (RG 17(a)'s own "secus die 2 ianuarii", the calendarium's "vel, ea deficiente, die 2 ianuarii") is now built: 2 January carries the feast whenever no Sunday falls 2-5 January that year. Before this fix colitur emitted no Holy Name office at all in such a year -- a real II-class feast simply missing, on 3,619 of 8,417 years in the domain (43%; 20 of the 46 years 2005-2050, independently re-derived against `date -d`). Both shapes band identically at RG 91 entry 14's MOVABLE half (Precedence_ef.entry_14_movable_band) -- one feast, not two differently classified ones, per the Mass propers' own single heading covering both dates. No live collision exists in the current sanctoral data (nothing fixed-date falls 2-5 January besides a Commemoration_only entry on 5 Jan), so the RG 16(a)/RG 112(a) interaction is proved with synthetic precedence-level fixtures mirroring Holy Family's own precedent, not a live-data regression. One asymmetry the synthetic tests found and kept honest rather than forced: a losing Holy-Name-Sunday is privileged (RG 109(a), "of a Sunday") and survives RG 111(a)'s cap; a losing Holy-Name-fallback is not (2 January is not a Sunday, and no other RG 109(a)-(f) category applies to it), so it gets no commemoration at all if ever outranked -- both individually correct readings of RG 109/111's closed lists. Golden pins follow the brief's own vacuity warning: the Sunday shape already emitted the right slug/rank/colour before this fix (an ordinary Sunday and Holy Name coincide on those fields), so its pin asserts subject explicitly rather than pinning the whole line uncritically; the fallback pin (2029-01-02) is the one with teeth against the larger defect, reddening on slug/rank/subject together if reverted. Mutation-tested: each half of the fix reverted separately reddens exactly its own tests (Sunday subject: the dedicated unit test, the golden pin's subject field, one precedence end-to-end test; fallback: the fallback unit/anchor/golden tests, the new C16 differential count, and both shapes of the precedence end-to-end tests) and nothing else. Adds data/ef/expected-divergences.sexp's C16 (lectio has no fallback for this window at all) -- the lectio differential's own count, independently re-derived against `date -d` for the fixture's 2005-2050 span, not transcribed from the domain-wide figure. The missalemeum oracle window (2026-2027) exercises only the Sunday shape and needed no new entry: subject is not one of its compared fields, and both years' rows already corroborate the Sunday-shape ranking (missalemeum's own 2026-01-04 and 2027-01-03 rows: rank 2, white, title "Holy Name of Jesus").

precedence(ef): fix round 1 (F2, F7) -- correct a false transcription claim, cite RG95's real primary authority

2026-08-13T07:16:30Z

F2: the disposition comment for RG112(a) claimed the electronic transcription "carries NEITHER this paragraph's own worked example NOR its own more specific Mass-proper rubric... exactly the transcription's documented defect" -- wrong on both counts, checked directly rather than taken on trust. RG 112 has no worked example in any of the three source documents (the transcription included) -- there was nothing to drop. The Holy Family Mass propers' own 13-January rubric is absent from the transcription not because it silently drops content it otherwise has (the project's own documented defect, dropped calendarium commemoration lines) but because that specific document (a 2006 web capture) contains almost no Mass-propers text of any day: 5 Introit-line occurrences across 26,322 lines, against 61 and 402 in the two photographic scans; its own "Proprium de Tempore" page is a table of contents linking to PDFs the capture never pulled in. A coverage gap in a partial web capture, not an instance of the narrower, different defect the methodology note actually describes. The ruling is unaffected -- the propers' quotation is real, on two independent photographic scans, and settles 13 January on its own -- only the attribution is struck, here and in test_precedence_ef.ml (the register and the task report are corrected separately). F7: RG 95's own second paragraph -- present in all three documents, including the transcription -- is a more direct, occurrence-level authority for the exact collision than RG 112(a): "if two feasts of the same Divine Person... occur together, the one higher in the table is kept, the other omitted." Added as the primary citation on disposition's RG112(a) branch, with RG 112(a) and the propers' note kept as corroboration. Also corrects RG 112(b)'s own framing: it is bidirectional ("et vicissim"), and only the Lord-feast-wins direction (RG 16(a)) is implemented -- the Sunday-wins direction has no witness anywhere in this codebase's data and was wrongly described as already covered.

precedence(ef): fix round 1 (F1) -- band's -14 sentinel inverted the table

2026-08-13T07:11:21Z

entry_14_movable_band = -14 was justified only by "cannot collide with a real table position (1..28) or unclassified" -- a collision argument, not an ordering one. band's own value is compared by plain <, so a negative number beats every I-class entry (1-13), not merely the fixed Baptism it was built to beat. Not live against the shipped universal sanctoral data (no I-class universal feast falls 7-13 January, which is why the original 1583-9999 sweep showed only the intended shape), but entries 12/13 admit PROPER and INDULT I-class feasts too, and those arrive via adjustments.sexp-style overlays -- this architecture's own advertised extension path. A diocesan patronal feast landing in Holy Family's window would have made the inversion live immediately: the I-class feast displaced and transferred to the next day, reproduced and confirmed before this fix, reverted after. Fixed by rescaling every band branch to the real RG 91 entry number times ten (unclassified untouched at max_int) instead of the bare 1..28, reserving genuine headroom immediately before each entry for a movable sub-rank -- not an arbitrary rescale: RG 91's own text records a "primum mobilia, deinde fixa" split at three rows, not only entry 14 (entry 12's own ninth sub-item, entry 14, and entry 20), so this leaves room for the other two without a second rescale if either ever gets a live witness. entry_14_movable_band now derives from entry_14_fixed_band - 1, tied together so they cannot drift apart. Added a synthetic I-class end-to-end test reproducing the exact counter-example that found this: a universal I-class candidate inside Holy Family's own 7-13 January window is now observed and does not transfer, with Holy Family itself falling to RG109(a)'s ordinary privileged-Sunday-commemoration path, the same treatment any genuine feast outranking an ordinary Sunday gets.

precedence(ef): RG91 e14 "primum mobilia" + RG112(a) inseparable exclusion

2026-08-13T06:15:48Z

Two related fixes, both needed for Holy Family to actually observe correctly on the seven years it coincides with the fixed Commemoration of the Baptism of the Lord (13 January): 2008, 2013, 2019, 2030, 2036, 2041, 2047. band's entry 14 ("Festa Domini II classis") gated on sanctoral origin only, so a movable Class2 Lord feast had no row in the table at all. RG 91 entry 14's own text says otherwise: "primum mobilia, deinde fixa" -- movable first, then fixed. entry_14_movable_band gives Holy Family (the one movable candidate this codebase builds with subject Lord) a value strictly better than the fixed half's own literal 14, negative so it can never collide with a real table position -- without it, the two would tie at 14 and fall to Precedence.resolve's own alphabetical tie-break, which picks the WRONG side ("commemoration-of-the-baptism-of-the-lord" sorts before "ef-time-after-epiphany-sunday-1"). disposition gets a new branch for RG 112(a) (Caput XVI, "De Commemorationibus"): the Office, Mass or commemoration of a mystery of one Divine Person excludes a commemoration of another mystery of the SAME Divine Person. Winner and loser both subject Lord -> Omit, not Commemorate -- without it, Holy Family observing correctly still left the Baptism admitted as an ordinary Class2 commemoration under RG111(b), which the primary text does not allow. Confirmed by a MORE SPECIFIC primary source than RG 112(a) alone: the Holy Family Mass propers' own note, found on both photographic scans, word for word, immediately after the Postcommunio -- "Si festum S. Familiae occurrerit die 13 ianuarii, Missa dicitur de festo S. Familiae, sine commemoratione Baptismatis D.N.I.C., et sine commemoratione dominicae." The electronic transcription carries neither this rubric nor RG 112(a)'s own worked example -- another instance of its documented defect, not merely absent from one check. Still expected red: the lectio differential (2005-2050) now shows a new, real divergence on those seven 13-January-Sunday dates -- lectio has no Holy Family at all, so it still shows the fixed Baptism observed there, unlike colitur now. The oracle's own M-series entries also still need widening for the observed-identity axis added two commits ago. Both are the next commits.