fix: accept unterminated block comments at EOF in scanner

src/scanner.c

@@ -224,6 +224,16 @@ static bool scan_multiline_comment(TSLexer *lexer) {
         }
         break;
       case '\0':
+        // Accept unterminated block comments at EOF rather than rejecting them.
+        // This matches JetBrains PSI behavior which recognizes unclosed /* as a
+        // BLOCK_COMMENT token (plus an error element). Without this, the scanner
+        // returns false and tree-sitter tries to parse the comment delimiters
+        // as operators/expressions.
+        if (lexer->eof(lexer)) {
+          lexer->result_symbol = MULTILINE_COMMENT;
+          lexer->mark_end(lexer);
+          return true;
+        }
         return false;
       default:
         advance(lexer);

test/corpus/jetbrains/BlockCommentAtBeginningOfFile1.txt

@@ -0,0 +1,12 @@
+==================
+BlockCommentAtBeginningOfFile1
+==================
+
+// COMPILATION_ERRORS
+
+/*
+---
+
+(source_file
+  (line_comment)
+  (multiline_comment))

test/corpus/jetbrains/BlockCommentAtBeginningOfFile2.txt

@@ -0,0 +1,13 @@
+==================
+BlockCommentAtBeginningOfFile2
+==================
+
+// COMPILATION_ERRORS
+
+/*
+/*
+---
+
+(source_file
+  (line_comment)
+  (multiline_comment))

test/corpus/jetbrains/BlockCommentAtBeginningOfFile3.txt

@@ -0,0 +1,14 @@
+==================
+BlockCommentAtBeginningOfFile3
+==================
+
+// COMPILATION_ERRORS
+
+/*
+
+fooo
+---
+
+(source_file
+  (line_comment)
+  (multiline_comment))

test/corpus/jetbrains/BlockCommentAtBeginningOfFile4.txt

@@ -0,0 +1,16 @@
+==================
+BlockCommentAtBeginningOfFile4
+==================
+
+// COMPILATION_ERRORS
+
+/*
+
+/*foo*/
+
+asdfas
+---
+
+(source_file
+  (line_comment)
+  (multiline_comment))

tools/cross-validation/excluded.txt

@@ -20,13 +20,9 @@
 #   3. Run: npm test (to verify the new corpus test passes)
 
 # =============================================================================
-# MISMATCH: Grammar produces wrong AST structure (26 files)
+# MISMATCH: Grammar produces wrong AST structure (24 files)
 # =============================================================================
 
-# --- unterminated_block_comment (tree-sitter limitation) ---
-BlockCommentAtBeginningOfFile3
-BlockCommentAtBeginningOfFile4
-
 # --- duplicate_accessor (error recovery difference) ---
 DuplicateAccessor
 
@@ -88,8 +84,6 @@ AbsentInnerType
 AnnotatedIntersections
 AssertNotNull
 BackslashInString
-BlockCommentAtBeginningOfFile1
-BlockCommentAtBeginningOfFile2
 CallsInWhen
 CollectionLiterals
 CommentsBinding