Fonts: "sans serif" is a sans name, not a serif one

Review pointed out that the CSS generic "sans-serif" matched the bare
"serif" needle first and classified as serif. The same ordering hit real
family names: "Microsoft Sans Serif", "MS Sans Serif" and "Lucida Sans
Serif" are sans faces that a Word import can name, and epub's own
generic_family already maps Microsoft Sans Serif to sans-serif.

So rather than special-casing the one hyphenated spelling, the bare
"serif" word now sorts after the sans needles. The Korean serif names
stay ahead of everything, which is what keeps "Noto Sans Myungjo" and
"Myungjo Gothic" serif; the test pins those two along with both "sans
serif" spellings.

crates/fonts/src/fontdb/fallback.rs

@@ -118,9 +118,9 @@ pub(super) const KOREAN_SANS_FALLBACKS: &[&str] = &[
 pub fn sans_by_name(family: &str) -> Option<bool> {
     let lower = family.to_lowercase();
     let has = |needles: &[&str]| needles.iter().any(|n| lower.contains(n));
-    // Serif first: "Myungjo Gothic" does not exist, but a name holding both reads as the serif
-    // it names, and "Sans" inside "Noto Sans Myungjo" would otherwise win.
-    if has(&["batang", "myungjo", "myeongjo", "gungsuh", "serif", "\u{BC14}\u{D5D5}", "\u{BA85}\u{C870}", "\u{AD81}\u{C11C}"]) {
+    // A Korean serif name comes first: "Myungjo Gothic" does not exist, but a name holding both
+    // reads as the serif it names, and "Sans" inside "Noto Sans Myungjo" would otherwise win.
+    if has(&["batang", "myungjo", "myeongjo", "gungsuh", "\u{BC14}\u{D5D5}", "\u{BA85}\u{C870}", "\u{AD81}\u{C11C}"]) {
         return Some(false);
     }
     if has(&[
@@ -138,6 +138,11 @@ pub fn sans_by_name(family: &str) -> Option<bool> {
     ]) {
         return Some(true);
     }
+    // Only now the bare word, which "sans serif" and "sans-serif" end with and do not mean
+    // (noted in review): "Microsoft Sans Serif" is a sans face, and so is the CSS generic.
+    if lower.contains("serif") {
+        return Some(false);
+    }
     None
 }
 

crates/fonts/src/fontdb/tests/fallback.rs

@@ -176,6 +176,26 @@ fn a_korean_family_name_says_sans_or_serif() {
     }
 }
 
+/// "sans serif" ends with the word it is not. The bare "serif" needle matched first, so
+/// "Microsoft Sans Serif" -- and the CSS generic -- classified as serif and a missing Hangul
+/// glyph took the serif-first chain (noted in review of #1692).
+#[test]
+fn a_sans_serif_name_is_not_read_as_serif() {
+    for sans in ["sans-serif", "Microsoft Sans Serif", "MS Sans Serif", "Lucida Sans Serif"] {
+        assert_eq!(sans_by_name(sans), Some(true), "{sans} is sans");
+    }
+    // The bare word still decides a name that holds nothing else.
+    for serif in ["serif", "Source Serif 4", "DejaVu Serif", "Noto Serif KR"] {
+        assert_eq!(sans_by_name(serif), Some(false), "{serif} is serif");
+    }
+    // And a Korean serif name still outranks "Sans", which is why it is tested first of all.
+    assert_eq!(sans_by_name("Noto Sans Myungjo"), Some(false));
+    assert_eq!(sans_by_name("Myungjo Gothic"), Some(false));
+    // Neither word present, so the caller still decides.
+    assert_eq!(sans_by_name("monospace"), None);
+    assert_eq!(sans_by_name("system-ui"), None);
+}
+
 /// The reported bug is the order of the chain: on Windows the first installed entry of the
 /// serif-first chain is Batang, so every missing Dotum face came out serif (#1692).
 #[test]