fix(of): Nativity Vigil no longer suppresses a Sunday of Advent

Temporal_of.named claimed 24 December unconditionally as
of-nativity-vigil (Sollemnitas), and temporal's dispatch matched named
before sunday_slug. On a Sunday 24 December, the Fourth Sunday of
Advent -- Tabula I.2, Normae n. 5's own privileged Sundays of Advent,
Lent and Easter -- was silently replaced by the Vigil.

named now returns None for 24 December when that date is itself a
Sunday, so temporal's dispatch falls through to sunday_slug and the
day resolves as of-advent-sunday-4. The Vigil Mass is an evening Mass
of the following solemnity, not an office belonging to 24 December's
own civil day -- the same civil-day-granularity reasoning already
applied to Holy Thursday (Precedence_of.band's entry-1 comment) and to
expected-divergences-litcal.sexp's L1 entry. Every non-Sunday
24 December is unaffected.

Audited every other named entry for the same latent shape (see
.superpowers/sdd/of-r8-advent-iv-report.md): none of them has it,
either because the date can never fall on a Sunday, or because the
date IS the definition of that Sunday (Palm Sunday, Easter, Pentecost,
Trinity, Christ the King), or because the competing movable-Sunday
windows are constructed to exclude that date.

Added test_nativity_vigil_yields_to_advent_sunday, covering
2028-12-24 and 2034-12-24 (both inside the litcal fixture window) and
the 2026-12-24 non-Sunday control. Re-ran the litcal layer: no
allow-list counts moved, since it compares only season and
Ordinary-Time week, and both readings of 24 December are "advent"
either way. Updated test_precedence_of.ml's "entry 3, not 9" comment,
which had carried this as a known, unfixed defect, to describe the
fix; its own pinned assertion (a hand-built Thursday candidate) needed
no change.

1 files changed, 25 insertions, 11 deletions

diff --git a/test/test_precedence_of.ml b/test/test_precedence_of.ml
index 4de9f9a..610cbfb 100644
--- a/test/test_precedence_of.ml
+++ b/test/test_precedence_of.ml
@@ -162,17 +162,31 @@ let part2_cases =
        entry 3. The pinned assertion above is still correct for the Thursday
        it names; only the earlier "band 30" framing read as unconditional,
        which it is not.
-       There is a real, KNOWN, UNFIXED Phase 1 defect behind this: [named]
-       (temporal_of.ml) claims 24 December unconditionally and is matched
-       BEFORE [sunday_slug] in [temporal]'s dispatch, so on a 24-December
-       Sunday the Fourth Sunday of Advent gets NO office at all -- it is
-       silently replaced by [of-nativity-vigil]. 2028-12-24 and 2034-12-24
-       both fall inside the litcal fixture window, and that layer cannot see
-       the gap: it compares season and Ordinary-Time week only, and colitur
-       still says "advent" on that date either way. NOT fixed in this wave
-       (final review Important #4, ledger ruling R8) -- it changes Phase 1
-       behaviour, needs its own tests, and needs the litcal layer re-run.
-       Carried as the first item of the next task. *)
+       RESOLVED (task R8, 2026-08-25): the Phase 1 defect this paragraph used
+       to describe -- [named] (temporal_of.ml) claiming 24 December
+       unconditionally, matched BEFORE [sunday_slug] in [temporal]'s
+       dispatch, so a 24-December Sunday silently lost the Fourth Sunday of
+       Advent to [of-nativity-vigil] -- is fixed. [named] now returns [None]
+       for 24 December when that date is itself a Sunday, citing Tabula
+       I.2/Normae n. 5 and the identical Holy Thursday granularity argument
+       (b) above already makes; [temporal]'s dispatch then falls through to
+       [sunday_slug] exactly as it does for every other Sunday [named] does
+       not claim, landing on [of-advent-sunday-4]. Paragraph (d) above
+       already anticipated the RIGHT band (20, via [is_privileged_sunday])
+       for that case; this fix is what makes that band actually reachable
+       from a real [temporal] call rather than only from this row's own
+       hand-built candidate. Confirmed on both 2028-12-24 and 2034-12-24
+       (test_temporal_of.ml, "nativity vigil yields to Advent Sunday IV"):
+       [named] claims nothing, [temporal] resolves to [of-advent-sunday-4],
+       and the non-Sunday case (2026-12-24 here) is unchanged. The litcal
+       layer was re-run: no allow-list count moved (L1's 36, the 1876
+       witnessed Ordinary Time weeks, and the 859 unwitnessed count are all
+       unchanged), because that layer compares season and Ordinary-Time week
+       only and both readings of 24 December are "advent" either way -- the
+       exact blind spot this same paragraph named above. This row's own
+       pinned assertion (a hand-built candidate, Thursday 2026-12-24, band
+       30) needed no change: it was never wrong, only the claim that no test
+       anywhere exercised the Sunday case. *)
     ( "entry 3, not 9: the Nativity Vigil is a solemnity, though 24 December \
        is also a late-Advent feria",
       ctx ~iso:"2026-12-24" ~season:Vocab_of.Advent,