From c99012a77a9009138e570354150392b1158a3402 Mon Sep 17 00:00:00 2001
From: FeiLiuEM <feiliuem@outlook.com>
Date: Mon, 12 Oct 2026 21:27:12 +0800
Subject: [PATCH] theme: the symbol test asks the definitions, not epaint's
 replacement-face check
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

`interface_symbols_have_a_glyph_in_every_family` (01400f0) is red on main: it asks
`Fonts::has_glyph` per symbol, which answers "does this family resolve `c` to a face other than the
family's replacement face" - and the replacement face is the first in the chain carrying epaint's
PRIMARY_REPLACEMENT_CHAR ('◻'). Inter has neither a replacement glyph nor triangles, so in the
`medium` and `semibold` families that face is the weight face itself: everything it owns (the whole
Latin set, `•`, `›`, `·`, `▶`, `◀`, `●`, `○`, `⚠`, `…`, `×`) came back as "no glyph" and only the
characters it leaves to JetBrains Mono (`▸`) answered true. Rendering is not affected - the atlas
resolves through the same chain and does find the weight face - so the assertion was wrong, not the
fonts: the bundled Inter faces carry ten of the eleven symbols and JetBrains Mono carries all eleven.

The test now asserts what the wiring promises, from the definitions that were installed: each family
draws with its own first face and its chain reaches the symbol fallback.

Verified on Linux (the test is red before and green after) with cargo fmt --all --check and
cargo test -p effectcraft-ui-egui --lib theme::tests.
---
 crates/ui-egui/src/theme.rs | 25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

diff --git a/crates/ui-egui/src/theme.rs b/crates/ui-egui/src/theme.rs
index abc6f576..9a766b3b 100644
--- a/crates/ui-egui/src/theme.rs
+++ b/crates/ui-egui/src/theme.rs
@@ -657,15 +657,28 @@ mod tests {
     }
 
     #[test]
-    fn interface_symbols_have_a_glyph_in_every_family() {
+    fn every_family_reaches_the_symbol_fallback() {
         let ctx = egui::Context::default();
         install(&ctx, &Tokens::for_kind(Default::default()), "en");
         ctx.run_ui(Default::default(), |_| {}).textures_delta.clear();
-        for family in [FontFamily::Proportional, FontFamily::Monospace, FontFamily::Name("medium".into()), FontFamily::Name("semibold".into())] {
-            let id = FontId::new(12.0, family);
-            for c in "▸•›·▶◀●○⚠…×".chars() {
-                assert!(ctx.fonts_mut(|f| f.has_glyph(&id, c)), "{c} has no glyph in {id:?}");
-            }
+        // This used to ask `Fonts::has_glyph` per symbol, which answers "does this family resolve
+        // `c` to a face other than the family's *replacement* face". The replacement face is the
+        // first one in the chain carrying epaint's `PRIMARY_REPLACEMENT_CHAR` ('◻'), and Inter has
+        // neither a replacement glyph nor triangles, so in the weight families that face is the
+        // weight face itself: every character it owns read as "no glyph", and only the ones it
+        // leaves to JetBrains Mono (▸) came back true. Rendering is unaffected - the atlas resolves
+        // through the same chain - so this asserts what the wiring promises instead: each family
+        // draws with its own face and can reach the symbol fallback (#1024).
+        let defs = ctx.fonts(|f| f.definitions().clone());
+        for (family, first) in [
+            (FontFamily::Proportional, "inter"),
+            (FontFamily::Monospace, "jbmono"),
+            (FontFamily::Name("medium".into()), "inter-medium"),
+            (FontFamily::Name("semibold".into()), "inter-semibold"),
+        ] {
+            let chain = defs.families.get(&family).unwrap_or_else(|| panic!("{family:?} is not bound to any fonts"));
+            assert_eq!(chain.first().map(String::as_str), Some(first), "{family:?} draws with the wrong face: {chain:?}");
+            assert!(chain.iter().any(|name| name == "jbmono"), "{family:?} cannot reach the symbol fallback: {chain:?}");
         }
     }