Changeset 58424 for trunk/src/wp-includes/kses.php

Timestamp:

06/17/2024 12:02:50 PM (6 months ago)

Author:

dmsnell

Message:

KSES: Fix tests and detection of HTML Bogus Comment spans.

In [58418] a test was added without the test_ prefix in its function
name, and because of that, it wasn't run in the test suite.
The prefix has been added to ensure that it runs.

In the original patch, due to a logical bug, a recursive loop to
transform the inside contents of the bogus comments was never run
more than once. This has been fixed.

This patch also includes one more case where kses wasn't
properly detecting the bogus comment state, and adds a test case
to cover this. It limits itself to some but not all constructions
of invalid markup declaration so that it doesn't conflict with
existing behaviors around those and other kinds of invalid comments.

Props ellatrix, dmsnell.
See #61009.
Follow-up to [58418].

File:

• trunk/src/wp-includes/kses.php (modified) (2 diffs)

trunk/src/wp-includes/kses.php

@@ -989 +989 @@
         |
         </[^a-zA-Z][^>]*>  #  - Closing tags with invalid tag names.
+        |
+        <![^>]*>           #  - Invalid markup declaration nodes. Not all invalid nodes
+                           #    are matched so as to avoid breaking legacy behaviors.
     )
     |
@@ -1115 +1118 @@
 
     /*
-     * When a closing tag appears with a name that isn't a valid tag name,
-     * it must be interpreted as an HTML comment. It extends until the
-     * first `>` character after the initial opening `</`.
+     * When certain invalid syntax constructs appear, the HTML parser
+     * shifts into what's called the "bogus comment state." This is a
+     * plaintext state that consumes everything until the nearest `>`
+     * and then transforms the entire span into an HTML comment.
      *
      * Preserve these comments and do not treat them like tags.
+     *
+     * @see https://html.spec.whatwg.org/#bogus-comment-state
      */
-    if ( 1 === preg_match( '~^</[^a-zA-Z][^>]*>$~', $content ) ) {
-        $content     = substr( $content, 2, -1 );
-        $transformed = null;
-
-        while ( $transformed !== $content ) {
-            $transformed = wp_kses( $content, $allowed_html, $allowed_protocols );
-            $content     = $transformed;
-        }
-
-        return "</{$transformed}>";
+    if ( 1 === preg_match( '~^(?:</[^a-zA-Z][^>]*>|<![a-z][^>]*>)$~', $content ) ) {
+        /**
+         * Since the pattern matches `</…>` and also `<!…>`, this will
+         * preserve the type of the cleaned-up token in the output.
+         */
+        $opener  = $content[1];
+        $content = substr( $content, 2, -1 );
+
+        do {
+            $prev    = $content;
+            $content = wp_kses( $content, $allowed_html, $allowed_protocols );
+        } while ( $prev !== $content );
+
+        // Recombine the modified inner content with the original token structure.
+        return "<{$opener}{$content}>";
     }