Do not leak explicit encoding

Fixes [Bug #20744]

Author: kddnewton

lib/prism/translation/ruby_parser.rb

@@ -1428,7 +1428,13 @@ def visit_statements_node(node)
  # "foo"
  # ^^^^^
  def visit_string_node(node)
- s(node, :str, node.unescaped)
+ unescaped = node.unescaped
+
+ if node.forced_binary_encoding?
+ unescaped.force_encoding(Encoding::BINARY)
+ end
+
+ s(node, :str, unescaped)
  end
 
  # super(foo)

src/prism.c

@@ -168,6 +168,7 @@ lex_mode_push_regexp(pm_parser_t *parser, uint8_t incrementor, uint8_t terminato
  breakpoints[index++] = incrementor;
  }
 
+ parser->explicit_encoding = NULL;
  return lex_mode_push(parser, lex_mode);
 }
 

test/prism/fixtures/regex_escape_encoding.txt

@@ -0,0 +1,3 @@
+# encoding: US-ASCII
+str = "hello \xFC"
+str =~ /hello \u{fc}/

test/prism/locals_test.rb

@@ -9,6 +9,10 @@
 # to test on the most recent versions.
 return if !defined?(RubyVM::InstructionSequence) || RUBY_VERSION < "3.4.0"
 
+# If we're on Ruby 3.4.0 and the default parser is Prism, then there is no point
+# in comparing the locals because they will be the same.
+return if RubyVM::InstructionSequence.compile("").to_a[4][:parser] == :prism
+
 # In Ruby 3.4.0, the local table for method forwarding changed. But 3.4.0 can
 # refer to the dev version, so while 3.4.0 still isn't released, we need to
 # check if we have a high enough revision.

test/prism/snapshots/regex_escape_encoding.txt

@@ -209,7 +209,7 @@ def self.each_encoding
 
  private
 
- if RUBY_ENGINE == "ruby"
+ if RUBY_ENGINE == "ruby" && RubyVM::InstructionSequence.compile("").to_a[4][:parser] != :prism
  # Check that the given source is valid syntax by compiling it with RubyVM.
  def check_syntax(source)
  ignore_warnings { RubyVM::InstructionSequence.compile(source) }