Look up symbolic TrueType codes in code-keyed cmap subtables

[Velli20]

Summary

A simple-font code without a /Differences entry was resolved only by Unicode, through skrifa's charmap, which selects Unicode subtables only and ignores the (3,0) Symbol and (1,0) Macintosh Roman subtables. A symbolic TrueType subset (descriptor flag 3, no /Encoding) whose only cmap is (1,0) therefore resolved no glyphs, and every character was drawn with a fallback face instead of the embedded program.

PDF 32000 §9.6.6.4 maps a symbolic TrueType font's codes through its (3,0) or (1,0) subtable directly. This change:

• adds FontFace::glyph_for_code(u8) (default None, so Type 1, CFF, Type 3 and Type 0 faces are unaffected);
• implements it for TrueType faces: (3,0) with the code, then 0xF000 | code, then (1,0) with the code; glyph 0 counts as a miss;
• in select_glyph, without an explicit selector, tries the code before Unicode for symbolic faces and after Unicode for other faces. Explicit glyph IDs, names and CIDs are unchanged.

Test plan

• A document with embedded symbolic Arial/Calibri subsets carrying only a (1,0) format 6 cmap: before, hundreds of glyph fallbacks per page; after, none, and the text is drawn with the embedded outlines
• PDFium corpus run on main and on this branch: identical document and page statuses, and no change in any page's mismatch metric
• cargo fmt --check, cargo clippy --all --workspace
• cargo test -p pdf-font -p pdf-text-engine -p pdf-canvas -p pdf-renderer

🤖 Generated with Claude Code