readings: PartIDs(new) stops declaring aklamacja, which the engine never emits

caldata.go:42 parses only first/psalm/second/gospel out of the lectionary
data, so no OF reading ever carries Part == "acclamation". PartIDs("new")
listed aklamacja anyway, so the app built a checkbox from it that filters
an ID that never appears -- a dead control, same defect class the app
previously shipped for the whole 1962 form.

ofPartOrder stays the full five-ID set: it also drives
render.LocalizeHeading's label matching, where a scraped heading can still
read "Aklamacja" even though this engine's own readings never produce
that section. PartIDs now draws from a new, narrower
ofEmittedPartOrder instead.

Rewrote TestPartLabelsMatchesWhatTheEngineEmits's OF half: it compared
PartLabels("of") to a hand-copied duplicate of ofPartOrder, asserting a
declaration against itself, which cannot fail for this class of bug. It
now sweeps Days over calendar year 2026 and asserts PartLabels("of")
matches the observed part-ID set exactly (0.5s). Confirmed red against
the pre-fix code, green after.

2 files changed, 14 insertions, 2 deletions

diff --git a/internal/readings/offline.go b/internal/readings/offline.go
index 608abd8..be7c430 100644
--- a/internal/readings/offline.go
+++ b/internal/readings/offline.go
@@ -98,6 +98,18 @@ var (
 	efPartOrder = []string{"epistola", "evangelium"}
 )
 
+// ofEmittedPartOrder is the subset of ofPartOrder the offline engine can
+// actually produce, in display order. It excludes "aklamacja":
+// internal/caldata/caldata.go:42 parses only "first", "psalm", "second" and
+// "gospel" out of the lectionary data, so no computed OF reading ever carries
+// Part == "acclamation" and ofPart's "aklamacja" mapping above is never
+// reached. ofPartOrder stays the full five-ID set on purpose -- it also
+// drives render.LocalizeHeading's *label* matching, where a scraped heading
+// can still read "Aklamacja" even though this engine's own readings never
+// produce that section -- so PartIDs, which promises IDs an app can filter
+// on, needs this narrower list rather than reusing or shrinking ofPartOrder.
+var ofEmittedPartOrder = []string{"pierwsze_czytanie", "psalm", "drugie_czytanie", "ewangelia"}
+
 // PartIDs returns the part IDs the given lectionary can emit, in display order.
 // lect takes config.Config.Lectionary's values: "new" or "traditional". An
 // unknown lectionary returns nil. Callers that build per-part UI (the dlectio
@@ -106,7 +118,7 @@ var (
 func PartIDs(lect string) []string {
 	switch lect {
 	case "new":
-		return append([]string(nil), ofPartOrder...)
+		return append([]string(nil), ofEmittedPartOrder...)
 	case "traditional":
 		return append([]string(nil), efPartOrder...)
 	}
diff --git a/internal/readings/partids_test.go b/internal/readings/partids_test.go
index 1ba9373..fdbd816 100644
--- a/internal/readings/partids_test.go
+++ b/internal/readings/partids_test.go
@@ -35,7 +35,7 @@ func TestDayInfoCarriesRank(t *testing.T) {
 // them instead, seven of the nine 1962 IDs were wrong and the epistle's
 // checkbox did nothing.
 func TestPartIDs(t *testing.T) {
-	wantNew := []string{"pierwsze_czytanie", "psalm", "drugie_czytanie", "aklamacja", "ewangelia"}
+	wantNew := []string{"pierwsze_czytanie", "psalm", "drugie_czytanie", "ewangelia"}
 	wantOld := []string{"epistola", "evangelium"}
 	if got := PartIDs("new"); !equalSlice(got, wantNew) {
 		t.Errorf("PartIDs(new) = %v, want %v", got, wantNew)