PDF import: a font the file doesn't carry keeps the style its name says

A PDF that names a font without embedding it (here a standard Helvetica)
is drawn with a stand-in, so comparing its glyphs with the installed
family's styles picks whichever style happens to look most like the
stand-in: with a large Helvetica family installed, Helvetica became
Helvetica CE Medium (tests_import_fidelity::text_becomes_point_type
failed on such a Mac). Fonts without a font program now keep the face
their name resolves to; embedded fonts are still compared as before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CC5DiGJEZ5ADaQMBEQk4W1

crates/pdf/src/import.rs

@@ -21,7 +21,7 @@ use vectorcraft_geom::{FillRule, PathData};
 
 use crate::import_color::{Colors, Native};
 use crate::import_mask::{MaskSpec, contains, is_rectangle, luminance, mask_spec, white_cover};
-use crate::import_scan::{MAX_NESTING, Ocgs, Scan, all_on, hides_forms, scan_page, tag_key};
+use crate::import_scan::{FontNames, MAX_NESTING, Ocgs, Scan, all_on, hides_forms, scan_page, tag_key};
 use crate::import_shading::{clipped, extend_clip, fold_stop_opacity, mesh_shading, shading_gradient};
 use crate::import_text::{Families, LineFacts, Look, Placement, TextLine, Upright};
 use crate::{CropTo, ImportOptions, ImportReport, PdfError, TextAs};
@@ -692,8 +692,9 @@ struct Builder<'p> {
     /// Mask keys of CMYK images whose mask turned out to hide nothing.
     opaque: HashSet<u128>,
     warnings: Vec<String>,
-    /// Fonts by cache key → base font name (from [`scan_page`]).
-    fonts: HashMap<u128, String>,
+    /// Fonts by cache key → base font name and whether the file carries its glyphs (from
+    /// [`scan_page`]).
+    fonts: FontNames,
     font_names: HashMap<u128, FontInfo>,
     families: Option<Families>,
     missing_fonts: Vec<String>,
@@ -1346,7 +1347,7 @@ impl<'p> Builder<'p> {
             return n.clone();
         }
         let data = o.font_data();
-        let name = self.fonts.get(&key).cloned().or_else(|| data.as_ref().and_then(|d| d.postscript_name.clone()));
+        let name = self.fonts.get(&key).map(|(n, _)| n.clone()).or_else(|| data.as_ref().and_then(|d| d.postscript_name.clone()));
         let (weight, italic) = data.as_ref().map_or((None, false), |d| (d.weight, d.is_italic));
         // A subset's six-letter tag.
         let ps = name.as_deref().map(|n| match n.split_once('+') {
@@ -1429,7 +1430,12 @@ impl<'p> Builder<'p> {
         let upright = width.is_some_and(|w| w.abs() <= 1.0);
         let top = m * kurbo::Point::new(f64::from(width.filter(|w| *w > 1.0).unwrap_or(1000.0)) * 0.5, 880.0);
         let mut info = self.font_name(key, o);
-        if let Some(face) = &info.face {
+        // A font the file doesn't carry is drawn with a stand-in: its glyphs can't tell which
+        // style of the installed family it is, so the name decides.
+        let embedded = self.fonts.get(&key).is_none_or(|(_, e)| *e);
+        if let Some(face) = &info.face
+            && embedded
+        {
             let decided = match self.matched.get(&key) {
                 Some(m) => m.clone(),
                 // A glyph that can't be compared (several characters, a vertical form) decides

crates/pdf/src/import_scan.rs

@@ -248,7 +248,7 @@ struct Walker<'w, 'p> {
     /// The groups that are off for the interpreter (none once [`all_on`] turned them on).
     off: HashSet<ObjectIdentifier>,
     xref: &'p hayro_syntax::xref::XRef,
-    fonts: &'w mut HashMap<u128, String>,
+    fonts: &'w mut FontNames,
     visible: Vec<bool>,
     ops: usize,
     max_ops: usize,
@@ -296,7 +296,7 @@ impl Walker<'_, '_> {
             if let Some(f) = res.fonts.get::<Dict<'_>>(key.as_ref())
                 && let Some(name) = f.get::<Name<'_>>(b"BaseFont")
             {
-                self.fonts.entry(f.cache_key()).or_insert_with(|| String::from_utf8_lossy(name.as_ref()).into_owned());
+                self.fonts.entry(f.cache_key()).or_insert_with(|| (String::from_utf8_lossy(name.as_ref()).into_owned(), embeds_glyphs(&f)));
             }
         }
     }
@@ -372,13 +372,32 @@ impl Walker<'_, '_> {
     }
 }
 
+/// The fonts of a document by cache key: base font name, and whether the file carries the glyphs
+/// ([`embeds_glyphs`]).
+pub(crate) type FontNames = HashMap<u128, (String, bool)>;
+
+/// Whether font `f` carries its own glyphs: a font program in its descriptor (a composite font's in
+/// its descendant's), or a Type 3 font's glyph procedures. A font that doesn't is drawn with a
+/// stand-in, whose glyphs say nothing about which installed face the document means.
+fn embeds_glyphs(f: &Dict<'_>) -> bool {
+    if f.get::<Name<'_>>(b"Subtype").is_some_and(|s| s.as_ref() == b"Type3") {
+        return true;
+    }
+    // A font program is a stream, so always an indirect object.
+    let program = |d: &Dict<'_>| {
+        d.get::<Dict<'_>>(b"FontDescriptor").is_some_and(|fd| [&b"FontFile"[..], b"FontFile2", b"FontFile3"].iter().any(|k| fd.get_ref(*k).is_some()))
+    };
+    program(f) || f.get::<Array<'_>>(b"DescendantFonts").and_then(|a| a.iter::<Dict<'_>>().next()).is_some_and(|d| program(&d))
+}
+
 /// Walk `page` as the interpreter will: `all_on` when [`all_on`] turned every group on. Fonts
-/// found are noted in `fonts` (their cache key → base font name).
-pub(crate) fn scan_page(page: &Page<'_>, ocgs: &mut Ocgs, all_on: bool, fonts: &mut HashMap<u128, String>) -> Scan {
+/// found are noted in `fonts` (their cache key → base font name, and whether the file carries its
+/// glyphs).
+pub(crate) fn scan_page(page: &Page<'_>, ocgs: &mut Ocgs, all_on: bool, fonts: &mut FontNames) -> Scan {
     walk_page(page, ocgs, all_on, fonts).out
 }
 
-fn walk_page<'w, 'p>(page: &Page<'p>, ocgs: &'w mut Ocgs, all_on: bool, fonts: &'w mut HashMap<u128, String>) -> Walker<'w, 'p> {
+fn walk_page<'w, 'p>(page: &Page<'p>, ocgs: &'w mut Ocgs, all_on: bool, fonts: &'w mut FontNames) -> Walker<'w, 'p> {
     let off = if all_on { HashSet::new() } else { ocgs.off.clone() };
     let mut w = Walker { ocgs, off, xref: page.xref(), fonts, visible: vec![], ops: 0, max_ops: max_ops(), out: Scan::default(), hidden_form: false };
     w.walk(page.typed_operations(), page.resources(), 0);