fix(scripts): preserve complete HTML src/href URLs in README link checks (#34817)

Only srcset is a comma-separated candidate list; src and href hold one URL.
Splitting them invented missing targets for paths containing commas or
spaces and could pass a missing target whose fragments were tracked.

Fixes #34815

config/scripts/check-readme-local-links.mjs

@@ -7,7 +7,7 @@ import { pathToFileURL } from 'node:url'
 // Docs-only CI skips some media owners; check index paths because GitHub cannot render uncommitted docs-site/feature-wall assets.
 const TRANSLATED_README_DIR = path.join('docs', 'readme')
 const EXTERNAL_TARGET = /^(?:[a-z][a-z0-9+.-]*:|#|\/\/)/i
-const HTML_ATTRIBUTE = /\b(?:src|srcset|href)\s*=\s*(?:"([^"]*)"|'([^']*)')/g
+const HTML_ATTRIBUTE = /\b(src|srcset|href)\s*=\s*(?:"([^"]*)"|'([^']*)')/g
 const MARKDOWN_LINK = /!?\[[^\]]*\]\(([^)\s]+)(?:\s+"[^"]*")?\)/g
 
 function readmeFiles(root) {
@@ -32,10 +32,14 @@ function trackedFiles(root, candidates) {
 }
 
 function* localTargets(markdown) {
-  for (const match of markdown.matchAll(HTML_ATTRIBUTE)) {
-    // srcset lists candidates; each begins with its URL.
-    for (const candidate of (match[1] ?? match[2]).split(',')) {
-      const target = candidate.trim().split(/\s+/)[0]
+  for (const [, attribute, doubleQuoted, singleQuoted] of markdown.matchAll(HTML_ATTRIBUTE)) {
+    const value = doubleQuoted ?? singleQuoted
+    // Only srcset lists candidates; each begins with its URL. src/href hold one whole URL.
+    const targets =
+      attribute === 'srcset'
+        ? value.split(',').map((candidate) => candidate.trim().split(/\s+/)[0])
+        : [value.trim()]
+    for (const target of targets) {
       if (target) {
         yield target
       }

config/scripts/check-readme-local-links.test.mjs

@@ -44,7 +44,9 @@ const validReadmes = {
     "<img src='resources/build/icon.png' />",
     '<img src="https://img.shields.io/badge/x-y-z" />',
     '[Contributing](.github/CONTRIBUTING.md) [Docs](https://example.com/docs) [Top](#top)',
-    '![hero](docs/assets/hero%20image.jpg "Hero")'
+    '![hero](docs/assets/hero%20image.jpg "Hero")',
+    // src/href hold one URL; only srcset is a comma-separated list.
+    '<img src="docs/assets/hero image,wide.jpg" /> <a href=\'https://example.com/docs?pages=one,two\'>Docs</a>'
   ].join('\n'),
   'docs/readme/README.ja.md': [
     '<img src="../../resources/build/icon.png" />',
@@ -56,6 +58,7 @@ const validReadmes = {
   'resources/onboarding/feature-wall/tile-01.poster.jpg': 'jpg',
   'docs/site/public/docs/tab-split.gif': 'gif',
   'docs/assets/hero image.jpg': 'jpg',
+  'docs/assets/hero image,wide.jpg': 'jpg',
   '.github/CONTRIBUTING.md': 'contributing',
   LICENSE: 'mit'
 }
@@ -77,6 +80,38 @@ describe('README local link check', () => {
     expect(findBrokenReadmeLinks(makeFixture(validReadmes))).toEqual([])
   })
 
+  it('rejects a missing full URL even when its split fragments are tracked', () => {
+    const root = makeFixture({
+      ...validReadmes,
+      'README.md': '<img src="docs/assets/missing image,wide.jpg" />',
+      'docs/assets/missing': 'fragment',
+      'wide.jpg': 'fragment'
+    })
+
+    expect(findBrokenReadmeLinks(root)).toEqual([
+      {
+        readme: 'README.md',
+        target: 'docs/assets/missing image,wide.jpg',
+        resolved: 'docs/assets/missing image,wide.jpg'
+      }
+    ])
+  })
+
+  it('continues checking each srcset candidate without its descriptor', () => {
+    const root = makeFixture({
+      ...validReadmes,
+      'README.md': '<source srcset="resources/build/icon.png 1x, docs/assets/missing.png 2x">'
+    })
+
+    expect(findBrokenReadmeLinks(root)).toEqual([
+      {
+        readme: 'README.md',
+        target: 'docs/assets/missing.png',
+        resolved: 'docs/assets/missing.png'
+      }
+    ])
+  })
+
   it('checks tracked media outside the detector sparse checkout', () => {
     const root = makeFixture(validReadmes)
     const workflow = parse(readFileSync(path.join(projectDir, '.github/workflows/pr.yml'), 'utf8'))