Codex Desktop PDF preview drops Chinese labels for PDFs using unembedded KaiTi/SimHei fonts
Environment
- Codex Desktop: 26.513.4821.0
- OS: Windows
- Surface: Codex Desktop built-in artifact/PDF preview
Problem
A PDF renders correctly in external Windows PDF viewers, but Codex Desktop's built-in PDF preview drops some Chinese labels. Table lines, QR code, amounts, and many field values remain visible, while some Chinese field labels disappear.
No personally identifiable document or screenshot is attached here because the original file contains private invoice/travel information.
Observed PDF characteristics
The missing labels appear to use unembedded Chinese fonts:
/KaiTi,/UniGB-UCS2-H, noToUnicode/SimHei,/UniGB-UCS2-H, noToUnicode
Other embedded/subset fonts render correctly.
The same Windows machine has Chinese fonts installed, including simhei.ttf and simkai.ttf, and external PDF viewers render the file correctly.
Codex preview behavior observed locally
The Codex Desktop bundle appears to use PDF.js for the preview. The preview code path appears to call PDF.js roughly as getDocument({ data }), without obvious explicit cMapUrl, standardFontDataUrl, or related font/CMap configuration in the preview wrapper.
Expected behavior
Codex Desktop PDF preview should match external Windows PDF viewers / browser PDF viewer behavior and should not drop Chinese field labels.
Actual behavior
Chinese field labels using unembedded /KaiTi or /SimHei fonts disappear in the Codex Desktop PDF preview, while table structure, QR code, numeric values, and other text remain visible.