fix: remove leading newline from `<pre>` contents (#14922)

... if it's not followed by another newline, according to the spec

Fixes #14767

---------

Co-authored-by: Simon Holthausen <simon.holthausen@vercel.com>

.changeset/tricky-radios-vanish.md

@@ -0,0 +1,5 @@
+---
+'svelte': patch
+---
+
+fix: remove leading newline from `<pre>` contents

packages/svelte/src/compiler/phases/3-transform/utils.js

@@ -5,6 +5,7 @@
 import {
 	regex_ends_with_whitespaces,
 	regex_not_whitespace,
+	regex_starts_with_newline,
 	regex_starts_with_whitespaces
 } from '../patterns.js';
 import * as b from '../../utils/builders.js';
@@ -270,6 +271,22 @@ export function clean_nodes(
 
 	var first = trimmed[0];
 
+	// initial newline inside a `<pre>` is disregarded, if not followed by another newline
+	if (parent.type === 'RegularElement' && parent.name === 'pre' && first.type === 'Text') {
+		const text = first.data.replace(regex_starts_with_newline, '');
+		if (text !== first.data) {
+			const tmp = text.replace(regex_starts_with_newline, '');
+			if (text === tmp) {
+				first.data = text;
+				first.raw = first.raw.replace(regex_starts_with_newline, '');
+				if (first.data === '') {
+					trimmed.shift();
+					first = trimmed[0];
+				}
+			}
+		}
+	}
+
 	// Special case: Add a comment if this is a lone script tag. This ensures that our run_scripts logic in template.js
 	// will always be able to call node.replaceWith() on the script tag in order to make it run. If we don't add this
 	// and would still call node.replaceWith() on the script tag, it would be a no-op because the script tag has no parent.

packages/svelte/tests/hydration/samples/pre-first-node-newline/_config.js

@@ -0,0 +1,5 @@
+import { test } from '../../test';
+
+// A note about _expected.html: It is different from body.html because we're
+// testing against target.innerHTML which already removed the redundant first newline
+export default test({});

packages/svelte/tests/hydration/samples/pre-first-node-newline/_expected.html

@@ -0,0 +1,7 @@
+<!--[--><pre>static content no line</pre> <pre>	static content ignored line
+</pre> <pre>
+	static content relevant line
+</pre> <pre><div><span></span></div>
+</pre> <pre>
+<div><span></span></div>
+</pre><!--]-->

packages/svelte/tests/hydration/samples/pre-first-node-newline/main.svelte

@@ -0,0 +1,23 @@
+<script>
+	let name = $state('');
+</script>
+
+<pre>static content no line</pre>
+
+<pre>
+	static content ignored line
+</pre>
+
+<pre>
+
+	static content relevant line
+</pre>
+
+<pre>
+<div><span>{name}</span></div>
+</pre>
+
+<pre>
+
+<div><span>{name}</span></div>
+</pre>

packages/svelte/tests/runtime-legacy/samples/pre-tag/_config.js

@@ -2,17 +2,9 @@ import { test } from '../../test';
 
 export default test({
 	mode: ['client', 'server'], // output is correct, but test suite chokes on the extra ssr comment which is harmless
-	withoutNormalizeHtml: true,
+	withoutNormalizeHtml: 'only-strip-comments', // because whitespace inside pre tags is significant
-	html: get_html(false),
+	// Note how we're testing against target.innerHTML which already removed the redundant first newline
-	ssrHtml: get_html(true)
+	html: `<pre id="pre">  A
-});
-
-/** @param {boolean} ssr */
-function get_html(ssr) {
-	// ssr rendered HTML has an extra newline prefixed within `<pre>` tag,
-	// if the <pre> tag starts with `\n`
-	// because when browser parses the SSR rendered HTML, it will ignore the 1st '\n' character
-	return `${ssr ? '<!--[-->' : ''}<pre id="pre">  A
   B
   <span>
     C
@@ -35,5 +27,5 @@ function get_html(ssr) {
 leading newlines</pre></div> <div id="pre-without-leading-newline"><pre>without spaces</pre> <pre>  with spaces  </pre> <pre>${' '}
 newline after leading space</pre></div> <pre id="pre-with-multiple-leading-newlines">
 
-multiple leading newlines</pre>${ssr ? '<!--]-->' : ''}`;
+multiple leading newlines</pre>`
-}
+});