Fix quadratic complexity parsing long backtick code spans with no matching closers

colinodell · colinodell · commit 540d850394a9 · 2024-12-07T10:17:04.000-05:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -110,6 +110,7 @@ Updates should follow the [Keep a CHANGELOG](https://keepachangelog.com/) princi
 - Fixed quadratic complexity parsing emphasis and strikethrough delimiters
 - Fixed issue where having 500,000+ delimiters could trigger a [known segmentation fault issue in PHP's garbage collection](https://bugs.php.net/bug.php?id=68606)
 - Fixed quadratic complexity deactivating link openers
+- Fixed quadratic complexity parsing long backtick code spans with no matching closers
 - Fixed catastrophic backtracking when parsing link labels/titles
 
 ## [2.4.1] - 2023-08-30
diff --git a/src/Extension/CommonMark/Parser/Inline/BacktickParser.php b/src/Extension/CommonMark/Parser/Inline/BacktickParser.php
@@ -18,12 +18,27 @@
 
 use League\CommonMark\Extension\CommonMark\Node\Inline\Code;
 use League\CommonMark\Node\Inline\Text;
+use League\CommonMark\Parser\Cursor;
 use League\CommonMark\Parser\Inline\InlineParserInterface;
 use League\CommonMark\Parser\Inline\InlineParserMatch;
 use League\CommonMark\Parser\InlineParserContext;
 
 final class BacktickParser implements InlineParserInterface
 {
+    /**
+     * Max bound for backtick code span delimiters.
+     *
+     * @see https://github.com/commonmark/cmark/commit/8ed5c9d
+     */
+    private const MAX_BACKTICKS = 1000;
+
+    /** @var \WeakReference<Cursor>|null */
+    private ?\WeakReference $lastCursor = null;
+    private bool $lastCursorScanned     = false;
+
+    /** @var array<int, int> backtick count => position of known ender */
+    private array $seenBackticks = [];
+
     public function getMatchDefinition(): InlineParserMatch
     {
         return InlineParserMatch::regex('`+');
@@ -38,11 +53,7 @@ public function parse(InlineParserContext $inlineContext): bool
         $currentPosition = $cursor->getPosition();
         $previousState   = $cursor->saveState();
 
-        while ($matchingTicks = $cursor->match('/`+/m')) {
-            if ($matchingTicks !== $ticks) {
-                continue;
-            }
-
+        if ($this->findMatchingTicks(\strlen($ticks), $cursor)) {
             $code = $cursor->getSubstring($currentPosition, $cursor->getPosition() - $currentPosition - \strlen($ticks));
 
             $c = \preg_replace('/\n/m', ' ', $code) ?? '';
@@ -67,4 +78,55 @@ public function parse(InlineParserContext $inlineContext): bool
 
         return true;
     }
+
+    /**
+     * Locates the matching closer for a backtick code span.
+     *
+     * Leverages some caching to avoid traversing the same cursor multiple times when
+     * we've already seen all the potential backtick closers.
+     *
+     * @see https://github.com/commonmark/cmark/commit/8ed5c9d
+     *
+     * @param int    $openTickLength Number of backticks in the opening sequence
+     * @param Cursor $cursor         Cursor to scan
+     *
+     * @return bool True if a matching closer was found, false otherwise
+     */
+    private function findMatchingTicks(int $openTickLength, Cursor $cursor): bool
+    {
+        // Reset the seenBackticks cache if this is a new cursor
+        if ($this->lastCursor === null || $this->lastCursor->get() !== $cursor) {
+            $this->seenBackticks     = [];
+            $this->lastCursor        = \WeakReference::create($cursor);
+            $this->lastCursorScanned = false;
+        }
+
+        if ($openTickLength > self::MAX_BACKTICKS) {
+            return false;
+        }
+
+        // Return if we already know there's no closer
+        if ($this->lastCursorScanned && isset($this->seenBackticks[$openTickLength]) && $this->seenBackticks[$openTickLength] <= $cursor->getPosition()) {
+            return false;
+        }
+
+        while ($ticks = $cursor->match('/`{1,' . self::MAX_BACKTICKS . '}/m')) {
+            $numTicks = \strlen($ticks);
+
+            // Did we find the closer?
+            if ($numTicks === $openTickLength) {
+                return true;
+            }
+
+            // Store position of closer
+            if ($numTicks <= self::MAX_BACKTICKS) {
+                $this->seenBackticks[$numTicks] = $cursor->getPosition() - $numTicks;
+            }
+        }
+
+        // Got through whole input without finding closer
+        $this->lastCursorScanned = true;
+
+        return false;
+    }
 }
