Merge pull request #3988 from ruby/track-newlines-in-escapes

Track newlines in character escape sequences

Author: kddnewton

src/prism.c

@@ -8594,6 +8594,7 @@ escape_write_escape_encoded(pm_parser_t *parser, pm_buffer_t *buffer, pm_buffer_
     }
 
     if (width == 1) {
+        if (*parser->current.end == '\n') pm_line_offset_list_append(&parser->line_offsets, PM_TOKEN_END(parser, &parser->current) + 1);
         escape_write_byte(parser, buffer, regular_expression_buffer, flags, escape_byte(*parser->current.end++, flags));
     } else if (width > 1) {
         // Valid multibyte character.  Just ignore escape.
@@ -8910,6 +8911,7 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, pm_buffer_t *regular_expre
                         return;
                     }
 
+                    if (peeked == '\n') pm_line_offset_list_append(&parser->line_offsets, PM_TOKEN_END(parser, &parser->current) + 1);
                     parser->current.end++;
                     escape_write_byte(parser, buffer, regular_expression_buffer, flags, escape_byte(peeked, flags | PM_ESCAPE_FLAG_CONTROL));
                     return;
@@ -8968,6 +8970,7 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, pm_buffer_t *regular_expre
                         return;
                     }
 
+                    if (peeked == '\n') pm_line_offset_list_append(&parser->line_offsets, PM_TOKEN_END(parser, &parser->current) + 1);
                     parser->current.end++;
                     escape_write_byte(parser, buffer, regular_expression_buffer, flags, escape_byte(peeked, flags | PM_ESCAPE_FLAG_CONTROL));
                     return;
@@ -9021,13 +9024,15 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, pm_buffer_t *regular_expre
                         return;
                     }
 
+                    if (peeked == '\n') pm_line_offset_list_append(&parser->line_offsets, PM_TOKEN_END(parser, &parser->current) + 1);
                     parser->current.end++;
                     escape_write_byte(parser, buffer, regular_expression_buffer, flags, escape_byte(peeked, flags | PM_ESCAPE_FLAG_META));
                     return;
             }
         }
         case '\r': {
             if (peek_offset(parser, 1) == '\n') {
+                pm_line_offset_list_append(&parser->line_offsets, PM_TOKEN_END(parser, &parser->current) + 2);
                 parser->current.end += 2;
                 escape_write_byte_encoded(parser, buffer, flags, escape_byte('\n', flags));
                 return;

test/prism/newline_offsets_test.rb

@@ -8,15 +8,38 @@ class NewlineOffsetsTest < TestCase
       define_method(fixture.test_name) { assert_newline_offsets(fixture) }
     end
 
+    def test_escape_control_newline
+      # Newlines consumed inside escape sequences like \C-, \c, and \M-
+      # must be tracked in line offsets across all literal types.
+      %w[\\C- \\c \\M-].each do |escape|
+        assert_newline_offsets_for("\"#{escape}\n\"", "#{escape} in string")
+        assert_newline_offsets_for("`#{escape}\n`", "#{escape} in xstring")
+        assert_newline_offsets_for("/#{escape}\n/", "#{escape} in regexp")
+        assert_newline_offsets_for("%Q{#{escape}\n}", "#{escape} in %Q")
+        assert_newline_offsets_for("%W[#{escape}\n]", "#{escape} in %W")
+        assert_newline_offsets_for("<<~H\n#{escape}\n\nH\n", "#{escape} in heredoc")
+        assert_newline_offsets_for("?#{escape}\n", "#{escape} in char literal")
+      end
+
+      # Combined meta + control escapes
+      assert_newline_offsets_for("\"\\M-\\C-\n\"", "\\M-\\C- in string")
+      assert_newline_offsets_for("\"\\M-\\c\n\"", "\\M-\\c in string")
+
+      # \r\n consumed inside escape context
+      assert_newline_offsets_for("\"\\C-\r\n\"", "\\C- with \\r\\n")
+    end
+
     private
 
     def assert_newline_offsets(fixture)
-      source = fixture.read
+      assert_newline_offsets_for(fixture.read)
+    end
 
+    def assert_newline_offsets_for(source, message = nil)
       expected = [0]
       source.b.scan("\n") { expected << $~.offset(0)[0] + 1 }
 
-      assert_equal expected, Prism.parse(source).source.offsets
+      assert_equal expected, Prism.parse(source).source.offsets, message
     end
   end
 end