UTF LE is fixed at least the first 2 bytes

* io.c (io_strip_bom): if the first 2 bytes are 0xFF0xFE, it
should be a little-endian UTF, 16 or 32.  [Bug #16099]

3 files changed, 5 insertions, 8 deletions

diff --git a/io.c b/io.c
index 69dbc93cff..6a4c5b978b 100644
--- a/io.c
+++ b/io.c
@@ -6136,12 +6136,9 @@ io_strip_bom(VALUE io)
 		    return ENCINDEX_UTF_32LE;
 		}
 		rb_io_ungetbyte(io, b4);
-		rb_io_ungetbyte(io, b3);
-	    }
-	    else {
-		rb_io_ungetbyte(io, b3);
-		return ENCINDEX_UTF_16LE;
 	    }
+            rb_io_ungetbyte(io, b3);
+            return ENCINDEX_UTF_16LE;
 	}
 	rb_io_ungetbyte(io, b2);
 	break;
diff --git a/test/ruby/test_file.rb b/test/ruby/test_file.rb
index 36c154d36c..3deab76e93 100644
--- a/test/ruby/test_file.rb
+++ b/test/ruby/test_file.rb
@@ -87,7 +87,7 @@ class TestFile < Test::Unit::TestCase
   end
 
   def test_bom_32le
-    assert_bom(["\xFF\xFE\0", "\0"], __method__)
+    assert_bom(["\xFF", "\xFE\0\0"], __method__)
   end
 
   def test_truncate_wbuf
diff --git a/test/ruby/test_io_m17n.rb b/test/ruby/test_io_m17n.rb
index 630f2eec08..8101bfb62f 100644
--- a/test/ruby/test_io_m17n.rb
+++ b/test/ruby/test_io_m17n.rb
@@ -2084,8 +2084,8 @@ EOT
     define_method("test_strip_bom:#{name}") do
       path = "#{name}-bom.txt"
       with_tmpdir {
-        text = "\uFEFFa"
-        stripped = "a"
+        text = "\uFEFF\u0100a"
+        stripped = "\u0100a"
         content = text.encode(name)
         generate_file(path, content)
         result = File.read(path, mode: 'rb:BOM|UTF-8')