[Morriar]

Summary

Prevent malformed UTF-8 in inline RBS comments from raising ArgumentError.

CommentBlock#trailing_annotation identified annotation comments using a regexp:

comment.location.slice.start_with?(/#[:\[]/)

For a comment containing malformed UTF-8, this operation raises before the annotation reaches the RBS parser:

String#start_with?: invalid byte sequence in UTF-8 (ArgumentError)

Fix

Use literal prefix matching:

comment.location.slice.start_with?("#:", "#[")

This check does not validate the rest of the comment as UTF-8. The malformed annotation therefore reaches the existing RBS lexer, which reports a parsing error. CommentBlock converts that error into an AnnotationSyntaxError diagnostic.

Reproduction

A Ruby source containing a literal invalid byte in an inline annotation comment previously raised ArgumentError:

class Foo
  attr_reader :value #: Integer # <0xDD>
end

Tests

Added an end-to-end Prism.parse → RBS::InlineParser.parse regression test covering:

• 0xDD
• 0xFF
• lone 0xC3
• truncated 0xE2 0x82

Each case now produces RBS::InlineParser::Diagnostic::AnnotationSyntaxError instead of raising ArgumentError.