gh-86427: Use the console code page for the stdio encoding on Windows (GH-155415)
In the legacy Windows stdio mode PyConfig.stdio_encoding is now initialized
with the console output code page instead of the ANSI code page.
Only one encoding can be specified, so sys.stdin.encoding is still wrong if
the input code page differs from the output one.
diff --git a/Lib/test/test_cmd_line.py b/Lib/test/test_cmd_line.py
index 555cebe..2200770 100644
--- a/Lib/test/test_cmd_line.py
+++ b/Lib/test/test_cmd_line.py
@@ -1067,6 +1067,41 @@ def test_python_legacy_windows_stdio(self):
support.skip_on_low_desktop_heap_memory_subprocess(p.returncode)
self.assertEqual(p.returncode, 0)
+ @unittest.skipUnless(support.MS_WINDOWS, 'Test only applicable on Windows')
+ def test_python_legacy_windows_stdio_encoding(self):
+ # gh-86427: In the legacy mode the encoding of the standard streams
+ # is the encoding of the console.
+ import ctypes
+ kernel32 = ctypes.WinDLL('kernel32', use_last_error=True)
+ try:
+ fin = open('CONIN$')
+ except OSError:
+ self.skipTest('no console')
+ # We cannot use PIPE, because the standard streams should be
+ # connected to the console. So we use the exit code.
+ code = ("import sys; sys.exit(sys.stdin.encoding != 'cp850' or "
+ "sys.stdout.encoding != 'cp850')")
+ env = os.environ.copy()
+ env['PYTHONLEGACYWINDOWSSTDIO'] = '1'
+ env['PYTHONUTF8'] = '0'
+ env.pop('PYTHONIOENCODING', None)
+ old_cp = kernel32.GetConsoleCP()
+ old_output_cp = kernel32.GetConsoleOutputCP()
+ with fin, open('CONOUT$', 'w') as fout:
+ try:
+ if not kernel32.SetConsoleCP(850):
+ self.skipTest('cannot set the console input code page')
+ if not kernel32.SetConsoleOutputCP(850):
+ self.skipTest('cannot set the console output code page')
+ proc = subprocess.run([sys.executable, '-c', code], env=env,
+ stdin=fin, stdout=fout,
+ stderr=subprocess.DEVNULL)
+ finally:
+ kernel32.SetConsoleCP(old_cp)
+ kernel32.SetConsoleOutputCP(old_output_cp)
+ support.skip_on_low_desktop_heap_memory_subprocess(proc.returncode)
+ self.assertEqual(proc.returncode, 0)
+
@unittest.skipIf("-fsanitize" in sysconfig.get_config_vars().get('PY_CFLAGS', ()),
"PYTHONMALLOCSTATS doesn't work with ASAN")
def test_python_malloc_stats(self):
diff --git a/Misc/NEWS.d/next/Windows/2026-08-09-07-00-00.gh-issue-86427.consolecp.rst b/Misc/NEWS.d/next/Windows/2026-08-09-07-00-00.gh-issue-86427.consolecp.rst
new file mode 100644
index 0000000..1cbf188
--- /dev/null
+++ b/Misc/NEWS.d/next/Windows/2026-08-09-07-00-00.gh-issue-86427.consolecp.rst
@@ -0,0 +1,3 @@
+Fix the encoding of the standard streams in the legacy Windows stdio mode
+(:envvar:`PYTHONLEGACYWINDOWSSTDIO`). It is now the code page of the
+console, as in Python 3.7, not the ANSI code page.
diff --git a/Python/initconfig.c b/Python/initconfig.c
index b9bacb1..69d47a5 100644
--- a/Python/initconfig.c
+++ b/Python/initconfig.c
@@ -2710,8 +2710,30 @@ config_init_stdio_encoding(PyConfig *config,
/* Choose the default error handler based on the current locale. */
if (config->stdio_encoding == NULL) {
- status = config_get_locale_encoding(config, preconfig,
- &config->stdio_encoding);
+#ifdef MS_WINDOWS
+ /* gh-86427: use the console code page. Only one encoding can be
+ specified, so the output code page is used: it affects two
+ streams of three. */
+ UINT cp = config->legacy_windows_stdio ? GetConsoleOutputCP() : 0;
+ if (cp != 0) {
+ if (cp == CP_UTF8) {
+ status = PyConfig_SetString(config, &config->stdio_encoding,
+ L"utf-8");
+ }
+ else {
+ wchar_t encoding[20];
+ swprintf(encoding, Py_ARRAY_LENGTH(encoding), L"cp%u",
+ (unsigned int)cp);
+ status = PyConfig_SetString(config, &config->stdio_encoding,
+ encoding);
+ }
+ }
+ else
+#endif
+ {
+ status = config_get_locale_encoding(config, preconfig,
+ &config->stdio_encoding);
+ }
if (_PyStatus_EXCEPTION(status)) {
return status;
}