From dc90c26a103ad62df73464cc1896edbcc90bd0c7 Mon Sep 17 00:00:00 2001
From: Earlopain <14981592+Earlopain@users.noreply.github.com>
Date: Fri, 8 May 2026 21:04:08 +0200
Subject: [ruby/prism] Respect `encoding` option in `Prism.lex` and friends

utf-8 is the default for source files but can be overwritten via options

https://github.com/ruby/prism/commit/355f451528
---
 prism/extension.c      |  2 +-
 test/prism/lex_test.rb | 18 ++++++++++++++++++
 2 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/prism/extension.c b/prism/extension.c
index 9f9169cfff..27df8dac50 100644
--- a/prism/extension.c
+++ b/prism/extension.c
@@ -793,7 +793,7 @@ parse_lex_input(const uint8_t *input, size_t input_length, const pm_options_t *o
     parse_lex_data_t parse_lex_data = {
         .source = source,
         .tokens = rb_ary_new(),
-        .encoding = rb_utf8_encoding(),
+        .encoding = rb_enc_find(pm_parser_encoding_name(parser)),
         .freeze = pm_options_freeze(options),
     };
 
diff --git a/test/prism/lex_test.rb b/test/prism/lex_test.rb
index 8ea7ce7e9b..1e06d52184 100644
--- a/test/prism/lex_test.rb
+++ b/test/prism/lex_test.rb
@@ -47,6 +47,24 @@ module Prism
       end
     end
 
+    def test_lex_encoding
+      tokens = Prism.lex('"わたし"', encoding: Encoding::Windows_31J).value
+      tokens.each do |t|
+        assert_equal(Encoding::Windows_31J, t[0].value.encoding)
+      end
+
+      # Shebangs must appear on the first line. For these cases, the encoding
+      # comment may appear second, but it should still change encoding.
+      tokens = Prism.lex(<<~RUBY, encoding: Encoding::Windows_31J).value
+        #! /usr/bin/env ruby
+        # encoding: utf-8
+        "わたし"
+      RUBY
+      tokens.each do |t|
+        assert_equal(Encoding::UTF_8, t[0].value.encoding)
+      end
+    end
+
     if RUBY_VERSION >= "3.3"
       def test_lex_compat
         source = "foo bar"
-- 
cgit v1.2.3