https://github.com/python/cpython/commit/16f4deefabe500dbd9e8621409d44073ba5cf059
commit: 16f4deefabe500dbd9e8621409d44073ba5cf059
branch: 3.13
author: Serhiy Storchaka <[email protected]>
committer: serhiy-storchaka <[email protected]>
date: 2026-09-01T07:53:03Z
summary:
[3.13] gh-99064: Ignore the encoding declaration when parsing decoded text
(GH-156734) (GH-156752)
ElementTree.parse() with a text file mis-decoded the text in the C
implementation: _parse_whole() encoded it as UTF-8, but left expat to honor
the encoding declared in the document. It now overrides the encoding, as
XMLParser.feed() already does for str data.
(cherry picked from commit c83013c92dfdc77b87a523b736b76d5abb8ede2a)
files:
A Misc/NEWS.d/next/Library/2026-08-31-20-30-00.gh-issue-99064.Rt4mZ9.rst
M Lib/test/test_xml_etree.py
M Modules/_elementtree.c
diff --git a/Lib/test/test_xml_etree.py b/Lib/test/test_xml_etree.py
index 4a76a5be1e5ede..c2642b2255bf49 100644
--- a/Lib/test/test_xml_etree.py
+++ b/Lib/test/test_xml_etree.py
@@ -1025,6 +1025,34 @@ def bxml(encoding, body=''):
self.assertRaises(ValueError, ET.XML, xml('undefined').encode('ascii'))
self.assertRaises(LookupError, ET.XML, xml('xxx').encode('ascii'))
+ def test_parse_text_source(self):
+ # gh-99064: The encoding declared in the document does not apply
+ # to a source which is already decoded.
+ def check(encoding, body):
+ xml = (f"<?xml version='1.0' encoding='{encoding}'?>"
+ f"<xml>{body}</xml>")
+ with self.subTest(encoding=encoding):
+ self.assertEqual(ET.parse(io.StringIO(xml)).getroot().text,
+ body)
+ # the same with an explicitly created parser
+ self.assertEqual(
+ ET.parse(io.StringIO(xml), ET.XMLParser()).getroot().text,
+ body)
+ check("ascii", 'a')
+ check("iso-8859-1", '\xbd')
+ check("iso-8859-15", '\u20ac')
+ check("cp437", '\u221a')
+ check("utf-8", '\u4e2d')
+ # not ASCII compatible, unsupported for a bytes source
+ check("utf-16", '\u4e2d')
+ check("utf-32", '\u4e2d')
+
+ def test_parse_text_source_multiple_chunks(self):
+ # the encoding is overridden before the first chunk is parsed
+ body = '\xe4' * 100_000
+ xml = "<?xml version='1.0' encoding='ISO-8859-1'?><xml>%s</xml>" % body
+ self.assertEqual(ET.parse(io.StringIO(xml)).getroot().text, body)
+
def test_methods(self):
# Test serialization methods.
diff --git
a/Misc/NEWS.d/next/Library/2026-08-31-20-30-00.gh-issue-99064.Rt4mZ9.rst
b/Misc/NEWS.d/next/Library/2026-08-31-20-30-00.gh-issue-99064.Rt4mZ9.rst
new file mode 100644
index 00000000000000..37a8c0b620310a
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2026-08-31-20-30-00.gh-issue-99064.Rt4mZ9.rst
@@ -0,0 +1,5 @@
+Fix :func:`xml.etree.ElementTree.parse` with a text file or other source
+of :class:`str` data in the C implementation.
+The encoding declared in the document was applied to the already decoded
+text, which produced mojibake. It is now ignored, as when parsing with
+:meth:`!XMLParser.feed` or :func:`~xml.etree.ElementTree.fromstring`.
diff --git a/Modules/_elementtree.c b/Modules/_elementtree.c
index f2e2f1237f11ef..f8fabdab505abb 100644
--- a/Modules/_elementtree.c
+++ b/Modules/_elementtree.c
@@ -4037,6 +4037,7 @@ _elementtree_XMLParser__parse_whole(XMLParserObject
*self, PyObject *file)
/* read from open file object */
elementtreestate *st = self->state;
+ int first = 1;
for (;;) {
buffer = PyObject_CallFunction(reader, "i", 64*1024);
@@ -4053,6 +4054,11 @@ _elementtree_XMLParser__parse_whole(XMLParserObject
*self, PyObject *file)
Py_DECREF(buffer);
break;
}
+ if (first) {
+ /* The text is already decoded, the encoding declared in the
+ document does not apply to it. Return code ignored. */
+ (void)EXPAT(st, SetEncoding)(self->parser, "utf-8");
+ }
temp = PyUnicode_AsEncodedString(buffer, "utf-8", "surrogatepass");
Py_DECREF(buffer);
if (!temp) {
@@ -4076,6 +4082,7 @@ _elementtree_XMLParser__parse_whole(XMLParserObject
*self, PyObject *file)
res = expat_parse(
st, self, PyBytes_AS_STRING(buffer), (int)PyBytes_GET_SIZE(buffer),
0);
+ first = 0;
Py_DECREF(buffer);
_______________________________________________
Python-checkins mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3//lists/python-checkins.python.org
Member address: [email protected]
