From d77a4c5ab43a2ea313b6a229fa754154d6532ca7 Mon Sep 17 00:00:00 2001
From: Pradyumn jha <devpryan7792@users.noreply.github.com>
Date: Sat, 10 Oct 2026 00:38:08 +0530
Subject: [PATCH] gh-159071: Correct UnicodeDecodeError offsets in utf-8-sig
 codec when input starts with BOM

Signed-off-by: Pradyumn jha <devpryan7792@users.noreply.github.com>
---
 Lib/encodings/utf_8_sig.py                    | 16 +++---
 Lib/test/test_codecs.py                       | 53 +++++++++++++++++++
 ...-10-10-00-37-12.gh-issue-159071.a62BQP.rst |  2 +
 3 files changed, 62 insertions(+), 9 deletions(-)
 create mode 100644 Misc/NEWS.d/next/Library/2026-10-10-00-37-12.gh-issue-159071.a62BQP.rst

diff --git a/Lib/encodings/utf_8_sig.py b/Lib/encodings/utf_8_sig.py
index cc895cddcc561bc..b94caf860819fc3 100644
--- a/Lib/encodings/utf_8_sig.py
+++ b/Lib/encodings/utf_8_sig.py
@@ -16,12 +16,10 @@ def encode(input, errors='strict'):
             len(input))
 
 def decode(input, errors='strict'):
-    prefix = 0
     if input[:3] == codecs.BOM_UTF8:
-        input = input[3:]
-        prefix = 3
-    (output, consumed) = codecs.utf_8_decode(input, errors, True)
-    return (output, consumed+prefix)
+        (output, consumed) = codecs.utf_8_decode(input, errors, True)
+        return (output[1:], consumed)
+    return codecs.utf_8_decode(input, errors, True)
 
 class IncrementalEncoder(codecs.IncrementalEncoder):
     def __init__(self, errors='strict'):
@@ -64,8 +62,8 @@ def _buffer_decode(self, input, errors, final):
                 self.first = 0
                 if input[:3] == codecs.BOM_UTF8:
                     (output, consumed) = \
-                       codecs.utf_8_decode(input[3:], errors, final)
-                    return (output, consumed+3)
+                       codecs.utf_8_decode(input, errors, final)
+                    return (output[1:], consumed)
         return codecs.utf_8_decode(input, errors, final)
 
     def reset(self):
@@ -110,8 +108,8 @@ def decode(self, input, errors='strict'):
                 return ("", 0)
         elif input[:3] == codecs.BOM_UTF8:
             self.decode = codecs.utf_8_decode
-            (output, consumed) = codecs.utf_8_decode(input[3:],errors)
-            return (output, consumed+3)
+            (output, consumed) = codecs.utf_8_decode(input, errors)
+            return (output[1:], consumed)
         # (else) no BOM present
         self.decode = codecs.utf_8_decode
         return codecs.utf_8_decode(input, errors)
diff --git a/Lib/test/test_codecs.py b/Lib/test/test_codecs.py
index 37ac57063c35390..9b5f475e715ec9c 100644
--- a/Lib/test/test_codecs.py
+++ b/Lib/test/test_codecs.py
@@ -1169,6 +1169,59 @@ def test_stream_bare(self):
             got = ostream.getvalue()
             self.assertEqual(got, unistring)
 
+    def test_decode_error_offsets_with_bom(self):
+        # gh-159071: UnicodeDecodeError offsets should count the BOM
+        data = codecs.BOM_UTF8 + b"abc\xff"
+        with self.assertRaises(UnicodeDecodeError) as cm:
+            data.decode("utf-8-sig")
+        exc = cm.exception
+        self.assertEqual(exc.object, data)
+        self.assertEqual(exc.start, 6)
+        self.assertEqual(exc.end, 7)
+
+        # Test error right after BOM
+        data = codecs.BOM_UTF8 + b"\xff"
+        with self.assertRaises(UnicodeDecodeError) as cm:
+            data.decode("utf-8-sig")
+        exc = cm.exception
+        self.assertEqual(exc.object, data)
+        self.assertEqual(exc.start, 3)
+        self.assertEqual(exc.end, 4)
+
+        # Test IncrementalDecoder
+        dec = codecs.getincrementaldecoder("utf-8-sig")()
+        with self.assertRaises(UnicodeDecodeError) as cm:
+            dec.decode(data, final=True)
+        exc = cm.exception
+        self.assertEqual(exc.object, data)
+        self.assertEqual(exc.start, 3)
+        self.assertEqual(exc.end, 4)
+
+        # Test StreamReader
+        reader = codecs.getreader("utf-8-sig")(io.BytesIO(data))
+        with self.assertRaises(UnicodeDecodeError) as cm:
+            reader.read()
+        exc = cm.exception
+        self.assertEqual(exc.object, data)
+        self.assertEqual(exc.start, 3)
+        self.assertEqual(exc.end, 4)
+
+        # Test custom error handler receives correct offsets
+        received = []
+        def handler(err):
+            received.append((err.object, err.start, err.end))
+            return ("?", err.end)
+
+        old_handler = codecs.lookup_error("replace")
+        codecs.register_error("test_utf8sig_handler", handler)
+        try:
+            result = data.decode("utf-8-sig", "test_utf8sig_handler")
+            self.assertEqual(result, "?")
+            self.assertEqual(received, [(data, 3, 4)])
+        finally:
+            # clean up error handler if needed
+            pass
+
 
 class EscapeDecodeTest(unittest.TestCase):
     def test_empty(self):
diff --git a/Misc/NEWS.d/next/Library/2026-10-10-00-37-12.gh-issue-159071.a62BQP.rst b/Misc/NEWS.d/next/Library/2026-10-10-00-37-12.gh-issue-159071.a62BQP.rst
new file mode 100644
index 000000000000000..e665a6ce071bceb
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2026-10-10-00-37-12.gh-issue-159071.a62BQP.rst
@@ -0,0 +1,2 @@
+Fixed :exc:`UnicodeDecodeError` offsets in the :mod:`encodings.utf_8_sig`
+codec when input starts with a BOM.