Small style fixes2

Author: isTravis

pubstash/server.ts

@@ -120,13 +120,6 @@ async function convertHtmlToPdf(html: string): Promise<Buffer> {
 		// Load the HTML
 		await page.setContent(html, { waitUntil: 'load', timeout: PAGE_TIMEOUT_MS });
 
-		// Chromium 146 renders characters like ↩ (U+21A9) as color emoji by
-		// default.  Force text presentation so they stay as simple glyphs —
-		// important for academic reference back-links.
-		await page.addStyleTag({
-			content: `*, *::before, *::after { font-variant-emoji: text; }`,
-		});
-
 		// Wait for all web fonts before paged.js measures text.
 		// (Key fix from the pubpub/pagedjs-cli fork.)
 		await page.evaluate(() => document.fonts.ready);
@@ -206,6 +199,32 @@ async function convertHtmlToPdf(html: string): Promise<Buffer> {
 			() => new Promise((r) => requestAnimationFrame(() => requestAnimationFrame(r))),
 		);
 
+		// Chromium 146+ renders certain Unicode characters (e.g. ↩ U+21A9 in
+		// footnote back-links) as color emoji instead of simple text glyphs.
+		// Fix at the Unicode level by appending U+FE0E (Variation Selector 15 =
+		// "text presentation") to every affected character in text nodes.
+		// This is more reliable than CSS font-variant-emoji because page.pdf()
+		// uses Chromium's internal print path which may ignore that property.
+		await page.evaluate(() => {
+			const VS15 = '\uFE0E';
+			// Characters with dual text/emoji presentation that commonly appear
+			// in academic documents and should remain as simple text glyphs.
+			const EMOJI_CAPABLE =
+				/[\u2139\u21A9\u21AA\u2194-\u2199\u23CF\u23ED-\u23EF\u23F1\u23F2\u25AA\u25AB\u25B6\u25C0\u25FB-\u25FE\u2600-\u27BF\u2934\u2935\u2B05-\u2B07\u2B1B\u2B1C\u2B50\u2B55\u3030\u303D]/;
+			const walker = document.createTreeWalker(document.body, NodeFilter.SHOW_TEXT);
+			let node: Text | null;
+			while ((node = walker.nextNode() as Text | null)) {
+				const text = node.nodeValue;
+				if (text && EMOJI_CAPABLE.test(text)) {
+					// Append VS15 after each matching char (skip if already followed by VS15)
+					node.nodeValue = text.replace(
+						/([\u2139\u21A9\u21AA\u2194-\u2199\u23CF\u23ED-\u23EF\u23F1\u23F2\u25AA\u25AB\u25B6\u25C0\u25FB-\u25FE\u2600-\u27BF\u2934\u2935\u2B05-\u2B07\u2B1B\u2B1C\u2B50\u2B55\u3030\u303D])(?!\uFE0E)/g,
+						`$1${VS15}`,
+					);
+				}
+			}
+		});
+
 		// Extract <meta> tags for PDF metadata
 		const meta: PdfMeta = await page.evaluate(() => {
 			const m: Record<string, string> = {};