gen-sanctoral-ef, caldata: fix round 3 -- guard coverage, rank pins, name test, Anastasia

Second external re-review returned Spec pass on round 2 (all ten prior
findings addressed, C1/C2 fixed at the root, mechanism claims held
under mutation) and found two blocking items in the I3 guard itself,
plus three recommended hardenings. All verified independently before
fixing.

BLOCKING 1: refYearExplainsAbsence's own doc comment said the id-rank
exclusion covers "6-31 December"; the code implemented `Day() >= 26`.
A typo (26 -> 6), but the wrong kind to have on the one comment this
whole review chain exists to keep honest -- fixed as part of the
rewrite below, not left as an isolated one-character correction.

BLOCKING 2: the guard covered only 6 of the 13 MM-DD values round 1's
own temporal_ef.go changes actually made refYearExplainsAbsence
couple to. Measured, not just reasoned about: running the function's
body against the branch-point engine versus HEAD over the six
reference years flips its verdict on 50 dates across 12-17 through
12-23 (round 1's late-Advent promotion, RG 91 entry 18, a 140-day
change) AND 12-26 through 12-31 (the Nativity-octave promotion, RG
67/68, already guarded). The late-Advent half was unguarded.

No live bug today -- confirmed by regenerating: the ini is
BYTE-IDENTICAL before and after this fix. The only sanctoral entry in
17-23 December is `thomas` (21 December), reached via harvestDate's
`obs` (observed-office) path, never refYearExplainsAbsence's
commemoration-id-trust path at all. Widened the guard to 17-31
December anyway, on the side of the known-safe default
(RankCommemoration) rather than wait for the first real entry in that
range to repeat I3's exact failure shape. The two sub-ranges are NOT
the same strength of evidence, and the doc comment says so explicitly
now: 26-31 December rests on a positive RG 68(d)/(e) citation naming
Thomas Becket and Silvester directly as bare "Commemoratio, no
class"; 17-23 December is a precautionary exclusion with no equivalent
citation -- RG 91 entry 18 ranks the FERIA, not any commemoration
that might coincide with it, and (per defect 2's own finding) a
genuine class-2 feast there would actually WIN against the ferias's
own privilege, so "the day is class-2" does not reliably explain a
class-2 saint's absence here the way it does for the Nativity octave's
two named cases. Recorded as a real asymmetry, not smoothed over.

RECOMMENDED 3: rank = commemoration is now pinned for all four entries
this task's two review rounds touched -- romanus, eusebius-confessor
(C2), thomas-becket, silvester (I3) -- not just their presence and
date. `git grep` before this commit returned zero hits for
thomas-becket or silvester rank anywhere in *_test.go; a future
regeneration could have rewritten either back to class-4, the exact
I3 failure mode, with the suite green throughout.

RECOMMENDED 4: TestTridentineNamesPreservedAcrossRegeneration
rewritten on both axes the review found loose. It counted name.pl
only -- a hardcoded single language, the identical "whitelist of two"
mistake C1's own generator fix exists to stop repeating, just moved
into the test instead. Now DISCOVERS every name.<lang> key actually
present in the data (so a regeneration dropping name.la, or losing a
future name.de, is caught the same way as name.pl) and asserts the
discovered language set is exactly {en, pl} -- not just each count.
The >= 315 floor had 7 entries of slack (the review proved 6 missing
Polish names still passed it); tightened to the exact, independently
verified true count (322), zero slack.

RECOMMENDED 5: the 25 December Anastasia gap and its cause are now
recorded, not just her absence. Two compounding, independently
checked causes: fetchOnce reads only `data[0]`, missalemeum's FIRST
of Christmas Day's three Masses, so a commemoration attached to a
second or third Mass (the calendarium's own "In secunda Missa:
Commemoratio S. Anastasiae Mart.") is structurally invisible to this
harvester for ANY date, not just this one -- but ALSO, checked live,
missalemeum's own "commemorations" list is empty on all three of the
date's records, not just the first, so a fetchOnce rewritten to merge
all of a date's Masses would still not recover her. Documented at
both fetchOnce's own definition and the PRIMARY-SOURCE NOTE, since
the two facts belong to different questions (a real code limitation
worth fixing someday vs. an upstream data gap this generator cannot
close at all).

Witnesses: TestTridentineCommemorationRanksPinned (new),
TestTridentineRomanusAndEusebiusPresent (extended with rank),
TestTridentineNamesPreservedAcrossRegeneration (rewritten). No pre-fix
failure to report for the two new/tightened tests specifically --
these are regression GUARDS against a bug class already fixed in
round 2, not fixes for a currently-red state; their value is in what
they would now catch, not what they catch today. internal/caldata/
tridentine-calendar.ini is NOT regenerated by this commit -- confirmed
byte-identical before and after re-running the generator, exactly as
"no live bug today" predicts.

2 files changed, 174 insertions, 41 deletions

diff --git a/internal/caldata/caldata_test.go b/internal/caldata/caldata_test.go
index 2ba052d..7dbbf84 100644
--- a/internal/caldata/caldata_test.go
+++ b/internal/caldata/caldata_test.go
@@ -246,18 +246,56 @@ func TestTridentineRomanusAndEusebiusPresent(t *testing.T) {
 	} else if rc.Fields["date"] != "08-09" {
 		t.Errorf("romanus: date = %q, want 08-09", rc.Fields["date"])
 	}
-	want := map[string]string{
-		"eusebius-confessor": "08-14", // "S. Eusebii Conf." -- a Confessor
-		"eusebius":           "12-16", // "S. Eusebii Ep. et Mart." -- a Bishop and Martyr, a different person
+	want := map[string]struct{ date, rank string }{
+		// St Eusebius Confessor (14 Aug) and St Romanus (9 Aug, checked
+		// above) are both bare "Commemoratio" in the calendarium (no class
+		// of their own) -- pinned here, not just presence/date, per
+		// TestTridentineCommemorationRanksPinned's own reasoning below:
+		// romanus's own rank is pinned there instead, since that test
+		// groups every commemoration-rank entry this task's fixes touched
+		// in one place. eusebius-confessor is listed here because it is
+		// the OTHER half of this specific test's own slug-disambiguation
+		// story.
+		"eusebius-confessor": {"08-14", "commemoration"}, // "S. Eusebii Conf." -- a Confessor
+		"eusebius":           {"12-16", "class-3"},       // "S. Eusebii Ep. et Mart." -- a Bishop and Martyr, a different person, unaffected by this round
 	}
-	for slug, date := range want {
+	for slug, w := range want {
 		rc, ok := l.Cels[slug]
 		if !ok {
-			t.Errorf("missing %q (%s)", slug, date)
+			t.Errorf("missing %q (%s)", slug, w.date)
 			continue
 		}
-		if rc.Fields["date"] != date {
-			t.Errorf("%s: date = %q, want %q", slug, rc.Fields["date"], date)
+		if rc.Fields["date"] != w.date {
+			t.Errorf("%s: date = %q, want %q", slug, rc.Fields["date"], w.date)
+		}
+		if rc.Fields["rank"] != w.rank {
+			t.Errorf("%s: rank = %q, want %q", slug, rc.Fields["rank"], w.rank)
+		}
+	}
+}
+
+// TestTridentineCommemorationRanksPinned: nothing in this test suite
+// asserted `rank` for thomas-becket, silvester, or romanus before this --
+// `TestTridentineRomanusAndEusebiusPresent` above pinned presence and date
+// for romanus but not rank, and Thomas Becket/Silvester (the I3 fix, RG
+// 68(d)/(e)) had no rank assertion anywhere at all. So a future
+// regeneration could silently rewrite any of the three back to class-4 --
+// the EXACT I3 failure mode -- with the suite green throughout. Pinned
+// together because they are the same failure shape (a bare "Commemoratio"
+// in the calendarium, no class of its own, silently promoted by
+// refYearExplainsAbsence's own coupling to temporal_ef.go, see that
+// function's doc comment) even though they were found in two different
+// review rounds (romanus in C2, Thomas Becket/Silvester in I3).
+func TestTridentineCommemorationRanksPinned(t *testing.T) {
+	l := Tridentine()
+	for _, slug := range []string{"romanus", "thomas-becket", "silvester"} {
+		rc, ok := l.Cels[slug]
+		if !ok {
+			t.Errorf("missing %q", slug)
+			continue
+		}
+		if rc.Fields["rank"] != "commemoration" {
+			t.Errorf("%s: rank = %q, want commemoration", slug, rc.Fields["rank"])
 		}
 	}
 }
@@ -300,24 +338,62 @@ func TestTridentineNoTransferArtifacts(t *testing.T) {
 // reaching mobile.Day(date, "ef", version, "pl") -- a shipped dlectio entry
 // point -- with no error anywhere in the chain.
 //
-// Checks both a broad coverage floor (not exactly 322: a handful of entries
-// added by this same task's own fixes -- St Agnes secundo, St Boniface
-// Martyr, St Evaristus, St Theodore, St Romanus, St Eusebius Confessor --
-// never had a curated Polish name to preserve in the first place, so 322 is
-// not achievable) and one specific, checkable value (the coordinator's own
-// example) so a coverage-only guard cannot itself be satisfied by silently
-// wrong values.
+// CORRECTED after review: the first version of this test counted name.pl
+// only (a hardcoded single language -- the exact "whitelist of two
+// languages" mistake C1's own fix was written to stop repeating, just
+// moved into the test), and its coverage floor (>= 315) had 7 entries of
+// slack -- the review proved dropping 6 entries' Polish names still passed
+// it. Fixed on both axes: the language set is DISCOVERED from the data
+// (every "name.<lang>" key actually present, not a hardcoded list, so a
+// regeneration dropping name.la or introducing a future name.de is
+// checked the same way as name.pl), and the floor for each discovered
+// language is its EXACT true count, not a loose approximation -- verified
+// directly against the branch point (`git show 2386a45:...`) once, by
+// hand, and hardcoded as the answer, not derived at test time from data
+// that could itself be wrong.
+//
+// True counts, independently verified: name.en on every entry (327, all
+// regenerated fresh from missalemeum, including the 5 entries this task's
+// own fixes added); name.pl on exactly 322 (preserved from the branch
+// point; the 5 new entries -- St Agnes secundo, St Boniface Martyr, St
+// Evaristus, St Theodore, St Eusebius Confessor -- never had a curated
+// Polish name to preserve in the first place, so 322, not 327, is the
+// correct target, not a shortfall); no OTHER name.<lang> exists in the
+// branch point at all (name.la is read by the generator but the
+// bootstrapped file has in fact never carried one), so the discovered
+// language set itself must be exactly {en, pl} -- a regeneration that
+// silently introduced or lost an entire language key, not just some
+// values within one, is caught by this assertion, not only by the count.
 func TestTridentineNamesPreservedAcrossRegeneration(t *testing.T) {
 	l := Tridentine()
-	n := 0
+	counts := map[string]int{}
 	for _, rc := range l.Cels {
-		if rc.Fields["name.pl"] != "" {
-			n++
+		for k, v := range rc.Fields {
+			if v == "" || !strings.HasPrefix(k, "name.") {
+				continue
+			}
+			lang := strings.TrimPrefix(k, "name.")
+			counts[lang]++
 		}
 	}
-	if n < 315 {
-		t.Errorf("name.pl coverage = %d entries, want >= 315 (was 322 before any of this task's fixes; a regeneration must preserve it, not merely not-crash)", n)
+	wantLangs := map[string]bool{"en": true, "pl": true}
+	for lang := range counts {
+		if !wantLangs[lang] {
+			t.Errorf("unexpected name.%s present (%d entries) -- discovered language set must be exactly {en, pl}", lang, counts[lang])
+		}
 	}
+	for lang := range wantLangs {
+		if counts[lang] == 0 {
+			t.Errorf("name.%s entirely absent -- discovered language set must be exactly {en, pl}", lang)
+		}
+	}
+	if got := counts["en"]; got != len(l.Cels) {
+		t.Errorf("name.en coverage = %d entries, want %d (every entry, all freshly regenerated)", got, len(l.Cels))
+	}
+	if got := counts["pl"]; got != 322 {
+		t.Errorf("name.pl coverage = %d entries, want exactly 322 (the branch point's own true count; the 5 entries this task added have no curated Polish name to preserve, so more or fewer than 322 is a bug either way)", got)
+	}
+
 	rc, ok := l.Cels["assumption-of-the-blessed-virgin-mary"]
 	if !ok {
 		t.Fatal("missing assumption-of-the-blessed-virgin-mary")
diff --git a/scripts/gen-sanctoral-ef.go b/scripts/gen-sanctoral-ef.go
index 0812472..0a43869 100644
--- a/scripts/gen-sanctoral-ef.go
+++ b/scripts/gen-sanctoral-ef.go
@@ -233,6 +233,19 @@ type mmDay struct {
 	} `json:"sections"`
 }
 
+// fetchOnce fetches one date's proper. KNOWN LIMITATION: missalemeum
+// returns one array element PER MASS on a date with more than one (25
+// December: three, ids ...m1/...m2/...m3) and this reads only `data[0]`,
+// the first -- so a commemoration attached specifically to a second or
+// third Mass (the calendarium's own "In secunda Missa: Commemoratio..."
+// pattern, e.g. St Anastasia on Christmas Day) is structurally invisible
+// to every caller of this function, not just harvestDate's own use of it.
+// See the PRIMARY-SOURCE NOTE above harvestDate for the fuller account --
+// in Anastasia's specific case this is moot (missalemeum's own
+// "commemorations" list is empty on all three of the date's records, not
+// just the first, so fixing this would not by itself recover her), but the
+// limitation is real and would matter for any date whose SECOND or third
+// Mass genuinely does carry a commemoration missalemeum's API records.
 func fetchOnce(date string) (*mmDay, error) {
 	req, _ := http.NewRequest("GET", "https://www.missalemeum.com/en/api/v5/proper/"+date, nil)
 	req.Header.Set("User-Agent", ua)
@@ -350,6 +363,28 @@ func readingsFrom(d *mmDay) (first, gospel string) {
 // wins. (14 August's "St. Eusebius, Conf." and 16 December's "St. Eusebius,
 // Ep. et Mart." are two different people, both genuinely in the calendarium
 // -- see slugOverride below, not a reason to drop either.)
+//
+// Of the four confirming examples above, three (Donatus, Romanus, Eusebius)
+// are fixed by this generator as of this note. **25 December's Anastasia is
+// NOT, and cannot be from this data source alone** -- a different, NOT
+// generator-fixable limitation, recorded so the next person does not
+// mistake her continued absence for an oversight of the fix above (she was
+// also missing at the branch point, so this is not a regression either):
+// the photographic scan places her "In secunda Missa: Commemoratio S.
+// Anastasiae Mart." -- specifically the SECOND of Christmas Day's three
+// Masses (missalemeum's own "2026-12-25" query independently confirmed to
+// return exactly three records, ids ...m1/...m2/...m3). Two compounding
+// problems, not one: fetchOnce (below) reads only `data[0]`, the FIRST
+// Mass, so this harvester cannot structurally see a commemoration attached
+// to a date's second or third Mass at all, for ANY date, not just this
+// one -- but ALSO, checked directly (not merely inferred from the
+// symptom), missalemeum's OWN "commemorations" list is empty on all THREE
+// of the 25 December records, not just the first -- so even a fetchOnce
+// rewritten to merge all of a date's Masses would still not recover her:
+// missalemeum's own data lacks her here, the same "transcription-shaped"
+// gap as Donatus/Romanus/Eusebius, just in the live API rather than the
+// static PDF this time. Not fixed: no source this generator reads carries
+// her.
 
 // slugOverride gives a proper, distinct slug to a small number of
 // commemorations whose title slugifies IDENTICALLY to an unrelated feast on
@@ -388,27 +423,48 @@ var slugOverride = map[string]string{
 //     lower, per the calendarium), but the function does not itself enforce
 //     it, so a future entry could reach it.
 //   - It is a proxy for "why was this saint never observed", not a citation.
-//     6-31 December is excluded from ever trusting the id-derived rank
-//     below, REGARDLESS of what this heuristic would otherwise say, because
-//     RG 68(d)/(e) settles two specific cases (Thomas Becket, 29 December;
-//     Silvester, 31 December) directly: the calendarium's own row names
-//     each a bare "Commemoratio" with NO class of its own, even though the
-//     DAY they fall on is II class (a day within the Nativity Octave). A
-//     rank-blind heuristic reading the day's own class here would (and,
-//     before this exclusion, did) wrongly conclude "the day was strong
-//     enough to explain the demotion, so the id's rank is trustworthy" --
-//     right about the day, wrong about what the rubric actually says about
-//     THESE two named commemorations specifically.
+//     17-31 December is excluded from ever trusting the id-derived rank
+//     below, REGARDLESS of what this heuristic would otherwise say --
+//     covering BOTH of round 1's own late-Advent/Christmas-octave rank
+//     promotions, on two DIFFERENT strengths of evidence, recorded
+//     separately because they are not the same case:
+//     -- 26-31 December (RG 68(d)/(e), a positive citation): "die 29
+//     decembris, fit commemoratio S. Thomae Episcopi et Mart.; die 31
+//     decembris, fit commemoratio S. Silvestri I Papae et Conf." -- the
+//     calendarium names each a bare "Commemoratio" with NO class of its
+//     own, even though the DAY they fall on is II class (within the
+//     Nativity Octave). Two REAL entries (Thomas Becket, Silvester) are
+//     live here; see the call site's own comment.
+//     -- 17-23 December (RG 91 entry 18, a PRECAUTIONARY exclusion, not a
+//     positive citation): round 1 also promoted these late-Advent
+//     ferias from class-3 to class-2 (same lineage, same coupling
+//     shape as 26-31 December's own promotion). No live entry tests
+//     this range today -- the sole sanctoral entry there, `thomas`
+//     (21 December), is an OBSERVED class-2 feast reached via the
+//     harvestDate `obs` path, not this function's id-rank-trust path
+//     at all -- but unlike 26-31 December, there is no RG citation
+//     stating that a saint commemorated here has NO independent class;
+//     RG 91 entry 18 only ranks the FERIA, and (per defect 2's own
+//     finding) a genuine class-2 FEAST commemorated here would
+//     actually WIN against it (entry 16 above entry 18), so "the day
+//     is class-2" does not reliably explain a class-2 saint's absence
+//     the way it does for 26-31 December's own two named cases.
+//     Excluded anyway, on the side of the KNOWN-safe default
+//     (RankCommemoration) rather than risk repeating the identical
+//     failure shape the day this range's own first real entry arrives.
 //
-// The 6-31 December exclusion is also a warning about a structural hazard,
-// not just a one-off fix: this function calls calendar.Compute, i.e. it
-// reads the ENGINE's OWN computed temporal ranks to decide what DATA to
-// generate. A change to temporal_ef.go's ranking (e.g. promoting 26-31
-// December from class-4 to class-2, RG 67/68, landed in this same lineage)
-// can silently flip this function's verdict and rewrite generated data with
-// no code change to this file at all. Any future temporal_ef.go rank change
-// should re-check this function's own boundary cases, not just its own
-// tests.
+// This is also a warning about a structural hazard, not just a boundary
+// fix: this function calls calendar.Compute, i.e. it reads the ENGINE's OWN
+// computed temporal ranks to decide what DATA to generate. A change to
+// temporal_ef.go's ranking (e.g. round 1's own 17-23/26-31 December
+// promotions, RG 91 entry 18 / RG 67-68) can silently flip this function's
+// verdict and rewrite generated data with no code change to this file at
+// all -- confirmed by direct measurement, not just reasoned about: running
+// this function's body against the branch-point engine versus the current
+// one over the six reference years flips the verdict on 50 dates across
+// exactly these 13 MM-DD values (12-17 through 12-23, 12-26 through
+// 12-31), no others. Any future temporal_ef.go rank change should re-check
+// this function's own boundary cases, not just its own tests.
 //
 // harvestDate keeps a saint's commemoration id rank ONLY if every reference
 // year in which it was seen returned true here -- a single false (an
@@ -418,8 +474,9 @@ var slugOverride = map[string]string{
 // RankCommemoration, proving the id's rank is a candidate, not a verdict --
 // see the call site's own comment).
 func refYearExplainsAbsence(date time.Time) bool {
-	if date.Month() == time.December && date.Day() >= 26 && date.Day() <= 31 {
-		// RG 68(d)/(e): "die 29 decembris, fit commemoratio S. Thomae
+	if date.Month() == time.December && date.Day() >= 17 && date.Day() <= 31 {
+		// RG 68(d)/(e) (26-31 Dec) and RG 91 entry 18 (17-23 Dec,
+		// precautionary): "die 29 decembris, fit commemoratio S. Thomae
 		// Episcopi et Mart.; die 31 decembris, fit commemoratio S.
 		// Silvestri I Papae et Conf." -- both named as a bare
 		// "Commemoratio", no class. See this function's own doc comment.