fix(citation): represent chapter-crossing verse ranges (W4)

Parse.verse_range's [last] endpoint gains an optional chapter
(Parse.verse_end: { chapter : int option; verse : verse_num }), so a
hyphen range whose two endpoints lie in different chapters
("1 John 1:5-2:2") can be represented at all. Rejected: a bare
[int option] living alongside [last] as a second field on
verse_range -- that would let 'a chapter with no verse' exist as a
constructible value.

parse_part now splits a part's leading "chapter:" at the FIRST
colon only (not every colon), so the verses side can itself carry a
second colon from a crossing range. parse_range detects a crossing
by checking whether the range's right-hand side contains ':', and
skips the same-chapter descending-range guard for that case (a later
chapter is always "ahead", whatever its own verse numbers are).
Handles the compound shape too -- a crossing range followed by
further, same-chapter verse references in the same comma list
("Matthew 9:35-10:1,5a,6-8") -- since those trailing pieces parse as
ordinary bare verses/ranges, unaffected by the preceding crossing.

Render's one_range renders a crossing [last] through the same
chapter_verse template one_part already uses for the part's own
leading "chapter:verses", so a style that reconfigures the
chapter/verse separator renders a crossing endpoint in that same
convention rather than a hardcoded ':'.

This closes the W4 known-wrong: 41 (now 49, after an intervening
Second-reading extraction) of the OF lectionary's citations printed
unconverted, every one this exact shape. test_citation_coverage_of.ml's
pinned residual is now empty and asserted exactly, over the full
1725-field data/of/lectionary.sexp population, including the round-trip
check (parse -> render -> parse structural equality). test_citation.ml
gains direct parse-suite cases for the basic crossing, the compound
shape, a mid-list crossing, a crossing with a sub-verse letter, and a
malformed-crossing rejection.

data/of/lectionary.sexp is regenerated via its own generator
(tools/bootstrap_lectionary_of.ml, whose own embedded header text is
updated to match); only comment lines change, confirmed by diff --
no lectionary entry differs. test_lectionary_of.ml's whole-file
SHA-256 pin is updated to match.

EF is unaffected: data/ef/ is untouched since v1.0.0, and a direct
byte comparison of `colitur day`/`colitur readings` for 2026, 1583
and 9999 against a git-worktree build of 1c0137d is identical on all
six outputs.

3 files changed, 88 insertions, 65 deletions

diff --git a/test/test_citation.ml b/test/test_citation.ml
index 3cbc8c6..ffbefe8 100644
--- a/test/test_citation.ml
+++ b/test/test_citation.ml
@@ -265,13 +265,18 @@ let suite =
 module P = Colitur_citation.Parse
 
 (* Render a parse back to a debug string so a test can assert shape
-   compactly: "book|chapter:v-v,v-v|chapter:v". *)
+   compactly: "book|chapter:v-v,v-v|chapter:v". A [last] that names its own,
+   later chapter (a chapter-crossing range) shows as "v-C:v", the same
+   "chapter:verse" shape a plain part already uses -- so a crossing is
+   visible in the debug string exactly where the source text put it. *)
 let show (t : P.t) =
   let num (v : P.verse_num) = string_of_int v.P.n ^ v.P.suffix in
   let range (r : P.verse_range) =
     match r.P.last with
     | None -> num r.P.first
-    | Some l -> Printf.sprintf "%s-%s" (num r.P.first) (num l)
+    | Some { P.chapter = None; verse } -> Printf.sprintf "%s-%s" (num r.P.first) (num verse)
+    | Some { P.chapter = Some c; verse } ->
+        Printf.sprintf "%s-%d:%s" (num r.P.first) c (num verse)
   in
   let part (p : P.part) =
     Printf.sprintf "%d:%s" p.P.chapter
@@ -319,6 +324,29 @@ let parse_suite =
     ("four parts",        `Quick, parses "Eccli 24:5; 14:7; 14:9-11; 24:30-31"
                                    "ecclesiasticus|24:5|14:7|14:9-11|24:30-31");
     ("modern name, vulgate id", `Quick, parses "Rev 12:1" "apocalypse|12:1");
+    (* Chapter-crossing ranges (W4): a hyphen range whose two endpoints lie
+       in different chapters -- real shapes from data/of/lectionary.sexp,
+       previously unparseable in their entirety (parse.mli's own
+       [verse_end]). *)
+    ("cross-chapter range", `Quick, parses "1 John 1:5-2:2" "john_1|1:5-2:2");
+    (* The compound shape: a chapter-crossing range FOLLOWED by further,
+       same-chapter verse references in the same comma list -- "5a" and
+       "6-8" belong to chapter 10, the chapter the preceding range just
+       crossed into, never chapter 9. A naive fix that only special-cased
+       "the whole part is one crossing range" breaks exactly here. *)
+    ("cross-chapter then more verses (compound)", `Quick,
+     parses "Matthew 9:35-10:1,5a,6-8" "matthew|9:35-10:1,5a,6-8");
+    (* The crossing is not the first comma-piece, and a normal same-chapter
+       range ("24-25a") sits between two others -- proves the crossing is
+       detected per-piece, not merely at a fixed position in the list. *)
+    ("cross-chapter mid-list", `Quick,
+     parses "2 Samuel 18:9-10,14b,24-25a,30-19:3" "samuel_2|18:9-10,14b,24-25a,30-19:3");
+    (* Verse sub-letters survive on a chapter-crossing endpoint too. *)
+    ("cross-chapter with sub-verse letter", `Quick,
+     parses "Genesis 1:20-2:4a" "genesis|1:20-2:4a");
+    (* A malformed crossing (no verse named after the second chapter) is
+       rejected, not silently misread. *)
+    ("rejects malformed cross-chapter", `Quick, rejects "Luke 1:5-2:");
     (* Ordinals 3 and 4 must survive [split_book]'s leading-digit scan.
        Narrowing its '1'..'4' range to '1'..'3' passes every OTHER case in
        this suite silently, while seven real citations depend on it. A
diff --git a/test/test_citation_coverage_of.ml b/test/test_citation_coverage_of.ml
index 7badd96..af8d161 100644
--- a/test/test_citation_coverage_of.ml
+++ b/test/test_citation_coverage_of.ml
@@ -3,48 +3,46 @@
 (* Fix wave I2 (final-review.md, 2026-08-25-colitur-of-phases-3-5): the OF
    counterpart of test_citation_coverage.ml. That file's own header states
    the rule this one follows: "every citation the engine can emit must
-   parse". EF reaches zero unconverted citations; OF does not, and this
-   file is the honest record of exactly where it falls short, mirroring the
-   review's own explicit instruction -- "enumerate them rather than
-   lowering the bar to fit".
+   parse". At the time this file was first written, EF reached zero
+   unconverted citations and OF did not; this file was the honest record of
+   exactly where it fell short.
 
-   MEASURED, not estimated: before this fix wave, 259 of 730 citation
-   fields in `colitur readings --rite of 2026` (35%) printed unconverted --
-   345 distinct book spellings unregistered ({!Colitur_citation.Book}),
-   plus verse sub-letter markers ("11a") and a handful of single-chapter-
-   book misreads ({!Colitur_citation.Book.is_single_chapter}) the parser
-   could not read at all. All of that is now fixed at the source (book.ml's
-   own [of_lectionary_table] and parse.ml's own [verse_num]/[single_chapter]
-   additions), closing every genuinely fixable shape. What remains, walking
-   the FULL data/of/lectionary.sexp (not merely what one civil year happens
-   to observe -- a stricter check than EF's own year-walk, and the same one
-   tools/bootstrap_lectionary_of.ml's own generated header now runs and
-   discloses): 41 distinct references, out of 1542 total, EVERY ONE a
-   hyphenated range crossing a chapter boundary ("2:29-3:6", "1 John") --
-   a {!Colitur_citation.Parse.t} shape [part]/[verse_range] do not
-   represent (a [part] is one chapter and a list of within-chapter ranges;
-   representing a cross-chapter range would need restructuring that type
-   and {!Colitur_citation.Render}'s own consumption of it), deliberately
-   not built this task.
+   MEASURED, not estimated: before fix wave I2, 259 of 730 citation fields
+   in `colitur readings --rite of 2026` (35%) printed unconverted -- 345
+   distinct book spellings unregistered ({!Colitur_citation.Book}), plus
+   verse sub-letter markers ("11a") and a handful of single-chapter-book
+   misreads ({!Colitur_citation.Book.is_single_chapter}) the parser could
+   not read at all. All of that was fixed at the source (book.ml's own
+   [of_lectionary_table] and parse.ml's own [verse_num]/[single_chapter]
+   additions). What remained, walking the FULL data/of/lectionary.sexp (not
+   merely what one civil year happens to observe -- a stricter check than
+   EF's own year-walk, and the same one tools/bootstrap_lectionary_of.ml's
+   own generated header runs and discloses): a hyphenated verse range whose
+   two endpoints lie in different chapters ("2:29-3:6", "1 John") -- a
+   {!Colitur_citation.Parse.t} shape [part]/[verse_range] did not
+   represent. Fix wave I7 (2026-08-26) grew the population from 1542 to
+   1725 citation fields (a new Second-reading extraction) and the residual
+   from 41 to 49 references, same shape, no new failure mode.
 
-   UPDATED, fix wave I7 (final-review.md, 2026-08-25-colitur-of-phases-3-5,
-   2026-08-26): {!Rite.t.citation_shapes} (kernel change) plus
-   tools/bootstrap_lectionary_of.ml's own [cite] now also extract a Second
-   reading on every Sunday-cycle (Sunday/solemnity) entry, so the total grew
-   1542 -> 1725 (+183, exactly the count data/of/lectionary.sexp's own
-   header COVERAGE (5) now discloses) and the unparseable residual grew
-   41 -> 49: the same chapter-crossing shape, now also hit by 8 of the 183
-   new Second-reading references. Not a new failure MODE, just a bigger
-   population drawing from the same, already-understood one -- confirmed
-   below by [looks_chapter_crossing], unchanged, still holding for all 49.
+   CLOSED (W4, this task): {!Colitur_citation.Parse.verse_end} now lets a
+   range's [last] endpoint name its own, later chapter, and
+   {!Colitur_citation.Render}'s [one_range] renders it back through the
+   part's own [chapter_verse] template. Every one of the 49 pinned
+   references now parses and round-trips -- including the COMPOUND shape
+   ("Matthew 9:35-10:1,5a,6-8": a crossing range followed by further,
+   same-chapter verse references in the same comma list) and a crossing
+   that is not the list's first or only piece ("2 Samuel
+   18:9-10,14b,24-25a,30-19:3"). The residual below is therefore now EMPTY,
+   over the full 1725-field population -- measured, not merely no longer
+   observed on a sample.
 
-   This file pins that residual EXACTLY -- not merely "at least these fail"
-   -- so it fails loudly in BOTH directions: a NEW unconverted reference
-   appearing (a regression) and one of these 49 starting to convert (this
-   list going stale, e.g. because a future task builds cross-chapter
-   support) are both caught, the same discipline every allow-list in this
-   project's suite already follows (`expected_rows`, `expected_bands`, the
-   litcal/missalemeum comparators). *)
+   The pinning discipline is kept rather than deleted: this file still
+   fails LOUDLY, in both directions, if the unconverted set ever stops
+   being empty -- a real regression (a genuinely new unparseable shape) and
+   a silently corrupted lectionary file are both caught by the same
+   mechanism every allow-list in this project's suite already follows
+   (`expected_rows`, `expected_bands`, the litcal/missalemeum
+   comparators). *)
 
 let lectionary_path = "../data/of/lectionary.sexp"
 
@@ -53,25 +51,12 @@ let lectionary () =
   | Ok l -> l
   | Error e -> Alcotest.failf "%s: %s" lectionary_path e
 
-(* The exact, pinned residual -- copied from data/of/lectionary.sexp's own
-   generated "Citations that do not parse" listing, itself produced by the
-   SAME {!Colitur_citation.Parse.parse} call this test makes, so the two
-   can never silently drift apart without one of them visibly failing to
-   regenerate/re-test clean. *)
-let known_cross_chapter_residual =
-  [ "1 Corinthians 10:31-11:1"; "1 Corinthians 12:31-13:13"; "1 John 1:5-2:2"; "1 John 2:29-3:6";
-    "1 John 3:22-4:6"; "1 John 4:19-5:4"; "1 Thessalonians 3:12-4:2"; "2 Corinthians 3:15-4:1,3-6";
-    "2 Corinthians 4:13-5:1"; "2 Samuel 18:9-10,14b,24-25a,30-19:3"; "2 Thessalonians 1:11-2:2";
-    "2 Thessalonians 2:16-3:5"; "2 Timothy 3:14-4:2"; "Colossians 1:24-2:3"; "Ecclesiastes 11:9-12:8";
-    "Ephesians 4:30-5:2"; "Ephesians 4:32-5:8"; "Exodus 11:10-12:14"; "Exodus 14:21-15:1";
-    "Ezekiel 2:8-3:4"; "Galatians 4:22-24,26-27,31-5:1"; "Genesis 1:1-2:2"; "Genesis 1:20-2:4a";
-    "Habakkuk 1:12-2:4"; "Hebrews 7:25-8:6"; "Isaiah 52:13-53:12"; "Isaiah 8:23b-9:3";
-    "James 3:16-4:3"; "John 15:26-16:4a"; "John 18:1-19:42"; "Jonah 1:1-2:2,11"; "Luke 7:36-8:3";
-    "Malachi 1:14b-2:2b,8-10"; "Mark 2:23-3:6"; "Mark 8:34-9:1"; "Matthew 10:34-11:1";
-    "Matthew 18:21-19:1"; "Matthew 9:35-10:1,5a,6-8"; "Matthew 9:36-10:8";
-    "Numbers 13:1-2,25-14:1,26a-29a,34-35"; "Philippians 3:17-4:1"; "Revelation 20:1-4,11-21:2";
-    "Sirach 27:30-28:7"; "The Acts 12:24-13:5a"; "The Acts 17:15,22-18:1"; "The Acts 7:51-8:1a";
-    "Wisdom 11:22-12:2"; "Wisdom 2:23-3:9"; "Wisdom 7:22b-8:1" ]
+(* The exact, pinned residual -- EMPTY since W4 closed the cross-chapter gap
+   (see this file's own header). Kept as a named binding, not inlined as
+   [[]] at each use site, so a future partial regression has one place to
+   list what it could not fix, exactly as the pre-W4 version of this file
+   did for the 49 references it once named here. *)
+let known_cross_chapter_residual = []
 
 (* Every failure this walk finds really is a chapter-crossing hyphen range
    (a "-" whose right side itself contains a ":") -- checked mechanically,
@@ -107,8 +92,8 @@ let test_of_citations_convert_or_are_the_known_cross_chapter_residual () =
   Alcotest.(check int) "1725 citation fields in data/of/lectionary.sexp today -- if this changes, the \
                         pinned residual below may need updating too, not just this count"
     1725 !total;
-  Alcotest.(check (list string)) "the unconverted set is EXACTLY the known cross-chapter residual, no \
-                                  more and no fewer"
+  Alcotest.(check (list string)) "the unconverted set is EXACTLY the known cross-chapter residual (empty, \
+                                  since W4) -- no more and no fewer"
     (List.sort compare known_cross_chapter_residual) bad;
   List.iter
     (fun r ->
@@ -119,7 +104,9 @@ let test_of_citations_convert_or_are_the_known_cross_chapter_residual () =
 
 (* Everything OUTSIDE the pinned residual round-trips through Sigla, the
    same structural (not textual) comparison test_citation_coverage.ml's own
-   EF version makes, and for the identical reason its own header gives. *)
+   EF version makes, and for the identical reason its own header gives.
+   Since the residual is now empty, this exercises the FULL 1725-field
+   population, cross-chapter shapes included. *)
 let test_round_trip_outside_the_residual () =
   let sg =
     Colitur_citation.Sigla.make ~style:Colitur_citation.Render.default_style
diff --git a/test/test_lectionary_of.ml b/test/test_lectionary_of.ml
index 706ee15..dc02176 100644
--- a/test/test_lectionary_of.ml
+++ b/test/test_lectionary_of.ml
@@ -178,7 +178,15 @@ let sha256_of_file path =
    don't transcribe it from generator stdout" discipline
    test_calendar_of_data.ml's own [test_sha256] already follows. *)
 let test_sha256_pinned () =
-  (* Fix wave I8 (final-review.md, 2026-08-25-colitur-of-phases-3-5): the
+  (* W4 (this task): header-only change, same discipline as I2 below --
+     {!Colitur_citation.Parse.verse_end} closed the chapter-crossing gap,
+     so the residual the header's own COVERAGE (4) and the trailing
+     "Citations that do not parse" listing disclose went from 49 to 0. No
+     lectionary ENTRY changed (`git diff --stat` on this file shows only
+     comment lines). Re-derived with `sha256sum data/of/lectionary.sexp`,
+     same discipline.
+
+     Fix wave I8 (final-review.md, 2026-08-25-colitur-of-phases-3-5): the
      hash changed again -- ONE genuine new entry this time ("of-nativity",
      the Christmas Day Mass hand-authored from OLM 1981; see
      test_coverage_2026's own citation), not header-only as the I2 change
@@ -193,7 +201,7 @@ let test_sha256_pinned () =
      and test_citation_coverage_of.ml), and the header's "Epistle" wording
      is fixed to "First". Re-derived the same way, same discipline. *)
   Alcotest.(check string) "data/of/lectionary.sexp SHA-256"
-    "0d82069e879d8312943dee543aabc23208265534fcf7edd85c3d9137c2f4581d" (sha256_of_file lectionary_path)
+    "788ddc6d2d783286b6e64882a5372ae213cd5462c887e68aa2bea22388cdb4f7" (sha256_of_file lectionary_path)
 
 (* lectio's OWN of-lectionary.ini SHA-256, pinned inside data/of/
    lectionary.sexp's own provenance header (tools/bootstrap_lectionary_of