fix(core): bound inert region scans in timing compiler

packages/core/src/compiler/timingCompiler.test.ts

@@ -18,6 +18,56 @@ it("source contains no raw NUL bytes", () => {
   expect(src.includes("\x00")).toBe(false);
 });
 
+describe("inert region scanning", () => {
+  const media = '<video id="v" data-start="1" data-duration="2">';
+
+  it.each([
+    `<!-- ${media} <!-- nested -->`,
+    `<!-- --!> ${media} -->`,
+    `<ScRiPt type="text/javascript">${media}</sCrIpT \n>`,
+    `<STYLE>${media}</STYLE\u00a0>`,
+    `<script-data>${media}</script>`,
+    `<style:${media}</style>`,
+    `<!-- <script>${media} -->`,
+    `<script><!-- ${media}</script>`,
+    `<style><script>${media}</style>`,
+  ])("preserves the existing complete-region boundaries in %j", (region) => {
+    const result = compileTimingAttrs(region + media);
+    expect(result).toEqual({ html: region + compileTimingAttrs(media).html, unresolved: [] });
+    expect(extractResolvedMedia(region + media)).toEqual(extractResolvedMedia(media));
+  });
+
+  it.each(["<!--", "<!-- --!>", "<script>", "<style>", "<scripture>", "<stylesheet>"])(
+    "keeps media outside a complete inert region after %j visible",
+    (prefix) => {
+      expect(compileTimingAttrs(prefix + media).html).toBe(prefix + compileTimingAttrs(media).html);
+      expect(extractResolvedMedia(prefix + media)).toEqual(extractResolvedMedia(media));
+    },
+  );
+
+  it.each(["<!--", "<script", "<style"])(
+    "handles many unclosed %j prefixes while masking other region kinds",
+    (prefix) => {
+      const unclosed = prefix.repeat(100_000);
+      const hidden = prefix === "<!--" ? `<style>${media}</style>` : `<!--${media}-->`;
+      expect(compileTimingAttrs(unclosed + hidden + media)).toEqual({
+        html: unclosed + hidden + compileTimingAttrs(media).html,
+        unresolved: [],
+      });
+      expect(extractResolvedMedia(unclosed + hidden + media)).toEqual(extractResolvedMedia(media));
+    },
+  );
+
+  it("uses the first closing delimiter and resumes scanning after it", () => {
+    const hidden = `<script>${media}</script>`;
+    const html = hidden + media + `</script><!--${media}--><style>${media}</style>`;
+    expect(compileTimingAttrs(html).html).toBe(
+      hidden + compileTimingAttrs(media).html + `</script><!--${media}--><style>${media}</style>`,
+    );
+    expect(extractResolvedMedia(html)).toEqual(extractResolvedMedia(media));
+  });
+});
+
 describe("compileTimingAttrs", () => {
   it.each(["", "   ", "0s", "0abc", "0px", "-1s", "Infinity", "NaN"])(
     "does not partially parse invalid literal data-duration=%j",

packages/core/src/compiler/timingCompiler.ts

@@ -113,18 +113,45 @@ function setAttr(tag: string, attr: string, value: string): string {
 // `<video>`/`<audio>` gets rewritten as if it were a real element (issue #1938).
 // Mask those inert regions with placeholders (no `<`, so the tag regexes skip
 // them) before scanning, then restore them verbatim.
-const INERT_REGION_RE =
-  /<!--[\s\S]*?-->|<script\b[\s\S]*?<\/script\s*>|<style\b[\s\S]*?<\/style\s*>/gi;
-
 // The NUL delimiters must stay as \u0000 escapes: raw 0x00 bytes make this file
 // binary to git and are corrupted by Bun's transpiler when bundled (issue #2139).
 function maskInertRegions(html: string): { masked: string; restore: (s: string) => string } {
   const stash: string[] = [];
-  const masked = html.replace(INERT_REGION_RE, (region) => {
+  const parts: string[] = [];
+  const opening = /<!--|<script\b|<style\b/gi;
+  const closings = new Map<string, string | RegExp>([
+    ["<!--", "-->"],
+    ["<script", /<\/script\s*>/gi],
+    ["<style", /<\/style\s*>/gi],
+  ]);
+  let cursor = 0;
+  let match: RegExpExecArray | null;
+  while ((match = opening.exec(html)) !== null) {
+    const kind = match[0].toLowerCase();
+    const closing = closings.get(kind);
+    if (!closing) continue;
+    let end: number;
+    if (typeof closing === "string") {
+      const index = html.indexOf(closing, opening.lastIndex);
+      end = index < 0 ? -1 : index + closing.length;
+    } else {
+      closing.lastIndex = opening.lastIndex;
+      end = closing.exec(html) ? closing.lastIndex : -1;
+    }
+    if (end < 0) {
+      // No later opener of this kind can close either. Search each unmatched
+      // suffix only once, while still allowing other kinds of inert regions.
+      closings.delete(kind);
+      continue;
+    }
     const token = `\u0000HFMASK${stash.length}\u0000`;
-    stash.push(region);
-    return token;
-  });
+    parts.push(html.slice(cursor, match.index), token);
+    stash.push(html.slice(match.index, end));
+    cursor = end;
+    opening.lastIndex = cursor;
+  }
+  parts.push(html.slice(cursor));
+  const masked = parts.join("");
   const restore = (s: string): string =>
     // oxlint-disable-next-line no-control-regex -- NUL cannot appear in HTML, which is what makes it a safe mask delimiter
     s.replace(/\u0000HFMASK(\d+)\u0000/g, (_, i) => stash[Number(i)] ?? "");