PDF export is printed by the browser now

Exporting a document to PDF used to draw the page with a separate PDF engine. It is now printed by the same headless browser that already produced pixel-faithful deck exports, so the portal has one PDF renderer instead of two.

Everything the export promised still arrives: the branded cover page, the table of contents, the running header and footer, the page numbers, and the page-break rules that start each chapter on a fresh page. Two things are better than before — a table that spans pages now repeats its header row, and a heading is no longer left stranded alone at the foot of a page.

One thing was gone, and it is worth naming: entries in the table of contents no longer showed the page they point at. They were still real links — clicking one jumps to that section — but the ..... 12 column was not there. Browsers provide no way to read a target's page number while laying the document out, and guessing it would risk printing numbers that are quietly wrong, which is worse than not printing them.

Since resolved. The column came back the next day, without any guessing: see The contents list tells you the page again.