dompurify 3.4.12 → 3.4.14

package/src/purify.ts

@@ -65,6 +65,48 @@ const NODE_TYPE = {
   notation: 12, // Deprecated
 };
 
+/* HTML-namespace elements whose child text nodes are serialized *literally*
+   (unescaped) by the HTML fragment-serialization algorithm. Two reparse-mXSS
+   shapes ride on that literal serialization:
+     (a) an element child - a tree the HTML parser can never build, but the DOM
+         API and an XML/XHTML parse can - after which a `</tag>`-bearing text
+         sibling breaks the element open on reparse; and
+     (b) text-only content that already carries the element's OWN end tag, e.g.
+         `<style>...</style><img onerror=x>` built as a node, which the literal
+         serializer emits verbatim for the HTML parser to re-open.
+   Shape (a) is handled by the firstElementChild branch in _isUnsafeNode; shape
+   (b) by the LITERAL_TEXT_CLOSE probe. Both read textContent (the raw-serialized
+   form for these elements) rather than innerHTML, because an XML/XHTML working
+   document serializes innerHTML with `<` escaped, which silently blinds the
+   innerHTML-based probes (rule 1's second probe and FALLBACK_TAG_CLOSE) there.
+   `script` is never allow-listed, but is kept here so the guard matches the
+   serializer's own literal-text list exactly. */
+const LITERAL_TEXT_ELEMENT_NAMES = [
+  'style',
+  'script',
+  'xmp',
+  'iframe',
+  'noembed',
+  'noframes',
+  'plaintext',
+  'noscript',
+];
+const LITERAL_TEXT_ELEMENTS = freeze(addToSet({}, LITERAL_TEXT_ELEMENT_NAMES));
+
+/* Per-element end-tag matcher. On an HTML reparse the ONLY token that
+   terminates a literal-text element's raw content is its own end tag; a foreign
+   literal-text close (e.g. `</xmp>` sitting inside `<style>`) does not break
+   out, so matching is per-element, not a shared alternation. The lookahead
+   requires an HTML tag-name terminator (whitespace, `/` or `>`) so a longer
+   name such as `</styles` is not mistaken for `</style`. */
+const LITERAL_TEXT_CLOSE = (function (): Record<string, RegExp> {
+  const map: Record<string, RegExp> = {};
+  arrayForEach(LITERAL_TEXT_ELEMENT_NAMES, (name) => {
+    map[name] = seal(new RegExp('</' + name + '(?=[\\t\\n\\f\\r />])', 'i'));
+  });
+  return freeze(map);
+})();
+
 const getGlobal = function (): WindowLike {
   return typeof window === 'undefined' ? null : window;
 };
@@ -163,6 +205,28 @@ const _resolveSetOption = function (
     : fallback;
 };
 
+/**
+ * Resolve an object-valued configuration option: a prototype-free clone
+ * of cfg[key] when it is an own, truthy object property, else a fresh
+ * fallback built by makeFallback (fresh on every parse, so a previous
+ * parse can never leak state into the next one).
+ *
+ * @param cfg the cloned, prototype-free configuration object
+ * @param key the configuration property to read
+ * @param makeFallback builds the fallback value when the option is absent
+ * @returns the resolved object
+ */
+const _resolveObjectOption = function <T extends Record<string, any>>(
+  cfg: Config,
+  key: keyof Config,
+  makeFallback: () => T
+): T {
+  const value = objectHasOwnProperty(cfg, key) ? cfg[key] : undefined;
+  return value && typeof value === 'object'
+    ? clone(value as T)
+    : makeFallback();
+};
+
 function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
   const DOMPurify: DOMPurify = (root: WindowLike) => createDOMPurify(root);
 
@@ -213,6 +277,22 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     Node && Node.prototype ? lookupGetter(Node.prototype, 'nodeType') : null;
   const getNodeName =
     Node && Node.prototype ? lookupGetter(Node.prototype, 'nodeName') : null;
+  const getOwnerDocument =
+    Node && Node.prototype
+      ? lookupGetter(Node.prototype, 'ownerDocument')
+      : null;
+
+  /* Clobber-safe nodeType / nodeName reads through the cached Node.prototype
+     getters, with a direct-property fallback for environments that lack
+     Node.prototype. Sites that need a different fallback (e.g. _isClobbered
+     returns early on a null name) intentionally keep their own reads. */
+  const _readNodeType = function (node: Node): number {
+    return getNodeType ? getNodeType(node) : (node as any).nodeType;
+  };
+
+  const _readNodeName = function (node: Node): string {
+    return getNodeName ? getNodeName(node) : (node as any).nodeName;
+  };
 
   // As per issue #47, the web-components registry is inherited by a
   // new document created via createHTMLDocument. As per the spec
@@ -724,26 +804,23 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     NAMESPACE =
       typeof cfg.NAMESPACE === 'string' ? cfg.NAMESPACE : HTML_NAMESPACE; // Default HTML namespace
 
-    MATHML_TEXT_INTEGRATION_POINTS =
-      objectHasOwnProperty(cfg, 'MATHML_TEXT_INTEGRATION_POINTS') &&
-      cfg.MATHML_TEXT_INTEGRATION_POINTS &&
-      typeof cfg.MATHML_TEXT_INTEGRATION_POINTS === 'object'
-        ? clone(cfg.MATHML_TEXT_INTEGRATION_POINTS)
-        : addToSet({}, DEFAULT_MATHML_TEXT_INTEGRATION_POINTS); // Default built-in map
-
-    HTML_INTEGRATION_POINTS =
-      objectHasOwnProperty(cfg, 'HTML_INTEGRATION_POINTS') &&
-      cfg.HTML_INTEGRATION_POINTS &&
-      typeof cfg.HTML_INTEGRATION_POINTS === 'object'
-        ? clone(cfg.HTML_INTEGRATION_POINTS)
-        : addToSet({}, DEFAULT_HTML_INTEGRATION_POINTS); // Default built-in map
-
-    const customElementHandling =
-      objectHasOwnProperty(cfg, 'CUSTOM_ELEMENT_HANDLING') &&
-      cfg.CUSTOM_ELEMENT_HANDLING &&
-      typeof cfg.CUSTOM_ELEMENT_HANDLING === 'object'
-        ? clone(cfg.CUSTOM_ELEMENT_HANDLING)
-        : create(null);
+    MATHML_TEXT_INTEGRATION_POINTS = _resolveObjectOption(
+      cfg,
+      'MATHML_TEXT_INTEGRATION_POINTS',
+      () => addToSet({}, DEFAULT_MATHML_TEXT_INTEGRATION_POINTS) // Default built-in map
+    );
+
+    HTML_INTEGRATION_POINTS = _resolveObjectOption(
+      cfg,
+      'HTML_INTEGRATION_POINTS',
+      () => addToSet({}, DEFAULT_HTML_INTEGRATION_POINTS) // Default built-in map
+    );
+
+    const customElementHandling = _resolveObjectOption(
+      cfg,
+      'CUSTOM_ELEMENT_HANDLING',
+      () => create(null)
+    );
 
     CUSTOM_ELEMENT_HANDLING = create(null);
 
@@ -841,24 +918,6 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       }
     }
 
-    if (
-      objectHasOwnProperty(cfg, 'ADD_URI_SAFE_ATTR') &&
-      arrayIsArray(cfg.ADD_URI_SAFE_ATTR)
-    ) {
-      addToSet(URI_SAFE_ATTRIBUTES, cfg.ADD_URI_SAFE_ATTR, transformCaseFunc);
-    }
-
-    if (
-      objectHasOwnProperty(cfg, 'FORBID_CONTENTS') &&
-      arrayIsArray(cfg.FORBID_CONTENTS)
-    ) {
-      if (FORBID_CONTENTS === DEFAULT_FORBID_CONTENTS) {
-        FORBID_CONTENTS = clone(FORBID_CONTENTS);
-      }
-
-      addToSet(FORBID_CONTENTS, cfg.FORBID_CONTENTS, transformCaseFunc);
-    }
-
     if (
       objectHasOwnProperty(cfg, 'ADD_FORBID_CONTENTS') &&
       arrayIsArray(cfg.ADD_FORBID_CONTENTS)
@@ -1171,6 +1230,37 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     }
   };
 
+  /**
+   * _stripAttributeNode
+   *
+   * Remove a single Attr node case/namespace-exactly on an attribute-teardown
+   * path. Name-based removeAttribute() ASCII-lowercases its lookup key for an
+   * HTML element in an HTML document and so silently misses a case-preserved
+   * handler (e.g. `ONERROR` off an XML/XHTML import) - the same defect
+   * _removeAttribute() was fixed for, which a name-based call would reintroduce
+   * on these IN_PLACE teardown paths. Unlike _removeAttribute this does not
+   * record into DOMPurify.removed: the neutralize passes intentionally do not
+   * book-keep. A clobbered/detached node falls back to best-effort name-based
+   * removal.
+   *
+   * @param element the element to strip the attribute from
+   * @param attribute the Attr node to remove
+   * @param name the attribute's name, for the fallback path
+   */
+  const _stripAttributeNode = function (
+    element: Element,
+    attribute: Attr,
+    name: string
+  ): void {
+    try {
+      element.removeAttributeNode(attribute);
+    } catch (_) {
+      try {
+        element.removeAttribute(name);
+      } catch (_) {}
+    }
+  };
+
   /**
    * _neutralizeRoot
    *
@@ -1217,11 +1307,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
         const attribute = attributes[i];
         const name = attribute && attribute.name;
         if (typeof name === 'string') {
-          try {
-            (root as Element).removeAttribute(name);
-          } catch (_) {
-            /* Clobbered removeAttribute — ignore (fail-closed best effort) */
-          }
+          _stripAttributeNode(root as Element, attribute, name);
         }
       }
     }
@@ -1230,24 +1316,53 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
   /**
    * _removeAttribute
    *
+   * Name-based getAttributeNode()/removeAttribute() ASCII-lowercase their
+   * lookup key for HTML elements in an HTML document, so they silently miss an
+   * attribute whose stored qualified name still contains uppercase ASCII
+   * letters. That happens when the node came from a case-preserving source
+   * (an XML/XHTML document imported via importNode(), or createAttributeNS()),
+   * where e.g. `ONERROR` survives the walk: the policy check lowercases to
+   * `onerror` and rejects it, but `removeAttribute('ONERROR')` looks up
+   * `onerror` and finds nothing. Remove the exact Attr node instead, which is
+   * case- and namespace-exact, and fall back to name-based removal only when
+   * the caller could not supply the node.
+   *
    * @param name an Attribute name
    * @param element a DOM node
+   * @param attr the exact Attr node to remove, when the caller has it
    */
-  const _removeAttribute = function (name: string, element: Element): void {
+  const _removeAttribute = function (
+    name: string,
+    element: Element,
+    attr?: Attr | null
+  ): void {
+    if (!attr) {
+      try {
+        attr = element.getAttributeNode(name);
+      } catch (_) {
+        attr = null;
+      }
+    }
+
+    arrayPush(DOMPurify.removed, {
+      attribute: attr || null,
+      from: element,
+    });
+
     try {
-      arrayPush(DOMPurify.removed, {
-        attribute: element.getAttributeNode(name),
-        from: element,
-      });
+      if (attr) {
+        element.removeAttributeNode(attr);
+      } else {
+        element.removeAttribute(name);
+      }
     } catch (_) {
-      arrayPush(DOMPurify.removed, {
-        attribute: null,
-        from: element,
-      });
+      /* Clobbered or already-detached node - best-effort fall back to a
+         name-based removal so the "is" handling below still runs. */
+      try {
+        element.removeAttribute(name);
+      } catch (_) {}
     }
 
-    element.removeAttribute(name);
-
     // We void attribute values for unremovable "is" attributes
     if (name === 'is') {
       if (RETURN_DOM || RETURN_DOM_FRAGMENT) {
@@ -1285,11 +1400,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
         continue;
       }
 
-      try {
-        element.removeAttribute(name);
-      } catch (_) {
-        /* Clobbered removeAttribute on a doomed node — ignore */
-      }
+      _stripAttributeNode(element, attribute, name);
     }
   };
 
@@ -1320,7 +1431,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
 
     while (stack.length > 0) {
       const node = stack.pop();
-      const nodeType = getNodeType ? getNodeType(node) : (node as any).nodeType;
+      const nodeType = _readNodeType(node);
 
       if (nodeType === NODE_TYPE.element) {
         _stripDisallowedAttributes(node as Element);
@@ -1370,6 +1481,32 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
    *
    * @param root the in-place root to sweep
    */
+  /**
+   * Central policy for declarative-partial-updates patch-linkage attributes,
+   * shared by the _neutralizePatchLinkage pre-pass and _isValidAttribute so
+   * the two sites cannot drift: `patchsrc` always links, `for` links
+   * everywhere except on <label>/<output>, and the whole policy is gated on
+   * SAFE_FOR_XML (see the rationale block in _isValidAttribute).
+   *
+   * @param lcName the transformCaseFunc'd attribute name
+   * @param lcTag the transformCaseFunc'd tag name of the carrying element
+   * @return true if the attribute is patch linkage and must be dropped
+   */
+  const _isPatchLinkageAttribute = function (
+    lcName: string,
+    lcTag: string
+  ): boolean {
+    if (!SAFE_FOR_XML) {
+      return false;
+    }
+
+    if (lcName === 'patchsrc') {
+      return true;
+    }
+
+    return lcName === 'for' && lcTag !== 'label' && lcTag !== 'output';
+  };
+
   const _neutralizePatchLinkage = function (root: Node): void {
     if (!SAFE_FOR_XML) {
       return;
@@ -1378,7 +1515,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     const stack: Node[] = [root];
     while (stack.length > 0) {
       const node = stack.pop();
-      const nodeType = getNodeType ? getNodeType(node) : (node as any).nodeType;
+      const nodeType = _readNodeType(node);
 
       /* Remove range markers (the target side of a patch linkage): every
          processing instruction, and any markup-bearing comment. */
@@ -1399,9 +1536,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       /* Strip patch-source attributes (the source side) off elements. */
       if (nodeType === NODE_TYPE.element) {
         const element = node as Element;
-        const lcTag = transformCaseFunc(
-          getNodeName ? getNodeName(node) : (node as any).nodeName
-        );
+        const lcTag = transformCaseFunc(_readNodeName(node));
         try {
           if (element.hasAttribute && element.hasAttribute('patchsrc')) {
             element.removeAttribute('patchsrc');
@@ -1410,8 +1545,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
           if (
             element.hasAttribute &&
             element.hasAttribute('for') &&
-            lcTag !== 'label' &&
-            lcTag !== 'output'
+            _isPatchLinkageAttribute('for', lcTag)
           ) {
             element.removeAttribute('for');
           }
@@ -1509,8 +1643,18 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
    * @return The created NodeIterator
    */
   const _createNodeIterator = function (root: Node): NodeIterator {
+    /* Read ownerDocument through the cached Node.prototype getter, never the
+       direct property. HTMLFormElement has [LegacyOverrideBuiltIns], so a
+       clobbering child (<input name="ownerDocument"> or a form-associated
+       external input) shadows the prototype getter and makes a direct read
+       return that <input>. createNodeIterator.call(<input>, ...) then throws
+       "Illegal invocation", and on the IN_PLACE path that throw lands before
+       the walk's fail-closed barrier - leaving the caller's live tree, with
+       any already-armed handler in it, un-neutralized. The cached getter
+       returns the real Document regardless of the clobber. */
+    const doc = getOwnerDocument ? getOwnerDocument(root) : root.ownerDocument;
     return createNodeIterator.call(
-      root.ownerDocument || root,
+      doc || root,
       root,
       // eslint-disable-next-line no-bitwise
       NodeFilter.SHOW_ELEMENT |
@@ -1558,8 +1702,12 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
    */
   const _scrubTemplateExpressions = function (node: Element): void {
     node.normalize();
+    /* Clobber-safe ownerDocument read, same reasoning as _createNodeIterator:
+       under SAFE_FOR_TEMPLATES this runs on the live IN_PLACE root, which may
+       carry a form-named-getter override of ownerDocument. */
+    const doc = getOwnerDocument ? getOwnerDocument(node) : node.ownerDocument;
     const walker = createNodeIterator.call(
-      node.ownerDocument || node,
+      doc || node,
       node,
       // eslint-disable-next-line no-bitwise
       NodeFilter.SHOW_TEXT |
@@ -1733,12 +1881,22 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       return true;
     }
 
-    /* Remove risky CSS construction leading to mXSS */
+    /* Remove rawtext/literal-text elements whose literal serialization
+       re-opens markup on an HTML reparse - shapes (a) and (b) documented at
+       LITERAL_TEXT_ELEMENTS. Both are invisible to rule 1 above (which
+       self-disables once there is an element child, and whose second probe
+       reads the innerHTML an XML/XHTML document serializes escaped), which
+       is why both probes here read textContent instead. Previously only
+       `style`-with-element-child was covered; every element in
+       LITERAL_TEXT_ELEMENTS shares this literal serialization and is
+       equally affected. */
     if (
       SAFE_FOR_XML &&
       currentNode.namespaceURI === HTML_NAMESPACE &&
-      tagName === 'style' &&
-      _isNode(currentNode.firstElementChild)
+      LITERAL_TEXT_ELEMENTS[tagName] &&
+      (_isNode(currentNode.firstElementChild) ||
+        (typeof currentNode.textContent === 'string' &&
+          regExpTest(LITERAL_TEXT_CLOSE[tagName], currentNode.textContent)))
     ) {
       return true;
     }
@@ -1760,6 +1918,34 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     return false;
   };
 
+  /**
+   * Evaluate a CUSTOM_ELEMENT_HANDLING check (a RegExp or a predicate
+   * function, per the validation in _parseConfig) against a name.
+   * Additional arguments are forwarded to predicate functions - the
+   * attributeNameCheck predicate receives the tag name as its second
+   * argument. A null/absent check never matches.
+   *
+   * @param check the configured tagNameCheck / attributeNameCheck value
+   * @param name the name to test
+   * @param args extra arguments forwarded to a predicate function
+   * @return true if the check matches the name
+   */
+  const _matchesNameCheck = function (
+    check: unknown,
+    name: string,
+    ...args: unknown[]
+  ): boolean {
+    if (check instanceof RegExp) {
+      return regExpTest(check, name);
+    }
+
+    if (check instanceof Function) {
+      return Boolean(check(name, ...args));
+    }
+
+    return false;
+  };
+
   /**
    * Handle a node whose tag is forbidden or not allowlisted: keep
    * allowed custom elements (false return exits _sanitizeElements
@@ -1779,23 +1965,16 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
    */
   const _sanitizeDisallowedNode = function (
     currentNode: any,
-    tagName: string
+    tagName: string,
+    root: Node
   ): boolean {
     /* Check if we have a custom element to handle */
-    if (!FORBID_TAGS[tagName] && _isBasicCustomElement(tagName)) {
-      if (
-        CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof RegExp &&
-        regExpTest(CUSTOM_ELEMENT_HANDLING.tagNameCheck, tagName)
-      ) {
-        return false;
-      }
-
-      if (
-        CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof Function &&
-        CUSTOM_ELEMENT_HANDLING.tagNameCheck(tagName)
-      ) {
-        return false;
-      }
+    if (
+      !FORBID_TAGS[tagName] &&
+      _isBasicCustomElement(tagName) &&
+      _matchesNameCheck(CUSTOM_ELEMENT_HANDLING.tagNameCheck, tagName)
+    ) {
+      return false;
     }
 
     /* Keep content except for bad-listed elements.
@@ -1813,31 +1992,31 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       if (childNodes && parentNode) {
         const childCount = childNodes.length;
 
-        /* In-place: hoist the *original* children so the iterator visits
-             and sanitises them through the same allowlist pass as every other
-             node. The caller built the tree in the live document, so the
-             originals carry already-queued resource events (`<img onerror>`,
-             `<video>`/`<audio>` error, lazy/`onload`, …); cloning would leave
-             those originals detached but still armed, firing in page scope
-             while the returned tree looked clean. Moving is safe in-place: the
-             root is pre-validated as an allowed tag and so is never the node
-             being removed, which keeps `parentNode` inside the iterator root
-             and the relocated child inside the serialised tree.
-
-             Otherwise (string / DOM-copy paths): clone. The iterator is rooted
-             at — and the result serialised from — `body`, so a restrictive
-             ALLOWED_TAGS that removes `body` itself must leave its content in
-             place, which only cloning does; and those paths parse into an
-             inert document, so their discarded originals never had a queued
-             event to neutralise.
+        /* Hoist by moving each child up one level rather than deep-cloning
+             it. Moving transfers every descendant exactly once, so a chain of
+             nested disallowed elements costs O(n) instead of the O(n^2) that
+             re-cloning the shrinking subtree at each level produced; it also
+             empties the removed original, so `DOMPurify.removed` no longer
+             pins whole subtrees. Moving preserves the in-place guarantee too:
+             an original carrying already-queued resource events (`<img
+             onerror>`, `<video>`/`<audio>` error, lazy/`onload`, …) is
+             relocated and sanitised rather than left detached but still armed.
+
+             The sole case that must clone is removing the walk root itself.
+             The result is serialised from the root's subtree, so a restrictive
+             ALLOWED_TAGS that strips the root (`body` on the string path) must
+             leave the content inside it, which only cloning does. In IN_PLACE
+             the root is pre-validated as an allowed tag and so is never removed
+             here, so that path always takes the move branch.
 
              `childNodes` is live; a tail-to-head walk keeps `childNodes[i]`
              valid whether we move (drops the trailing entry) or clone (leaves
              the list intact). */
         for (let i = childCount - 1; i >= 0; --i) {
-          const hoisted = IN_PLACE
-            ? childNodes[i]
-            : cloneNode(childNodes[i], true);
+          const hoisted =
+            currentNode === root
+              ? cloneNode(childNodes[i], true)
+              : childNodes[i];
           parentNode.insertBefore(hoisted, getNextSibling(currentNode));
         }
       }
@@ -1847,6 +2026,79 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     return true;
   };
 
+  /**
+   * Fork a hook-mutable allowlist off its shared binding the first time a
+   * (possibly lazily-installed) uponSanitize* hook is about to see it, so the
+   * hook cannot widen the per-instance default or the setConfig binding by
+   * reference and leak past the call. Returns the set unchanged once it is
+   * already call-local, so repeated calls across elements are idempotent.
+   *
+   * @param hookList the uponSanitize* hook array for this event
+   * @param set the current ALLOWED_TAGS / ALLOWED_ATTR binding
+   * @param defaultSet the per-instance DEFAULT_ALLOWED_* constant
+   * @param setConfigSet the captured setConfig() binding, or null
+   * @return a call-local clone if a hook is present and set is still shared,
+   *   else set unchanged
+   */
+  const _forkSharedAllowlist = function <T extends Record<string, any>>(
+    hookList: unknown[],
+    set: T,
+    defaultSet: T,
+    setConfigSet: T | null
+  ): T {
+    if (hookList.length === 0) {
+      return set;
+    }
+
+    return set === defaultSet || set === setConfigSet ? clone(set) : set;
+  };
+
+  /**
+   * Shared guard for a node that a hook has detached from the walk tree,
+   * used after each element-hook site in _sanitizeElements. Detaching is a
+   * long-standing user pattern (issue #469; draw.io-style foreignObject
+   * filtering). Per the cached, unclobberable parentNode getter the node is
+   * genuinely out of the tree, so it can reach neither the serialized
+   * output nor an IN_PLACE live tree; treat it as removed and stop
+   * processing it. Without this guard, the unsafe-node / namespace checks
+   * would call _forceRemove on a parentless node and hit the REPORT-3
+   * fail-closed throw — which exists for nodes DOMPurify wants gone but
+   * *cannot* detach (clobbered / parentless roots), the opposite of a node
+   * that is already safely gone. The walk root is exempt: a detached
+   * IN_PLACE root is legitimate input and must still be fully sanitized,
+   * and a kill-decision on it must keep hitting the REPORT-3 throw.
+   *
+   * Nodes detached by hooks stay the hook's responsibility for placement:
+   * they are not recorded in DOMPurify.removed, so the post-walk IN_PLACE
+   * pass (which iterates DOMPurify.removed) does not reach them. But a
+   * hook-detached subtree can still hold a queued resource-event handler -
+   * e.g. an <img onload> that began loading when the caller built the live
+   * tree - which fires in page scope after sanitize returns even though the
+   * handler never reached the returned tree. That is the audit-5 F1 hazard,
+   * and the documented node.remove() hook pattern walks straight into it.
+   * So on the IN_PLACE path we neutralize the detached subtree inline,
+   * stripping its non-allow-listed attributes before returning, exactly as
+   * the post-walk pass does for _forceRemove'd subtrees.
+   *
+   * @param currentNode the node a hook may have detached
+   * @param root the current walk root
+   * @return true if the node is detached and now handled, false otherwise
+   */
+  const _handleHookDetachedNode = function (
+    currentNode: Node,
+    root: Node
+  ): boolean {
+    if (currentNode === root || getParentNode(currentNode) !== null) {
+      return false;
+    }
+
+    if (IN_PLACE) {
+      _neutralizeSubtree(currentNode);
+    }
+
+    return true;
+  };
+
   /**
    * _sanitizeElements
    *
@@ -1856,14 +2108,13 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
    * @param currentNode to check for permission to exist
    * @return true if node was killed, false if left alive
    */
-  // eslint-disable-next-line complexity
   const _sanitizeElements = function (currentNode: any, root: Node): boolean {
     /* Execute a hook if present */
     _executeHooks(hooks.beforeSanitizeElements, currentNode, null);
 
-    /* A hook may have detached the node — treat it as removed (see the
-       detached-node comment after the uponSanitizeElement hook below). */
-    if (currentNode !== root && getParentNode(currentNode) === null) {
+    /* A hook may have detached the node - treat it as removed (see
+       _handleHookDetachedNode for the full rationale). */
+    if (_handleHookDetachedNode(currentNode, root)) {
       return true;
     }
 
@@ -1874,8 +2125,18 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
     }
 
     /* Now let's check the element's type and name */
-    const tagName = transformCaseFunc(
-      getNodeName ? getNodeName(currentNode) : currentNode.nodeName
+    const tagName = transformCaseFunc(_readNodeName(currentNode));
+
+    /* Close the pre-walk clone-guard's timing gap: an uponSanitizeElement
+       hook may have been installed after that guard sampled the hook arrays
+       (e.g. lazily from beforeSanitizeElements), leaving ALLOWED_TAGS still
+       aliasing a shared binding that a widening hook would mutate by
+       reference. Fork it before exposing it to the hook. */
+    ALLOWED_TAGS = _forkSharedAllowlist(
+      hooks.uponSanitizeElement,
+      ALLOWED_TAGS,
+      DEFAULT_ALLOWED_TAGS,
+      SET_CONFIG_ALLOWED_TAGS
     );
 
     /* Execute a hook if present */
@@ -1884,22 +2145,9 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       allowedTags: ALLOWED_TAGS,
     });
 
-    /* A hook may have detached the node from the tree — a long-standing
-       user pattern (issue #469; draw.io-style foreignObject filtering).
-       Per the cached, unclobberable parentNode getter the node is
-       genuinely out of the tree, so it can reach neither the serialized
-       output nor an IN_PLACE live tree; treat it as removed and stop
-       processing it. Without this guard, the unsafe-node / namespace
-       checks below would call _forceRemove on a parentless node and hit
-       the REPORT-3 fail-closed throw — which exists for nodes DOMPurify
-       wants gone but *cannot* detach (clobbered / parentless roots), the
-       opposite of a node that is already safely gone. The walk root is
-       exempt: a detached IN_PLACE root is legitimate input and must still
-       be fully sanitized, and a kill-decision on it must keep hitting the
-       REPORT-3 throw. Nodes detached by hooks are the hook's
-       responsibility: they are not recorded in DOMPurify.removed and are
-       not neutralized by the post-walk IN_PLACE pass. */
-    if (currentNode !== root && getParentNode(currentNode) === null) {
+    /* The uponSanitizeElement hook may have detached the node, exactly as
+       above (see _handleHookDetachedNode for the full rationale). */
+    if (_handleHookDetachedNode(currentNode, root)) {
       return true;
     }
 
@@ -1918,7 +2166,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       ) &&
         !ALLOWED_TAGS[tagName])
     ) {
-      const removed = _sanitizeDisallowedNode(currentNode, tagName);
+      const removed = _sanitizeDisallowedNode(currentNode, tagName, root);
 
       /* A false return means the node is a custom element kept via
          CUSTOM_ELEMENT_HANDLING - the only keep path through
@@ -1944,7 +2192,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
        bound and short-circuits to false for any node minted in a different
        realm — letting a foreign-realm element with a forbidden namespace
        slip past the namespace check entirely. */
-    const nt = getNodeType ? getNodeType(currentNode) : currentNode.nodeType;
+    const nt = _readNodeType(currentNode);
     if (nt === NODE_TYPE.element && !_checkValidNamespace(currentNode)) {
       _forceRemove(currentNode);
       return true;
@@ -2020,16 +2268,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
        other structural-threat checks and stays overridable, consistent with
        the rest of the codebase. PI range markers are already removed by
        _isUnsafeNode. */
-    if (SAFE_FOR_XML && lcName === 'patchsrc') {
-      return false;
-    }
-
-    if (
-      SAFE_FOR_XML &&
-      lcName === 'for' &&
-      lcTag !== 'label' &&
-      lcTag !== 'output'
-    ) {
+    if (_isPatchLinkageAttribute(lcName, lcTag)) {
       return false;
     }
 
@@ -2052,73 +2291,76 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
         XML-compatible (https://html.spec.whatwg.org/multipage/infrastructure.html#xml-compatible and http://www.w3.org/TR/xml/#d0e804)
         We don't need to check the value; it's always URI safe. */
     if (ALLOW_DATA_ATTR && regExpTest(DATA_ATTR, lcName)) {
-      // This attribute is safe
-    } else if (ALLOW_ARIA_ATTR && regExpTest(ARIA_ATTR, lcName)) {
-      // This attribute is safe
-      /* Otherwise, check the name is permitted */
-    } else if (!nameIsPermitted) {
-      if (
-        // First condition does a very basic check if a) it's basically a valid custom element tagname AND
-        // b) if the tagName passes whatever the user has configured for CUSTOM_ELEMENT_HANDLING.tagNameCheck
-        // and c) if the attribute name passes whatever the user has configured for CUSTOM_ELEMENT_HANDLING.attributeNameCheck
+      return true;
+    }
+
+    /* Allow valid aria-* attributes, the value is always URI safe */
+    if (ALLOW_ARIA_ATTR && regExpTest(ARIA_ATTR, lcName)) {
+      return true;
+    }
+
+    /* A name outside the allowlist is acceptable on custom-element terms
+       only. The value checks below are intentionally skipped in that case:
+       if the user supplied a tagNameCheck we also allow derived custom
+       elements using the same test, and attributes passing the configured
+       attributeNameCheck are allowed as custom elements define these at
+       their own discretion. */
+    if (!nameIsPermitted) {
+      return (
+        // Condition a) covers a basically valid custom element tag name whose
+        // tag passes the configured tagNameCheck and whose attribute name
+        // passes the configured attributeNameCheck ...
         (_isBasicCustomElement(lcTag) &&
-          ((CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof RegExp &&
-            regExpTest(CUSTOM_ELEMENT_HANDLING.tagNameCheck, lcTag)) ||
-            (CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof Function &&
-              CUSTOM_ELEMENT_HANDLING.tagNameCheck(lcTag))) &&
-          ((CUSTOM_ELEMENT_HANDLING.attributeNameCheck instanceof RegExp &&
-            regExpTest(CUSTOM_ELEMENT_HANDLING.attributeNameCheck, lcName)) ||
-            (CUSTOM_ELEMENT_HANDLING.attributeNameCheck instanceof Function &&
-              CUSTOM_ELEMENT_HANDLING.attributeNameCheck(lcName, lcTag)))) ||
-        // Alternative, second condition checks if it's an `is`-attribute, AND
-        // the value passes whatever the user has configured for CUSTOM_ELEMENT_HANDLING.tagNameCheck
+          _matchesNameCheck(CUSTOM_ELEMENT_HANDLING.tagNameCheck, lcTag) &&
+          _matchesNameCheck(
+            CUSTOM_ELEMENT_HANDLING.attributeNameCheck,
+            lcName,
+            lcTag
+          )) ||
+        // Condition b) covers an `is` attribute whose value passes the
+        // configured tagNameCheck while customized built-in elements are
+        // allowed.
         (lcName === 'is' &&
           CUSTOM_ELEMENT_HANDLING.allowCustomizedBuiltInElements &&
-          ((CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof RegExp &&
-            regExpTest(CUSTOM_ELEMENT_HANDLING.tagNameCheck, value)) ||
-            (CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof Function &&
-              CUSTOM_ELEMENT_HANDLING.tagNameCheck(value))))
-      ) {
-        // If user has supplied a regexp or function in CUSTOM_ELEMENT_HANDLING.tagNameCheck, we need to also allow derived custom elements using the same tagName test.
-        // Additionally, we need to allow attributes passing the CUSTOM_ELEMENT_HANDLING.attributeNameCheck user has configured, as custom elements can define these at their own discretion.
-      } else {
-        return false;
-      }
-      /* Check value is safe. First, is attr inert? If so, is safe */
-    } else if (URI_SAFE_ATTRIBUTES[lcName]) {
-      // This attribute is safe
-      /* Check no script, data or unknown possibly unsafe URI
+          _matchesNameCheck(CUSTOM_ELEMENT_HANDLING.tagNameCheck, value))
+      );
+    }
+
+    /* Check value is safe. First, is attr inert? If so, is safe */
+    if (URI_SAFE_ATTRIBUTES[lcName]) {
+      return true;
+    }
+
+    /* Check no script, data or unknown possibly unsafe URI
         unless we know URI values are safe for that attribute */
-    } else if (
-      regExpTest(IS_ALLOWED_URI, stringReplace(value, ATTR_WHITESPACE, ''))
-    ) {
-      // This attribute is safe
-      /* Keep image data URIs alive if src/xlink:href is allowed */
-      /* Further prevent gadget XSS for dynamically built script tags */
-    } else if (
+    if (regExpTest(IS_ALLOWED_URI, stringReplace(value, ATTR_WHITESPACE, ''))) {
+      return true;
+    }
+
+    /* Keep image data URIs alive if src/xlink:href is allowed */
+    /* Further prevent gadget XSS for dynamically built script tags */
+    if (
       (lcName === 'src' || lcName === 'xlink:href' || lcName === 'href') &&
       lcTag !== 'script' &&
       stringIndexOf(value, 'data:') === 0 &&
       DATA_URI_TAGS[lcTag]
     ) {
-      // This attribute is safe
-      /* Allow unknown protocols: This provides support for links that
+      return true;
+    }
+
+    /* Allow unknown protocols: This provides support for links that
         are handled by protocol handlers which may be unknown ahead of
         time, e.g. fb:, spotify: */
-    } else if (
+    if (
       ALLOW_UNKNOWN_PROTOCOLS &&
       !regExpTest(IS_SCRIPT_OR_DATA, stringReplace(value, ATTR_WHITESPACE, ''))
     ) {
-      // This attribute is safe
-      /* Check for binary attributes */
-    } else if (value) {
-      return false;
-    } else {
-      // Binary attributes are safe at this point
-      /* Anything else, presume unsafe, do not add it back */
+      return true;
     }
 
-    return true;
+    /* Only an empty (binary) value remains safe at this point;
+       anything else is presumed unsafe, do not add it back */
+    return !value;
   };
 
   /* Names the HTML spec reserves from valid-custom-element-name; these must
@@ -2250,6 +2492,15 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       return;
     }
 
+    /* Same lazy-install guard as uponSanitizeElement (see there): fork the
+       attribute allowlist off its shared binding before a hook can see it. */
+    ALLOWED_ATTR = _forkSharedAllowlist(
+      hooks.uponSanitizeAttribute,
+      ALLOWED_ATTR,
+      DEFAULT_ALLOWED_ATTR,
+      SET_CONFIG_ALLOWED_ATTR
+    );
+
     const hookEvent = {
       attrName: '',
       attrValue: '',
@@ -2286,7 +2537,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
         stringIndexOf(value, SANITIZE_NAMED_PROPS_PREFIX) !== 0
       ) {
         // Remove the attribute with this value
-        _removeAttribute(name, currentNode);
+        _removeAttribute(name, currentNode, attr);
         // Prefix the value and later re-create the attribute with the sanitized value
         value = SANITIZE_NAMED_PROPS_PREFIX + value;
       }
@@ -2301,13 +2552,13 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
           value
         )
       ) {
-        _removeAttribute(name, currentNode);
+        _removeAttribute(name, currentNode, attr);
         continue;
       }
 
       /* Make sure we cannot easily use animated hrefs, even if animations are allowed */
       if (lcName === 'attributename' && stringMatch(value, 'href')) {
-        _removeAttribute(name, currentNode);
+        _removeAttribute(name, currentNode, attr);
         continue;
       }
 
@@ -2318,7 +2569,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
 
       /* Did the hooks approve of the attribute? */
       if (!hookEvent.keepAttr) {
-        _removeAttribute(name, currentNode);
+        _removeAttribute(name, currentNode, attr);
         continue;
       }
 
@@ -2327,7 +2578,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
         !ALLOW_SELF_CLOSE_IN_ATTR &&
         regExpTest(EXPRESSIONS.SELF_CLOSING_TAG, value)
       ) {
-        _removeAttribute(name, currentNode);
+        _removeAttribute(name, currentNode, attr);
         continue;
       }
 
@@ -2338,7 +2589,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
 
       /* Is `value` valid for this attribute? */
       if (!_isValidAttribute(lcTag, lcName, value)) {
-        _removeAttribute(name, currentNode);
+        _removeAttribute(name, currentNode, attr);
         continue;
       }
 
@@ -2395,10 +2646,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
          Walk it explicitly. The nodeType guard avoids reading
          shadowRoot off text / comment / CDATA / PI nodes that the
          iterator also surfaces. */
-      const shadowNodeType = getNodeType
-        ? getNodeType(shadowNode)
-        : shadowNode.nodeType;
-      if (shadowNodeType === NODE_TYPE.element) {
+      if (_readNodeType(shadowNode) === NODE_TYPE.element) {
         const innerSr = getShadowRoot(shadowNode);
         if (_isDocumentFragment(innerSr)) {
           _sanitizeAttachedShadowRoots(innerSr);
@@ -2460,7 +2708,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
       }
 
       const node = item.node;
-      const nodeType = getNodeType ? getNodeType(node) : (node as any).nodeType;
+      const nodeType = _readNodeType(node);
       const isElement = nodeType === NODE_TYPE.element;
 
       /* (pushed last → processed first) Children, snapshotted in reverse so
@@ -2589,9 +2837,7 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
          child named "nodeName" on the form root would otherwise shadow
          the property and let this check skip the root-allowlist
          validation entirely. */
-      const nn = getNodeName
-        ? getNodeName(dirty as Node)
-        : (dirty as Node).nodeName;
+      const nn = _readNodeName(dirty as Node);
       if (typeof nn === 'string') {
         const tagName = transformCaseFunc(nn);
         if (!ALLOWED_TAGS[tagName] || FORBID_TAGS[tagName]) {
@@ -2692,18 +2938,21 @@ function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
 
     /* Get node iterator */
     const walkRoot: Node = inPlace ? (dirty as Node) : body;
-    const nodeIterator = _createNodeIterator(walkRoot);
 
     /* Now start iterating over the created document.
        The walk runs inside an exception barrier (campaign-3 F2): a re-entrant
        engine/custom-element mutation can detach a node mid-walk so
        `_forceRemove`'s parentless guard throws, aborting the loop. Without the
        barrier the caller's in-place tree would be left half-sanitized with the
-       unvisited tail still armed. On any throw we fail closed — strip the
-       in-place root bare — then rethrow so the existing throw contract is
-       preserved. (String/DOM-copy paths never return the partial body, so the
-       propagating throw is already fail-closed there.) */
+       unvisited tail still armed. _createNodeIterator itself is inside the
+       barrier too: constructing the iterator dereferences the root's document,
+       and any failure there (e.g. an exotic/clobbered root) must still fail
+       closed rather than skip the neutralize. On any throw we fail closed -
+       strip the in-place root bare - then rethrow so the existing throw
+       contract is preserved. (String/DOM-copy paths never return the partial
+       body, so the propagating throw is already fail-closed there.) */
     try {
+      const nodeIterator = _createNodeIterator(walkRoot);
       while ((currentNode = nodeIterator.nextNode())) {
         /* Sanitize tags and elements */
         _sanitizeElements(currentNode, walkRoot);