fix(print): make an editor-mode PDF export contain the document

[PathGao]

3 / 5 of a stack. Base: #406.

Exporting to PDF from plain edit mode produced a blank page. Two independent causes — and fixing either one alone is worse than fixing neither.

Cause 1: a specificity fight the print sheet cannot win

The print sheet hides the editor pane and restores the preview with .pane.viewer-pane { width: 100% !important } — two classes. Svelte scopes the edit-mode collapse to:

.layout-container:not(.split).editing.svelte-xxx .viewer-pane:where(.svelte-xxx)

— five. Both carry !important, so specificity decides and width: 0 wins. And opacity: 0 had no print reset at all, so even a width-only fix prints a correctly sized blank rectangle.

#app (the root in app.html) is the only handle a global sheet has that outranks a component's scoped selector.

Side effect worth noting: the old .pane.viewer-pane { flex: none !important } was also losing in reading mode, beaten by the same five-class flex: 1 !important. Harmless there, now correct everywhere.

Cause 2: the preview content is stale in edit mode

The preview-render effect runs on tab.isSplit || (isEditing && settings.showToc), so in plain edit mode with the outline closed, tab.content never updates.

Showing the pane without fixing this yields the document as it was before the edits — present, stale, and looking like it worked. That is worse than a blank page, which at least announces the failure.

syncPreviewForPrint() renders and awaits renderRichContent before printing. Skipped in reading mode — that DOM already came from loadMarkdown on the same buffer, and re-rendering would discard scroll, fold and find state — and skipped when the rendered content already matches. A failure toasts rather than silently exporting a stale document.

#382's print work is intact; exportFoldParity.test.ts still green.

Tests

scripts/editorPdfExport.test.ts resolves the cascade over the real compiled component CSS (svelte/compiler) plus styles.css, modelling #app, :where()/:not() specificity, and the pane's inline flex.

vs #406	7 red / 3 green → 10 / 10

The three that pass by design are premise anchors: edit mode still hides the pane on screen, the editor still stays off the page, and the effect condition is what the fix assumes.

npm run check   0 errors
npm test        452 / 452
cargo test      131 / 131

Not covered

• Whether the printed page is legible — the test proves the pane is laid out and filled, not typeset well.
• .markdown-container carries zoom: 1 in edit mode but zoomLevel/100 in reading mode, so a PDF from the editor is always 100% while one from reading mode inherits the user's zoom. Left alone.

🤖 Generated with Claude Code