fix(of): anticipate St Joseph onto 18 March per Normae n. 56(f)

W1: when 19 March (St Joseph) falls on Palm Sunday, colitur sent him
forward to Easter+9 instead of anticipating the solemnity backward to
18 March, the one rule in the whole Normae transfer machinery that
names an earlier date. Live in 16 of 1583-2100 (1595, 1606, 1617,
1690, 1758, 1769, 1780, 1815, 1826, 1837, 1967, 1978, 1989, 2062,
2073, 2084).

Blocked on a kernel contract: Rite.t.transfer_target required its
result be strictly later than the impeded date, an obligation argued
purely from the EF's RG 96. Traced against Calendar's actual placement
pass rather than assumed unsafe: the round loop re-resolves the whole
year fresh every round, and its two safety nets (~start/~stop; the
round-count guard) both test the target's own value with no comparison
against origin, so a bounded backward jump converges the same way a
bounded forward one does. The one requirement that stays load-bearing
is target <> origin: Precedence.resolve never deduplicates by slug, so
a candidate returning its own argument unchanged would collide with
itself in the same round's contest. Relaxed the obligation accordingly
(rite.mli) and added the fourth transfer_target branch in
Precedence_of, ahead of the existing three (Precedence_of.transfer_target,
Rule 0). calendar.ml needed no logic change, only two comments that
asserted "every genuine transfer moves forward" as a blanket fact.

Unpinned the two tests that asserted the old wrong (forward) behaviour
and replaced them with assertions of the correct one, plus a new
direct unit test of the backward branch across two independent years.

EF output is untouched: data/ef/ and lib/rites/rite_ef/ carry no diff,
and colitur day/readings for 2026, 1583 and 9999 are byte-identical
against a build of the prior commit. OF's own 2005-2050 window carries
none of the 16 affected years and shows zero unexpected diffs; the
full 1583-9999 domain still reaches a fixed point under the exhaustive
sweep.

1 files changed, 57 insertions, 53 deletions

diff --git a/test/test_rite_of.ml b/test/test_rite_of.ml
index 5e369b8..a7d3237 100644
--- a/test/test_rite_of.ml
+++ b/test/test_rite_of.ml
@@ -242,39 +242,35 @@ let test_transfers_reach_a_fixed_point_2005_2050 () =
 
 let slug_of (c : V.rank Cel.t) = Slug.to_string c.Cel.slug
 
-(* precedence_of.mli's own documented, KNOWN-UNIMPLEMENTED fourth
-   transfer_target rule (final review Important #5, ledger ruling R9):
-   Normae n. 56(f) -- St Joseph's solemnity, falling on Palm Sunday, is
-   ANTICIPATED BACKWARD to 18 March, not carried forward. colitur's
-   [transfer_target] has no backward-search branch (rite.mli's own
-   [transfer_target] contract requires the result be strictly LATER than
-   the origin -- an EF-shaped kernel obligation, per that value's own
-   citation), so it instead sends Joseph forward via the general rule 3,
-   the same target date the Annunciation would already be heading to that
-   same year (Easter + 8).
+(* RESOLVED ([of-normae-56f]/W1, 2026-08-26; previously the
+   precedence_of.mli-documented KNOWN-UNIMPLEMENTED fourth transfer_target
+   rule, final review Important #5, ledger ruling R9). Normae n. 56(f) --
+   St Joseph's solemnity, falling on Palm Sunday, is ANTICIPATED BACKWARD
+   to 18 March, not carried forward. {!Colitur_kernel.Rite.t.transfer_target}'s
+   own obligation was relaxed from "strictly later than origin" to
+   "different from origin" (rite.mli's own comment carries the argument,
+   traced against {!Colitur_kernel.Calendar}'s actual placement pass, not
+   merely assumed safe), and [Precedence_of.transfer_target] gained a
+   Rule 0 for exactly this collision -- see that value's own .ml/.mli for
+   the full citation.
 
-   This is NOT fixed by this task (explicitly out of scope, per the task
-   brief) -- pinned here, on a REAL affected year, so the current
-   (known-wrong) behaviour is asserted deliberately rather than passing
-   silently as if it were correct. All years with Gregorian Easter = 26
-   March put Palm Sunday on 19 March, coinciding with Joseph; 2062 is the
-   nearest one after today (2026) still inside the differential-adjacent
-   range this project favours for worked examples.
+   Previously pinned here as KNOWN WRONG (colitur sent Joseph forward to
+   Easter+9 via the general rule 3 instead); now pinned as CORRECT. All
+   years with Gregorian Easter = 26 March put Palm Sunday on 19 March,
+   coinciding with Joseph; 2062 is the nearest one after today (2026)
+   still inside the differential-adjacent range this project favours for
+   worked examples.
 
    2062 is a doubly-worked example, not chosen only for Joseph: that same
    year, 25 March (the Annunciation) falls on HOLY SATURDAY (Easter - 1),
-   so BOTH solemnities are impeded and BOTH nominally target Easter + 8 via
-   the general rule -- {!Colitur_kernel.Calendar}'s own placement pass
-   settles that collision by landing whichever it processes first (Joseph,
-   departing the earlier date, 19 March) on Easter + 8 and pushing the
-   other to the next admissible day; empirically, on shipped data, it is
-   JOSEPH who is pushed one day further still, to Easter + 9 -- read as
-   further confirmation this is genuinely rule 3's general forward search
-   (sensitive to whatever else is competing that year), not the fixed,
-   collision-proof Easter+8 destination rule 3's cousin (the Annunciation's
-   own Holy-Week branch) names outright -- not as a claim about exactly
-   which of the two wins the tie, which this test does not depend on. *)
-let test_joseph_palm_sunday_2062_known_wrong_forward_transfer () =
+   so the Annunciation is ALSO impeded and targets Easter + 8 via the
+   general rule -- independently of Joseph now, since Rule 0 is a fixed
+   one-day displacement that never enters the general search or competes
+   with the Annunciation's own target at all (Joseph departs 19 March for
+   18 March; the Annunciation departs 25 March for Easter+8 -- the two no
+   longer collide the way they did when both were pushed toward the same
+   forward date). *)
+let test_joseph_palm_sunday_2062_anticipated_backward () =
   let easter_2062 = Computus.gregorian_easter 2062 in
   Alcotest.(check string) "2062 Easter is really 26 March (the Normae n.56(f) trigger)" "2062-03-26"
     (Date.to_iso8601 easter_2062);
@@ -285,36 +281,44 @@ let test_joseph_palm_sunday_2062_known_wrong_forward_transfer () =
   Alcotest.(check string) "25 March 2062 is really Holy Saturday, coinciding with the Annunciation too"
     "2062-03-25" (Date.to_iso8601 holy_saturday);
   let days = Cal.year real_of_rite real_of_layer 2061 in
-  let joseph_day =
-    Array.to_list days |> List.find (fun d -> Date.compare d.LD.date palm_sunday = 0)
-  in
+  let day_of d = Array.to_list days |> List.find (fun day -> Date.compare day.LD.date d = 0) in
+  let joseph_day = day_of palm_sunday in
   Alcotest.(check string) "Palm Sunday itself is observed on 19 March (Joseph impeded, correctly)"
     "of-palm-sunday" (slug_of joseph_day.LD.observed);
   let joseph_slug = "joseph-husband-of-the-blessed-virgin-mary" in
   let departure =
     List.find_opt (fun (c, _) -> slug_of c = joseph_slug) joseph_day.LD.transferred_out
   in
-  match departure with
+  let eighteen_march = Date.add_days palm_sunday (-1) in
+  (match departure with
   | None -> Alcotest.fail "expected Joseph to be transferred_out from 19 March 2062 (impeded by Palm Sunday)"
   | Some (_, target) ->
-      (* Concrete, not merely qualitative: 2062-04-04 is the actual pinned
-         value on shipped data today (Easter + 9, one day past the
-         Annunciation's own Easter + 8 -- see this test's own header). A
-         future data or Calendar change that moves it is meant to be
-         caught here; the point being pinned is the SHAPE (forward, past
-         Easter, not 18 March), asserted again below in a form that does
-         not depend on the exact day. *)
       Alcotest.(check string)
-        "KNOWN WRONG (Normae n.56(f) unimplemented, precedence_of.mli): Joseph lands forward, past \
-         Easter (currently Easter+9, one day beyond the Annunciation's own Easter+8), not the \
-         rubrically-correct 18 March (backward)"
-        "2062-04-04" (Date.to_iso8601 target);
-      Alcotest.(check bool) "not 18 March -- confirming this really is the anticipation defect, not a \
-                             coincidence" false
-        (Date.compare target (Date.add_days palm_sunday (-1)) = 0);
-      Alcotest.(check bool) "and it lands on or after Easter itself (rule 3's forward search, not a \
-                             backward anticipation)" true
-        (Date.compare target easter_2062 >= 0)
+        "Normae n.56(f): Joseph is anticipated BACKWARD to 18 March 2062, not carried forward"
+        "2062-03-18" (Date.to_iso8601 target);
+      Alcotest.(check bool) "which really is the day before Palm Sunday" true
+        (Date.compare target eighteen_march = 0));
+  let march18_day = day_of eighteen_march in
+  Alcotest.(check string) "and 18 March 2062 itself observes St Joseph (not Cyril of Jerusalem's \
+                           ad-libitum memorial, not the ordinary Lenten Saturday feria)"
+    joseph_slug (slug_of march18_day.LD.observed);
+  let joseph_arrived =
+    match march18_day.LD.transferred_in with
+    | Some c -> slug_of c = joseph_slug
+    | None -> false
+  in
+  Alcotest.(check bool) "and 18 March 2062 records the arrival (transferred_in), not a silent placement"
+    true joseph_arrived;
+  (* {!Colitur_kernel.Rite.t.transfer_target}'s own second obligation
+     (terminate) reduces, for this liturgical year, to the whole-year
+     placement pass reaching a fixed point at all -- {!Val.run}'s own
+     "unconverged" check is exactly {!Colitur_kernel.Calendar}'s round
+     guard surfaced as a named failure (calendar.ml's own
+     [unconverged_reason]). *)
+  let failures = run 2061 in
+  Alcotest.(check bool) "the whole 2061 liturgical year still reaches a fixed point (no unconverged \
+                         transfer placement anywhere in it)" false
+    (List.exists (fun f -> f.Val.check = "unconverged") failures)
 
 (* Second NEWLY-FOUND defect, pinned the same way (see
    [is_known_holy_family_fallback_gap]'s own citation just above for the
@@ -402,9 +406,9 @@ let suite =
       Alcotest.test_case "2005-2050: transfers reach a fixed point, nothing lost or duplicated" `Slow
         test_transfers_reach_a_fixed_point_2005_2050;
       Alcotest.test_case
-        "KNOWN OPEN DEFECT pinned, not fixed: St Joseph on Palm Sunday 2062 transfers forward, not \
-         backward per Normae n.56(f)" `Quick
-        test_joseph_palm_sunday_2062_known_wrong_forward_transfer;
+        "RESOLVED (W1): St Joseph on Palm Sunday 2062 is anticipated backward to 18 March per Normae \
+         n.56(f), converges, and is recorded as an arrival" `Quick
+        test_joseph_palm_sunday_2062_anticipated_backward;
       Alcotest.test_case
         "NEWLY-FOUND DEFECT pinned, not fixed: Holy Family's Normae n.35(a) 30-December fallback is \
          unreached when Christmas Day is a Sunday (1583)" `Quick