Merge pull request #2974 from ksss/ksss/fix-invalid-position-range

Reject reversed position ranges in parser/lexer entrypoints

Author: ksss

ext/rbs_extension/main.c

@@ -145,10 +145,17 @@ static VALUE parse_type_try(VALUE a) {
     return rbs_struct_to_ruby_value(ctx, type);
 }
 
-static rbs_lexer_t *alloc_lexer_from_buffer(rbs_allocator_t *allocator, VALUE string, rb_encoding *encoding, int start_pos, int end_pos) {
+static void validate_position_range(int start_pos, int end_pos) {
     if (start_pos < 0 || end_pos < 0) {
         rb_raise(rb_eArgError, "negative position range: %d...%d", start_pos, end_pos);
     }
+    if (start_pos > end_pos) {
+        rb_raise(rb_eArgError, "invalid position range: %d...%d", start_pos, end_pos);
+    }
+}
+
+static rbs_lexer_t *alloc_lexer_from_buffer(rbs_allocator_t *allocator, VALUE string, rb_encoding *encoding, int start_pos, int end_pos) {
+    validate_position_range(start_pos, end_pos);
 
     const char *encoding_name = rb_enc_name(encoding);
 
@@ -162,9 +169,7 @@ static rbs_lexer_t *alloc_lexer_from_buffer(rbs_allocator_t *allocator, VALUE st
 }
 
 static rbs_parser_t *alloc_parser_from_buffer(VALUE buffer, int start_pos, int end_pos) {
-    if (start_pos < 0 || end_pos < 0) {
-        rb_raise(rb_eArgError, "negative position range: %d...%d", start_pos, end_pos);
-    }
+    validate_position_range(start_pos, end_pos);
 
     VALUE string = rb_funcall(buffer, rb_intern("content"), 0);
     StringValue(string);

test/rbs/parser_test.rb

@@ -1030,6 +1030,21 @@ class Foo[T < Integer] < Bar # Comment
     assert_equal [:pEOF, '', 57...57], tokens.shift.then { |t| [t[0], t[1].source, t[1].range] }
   end
 
+  def test_invalid_position_range_raises
+    # Regression: start_pos > end_pos used to cause an infinite loop in the lexer.
+    assert_raises(ArgumentError) do
+      RBS::Parser._parse_signature(buffer(""), 1, 0)
+    end
+  end
+
+  def test_invalid_byte_range_in_parse_type_raises
+    # Regression: parse_type's byte_range: keyword reaches _parse_type directly,
+    # which used to hang on reversed ranges.
+    assert_raises(ArgumentError) do
+      RBS::Parser.parse_type("", byte_range: 1..0)
+    end
+  end
+
   def test_invalid_utf8_byte_in_comment_does_not_hang
     # Regression: invalid UTF-8 byte in a comment used to loop forever in the lexer.
     source = "# \xC2".dup.force_encoding(Encoding::UTF_8)