Robustesse : appliquer la limite de taille après normalisation Unicode, nettoyer .gitignore

.gitignore

@@ -9,10 +9,6 @@
 .phpunit.result.cache
 .phpunit.cache
 
-# PHPUnit
-/app/phpunit.xml
-/phpunit.xml
-
 # Logs
 /logs/
 *.log

src/Lexer/Lexer.php

@@ -651,7 +651,7 @@ private function quoteLineState(string $content, bool $paragraphOpen, bool $fenc
         }
         $isText = $this->isParagraphContinuation($content)
             || ($paragraphOpen && $content !== '' && !ctype_space($content) && !preg_match(self::PATTERN_HEADING, $content));
-        return [$isText && !preg_match(self::PATTERN_TABLE_SEPARATOR, $content), false];
+        return [$isText, false];
     }
 
     private function matchLine(string $line): Token

src/MarkdownParser.php

@@ -49,7 +49,8 @@ public function __construct(
     }
 
     /**
-     * @throws ParseException if input exceeds maxBytes or is not valid UTF-8.
+     * @throws ParseException if input exceeds maxBytes (before or after NFC normalization)
+     *                        or is not valid UTF-8.
      */
     private function guard(string &$markdown): void
     {
@@ -63,6 +64,14 @@ private function guard(string &$markdown): void
         }
         $normalized = $this->normalizer->normalize($markdown);
         $markdown = $normalized !== false ? $normalized : $markdown;
+
+        // NFC can grow the input (e.g. U+0958 has no composed form and doubles in size):
+        // the limit applies to what the parser actually processes.
+        if (strlen($markdown) > $this->maxBytes) {
+            throw new ParseException(
+                sprintf('Input exceeds maximum allowed size of %d bytes after Unicode normalization.', $this->maxBytes),
+            );
+        }
     }
 
     /**

tests/Integration/MarkdownParserTest.php

@@ -131,6 +131,17 @@ public function testMaxBytesAllowsInputAtExactByteLimit(): void
         $this->assertNotEmpty($html);
     }
 
+    public function testMaxBytesIsEnforcedAfterNormalization(): void
+    {
+        // U+0958 (3 bytes) has no composed form: NFC turns it into 2 code points (6 bytes).
+        // 10 × 3 = 30 bytes fits the limit before normalization, 60 bytes does not after.
+        $parser = new MarkdownParser(maxBytes: 40);
+
+        $this->expectException(ParseException::class);
+        $this->expectExceptionMessage('after Unicode normalization');
+        $parser->parse(str_repeat("\u{0958}", 10));
+    }
+
     // ── NFC normalisation ─────────────────────────────────────────────────────
 
     public function testNfcNormalisationConvertsNfdInput(): void