Issue #26919: On Android, operating system data is now always encoded/decoded

to/from UTF-8, instead of the locale encoding to avoid inconsistencies with
os.fsencode() and os.fsdecode() which are already using UTF-8.

Lib/test/test_cmd_line.py

@@ -8,7 +8,7 @@ import shutil
 import sys
 import subprocess
 import tempfile
-from test.support import script_helper
+from test.support import script_helper, is_android
 from test.support.script_helper import (spawn_python, kill_python, assert_python_ok,
     assert_python_failure)
 
@@ -178,15 +178,16 @@ class CmdLineTest(unittest.TestCase):
         if not stdout.startswith(pattern):
             raise AssertionError("%a doesn't start with %a" % (stdout, pattern))
 
-    @unittest.skipUnless(sys.platform == 'darwin', 'test specific to Mac OS X')
-    def test_osx_utf8(self):
+    @unittest.skipUnless((sys.platform == 'darwin' or
+                is_android), 'test specific to Mac OS X and Android')
+    def test_osx_android_utf8(self):
         def check_output(text):
             decoded = text.decode('utf-8', 'surrogateescape')
             expected = ascii(decoded).encode('ascii') + b'\n'
 
             env = os.environ.copy()
             # C locale gives ASCII locale encoding, but Python uses UTF-8
-            # to parse the command line arguments on Mac OS X
+            # to parse the command line arguments on Mac OS X and Android.
             env['LC_ALL'] = 'C'
 
             p = subprocess.Popen(

Misc/NEWS

@@ -10,6 +10,10 @@ What's New in Python 3.6.1 release candidate 1
 Core and Builtins
 -----------------
 
+- Issue #26919: On Android, operating system data is now always encoded/decoded
+  to/from UTF-8, instead of the locale encoding to avoid inconsistencies with
+  os.fsencode() and os.fsdecode() which are already using UTF-8.
+
 - Issue #28147: Fix a memory leak in split-table dictionaries: setattr()
   must not convert combined table into split table. Patch written by INADA
   Naoki.

Objects/unicodeobject.c

@@ -5083,10 +5083,10 @@ onError:
     return NULL;
 }
 
-#ifdef __APPLE__
+#if defined(__APPLE__) || defined(__ANDROID__)
 
 /* Simplified UTF-8 decoder using surrogateescape error handler,
-   used to decode the command line arguments on Mac OS X.
+   used to decode the command line arguments on Mac OS X and Android.
 
    Return a pointer to a newly allocated wide character string (use
    PyMem_RawFree() to free the memory), or NULL on memory allocation error. */
@@ -5137,7 +5137,7 @@ _Py_DecodeUTF8_surrogateescape(const char *s, Py_ssize_t size)
     return unicode;
 }
 
-#endif /* __APPLE__ */
+#endif /* __APPLE__ or __ANDROID__ */
 
 /* Primary internal function which creates utf8 encoded bytes objects.
 

Python/fileutils.c

@@ -20,7 +20,7 @@ extern int winerror_to_errno(int);
 #include <fcntl.h>
 #endif /* HAVE_FCNTL_H */
 
-#ifdef __APPLE__
+#if defined(__APPLE__) || defined(__ANDROID__)
 extern wchar_t* _Py_DecodeUTF8_surrogateescape(const char *s, Py_ssize_t size);
 #endif
 
@@ -273,7 +273,7 @@ decode_ascii_surrogateescape(const char *arg, size_t *size)
 wchar_t*
 Py_DecodeLocale(const char* arg, size_t *size)
 {
-#ifdef __APPLE__
+#if defined(__APPLE__) || defined(__ANDROID__)
     wchar_t *wstr;
     wstr = _Py_DecodeUTF8_surrogateescape(arg, strlen(arg));
     if (size != NULL) {
@@ -406,7 +406,7 @@ oom:
     if (size != NULL)
         *size = (size_t)-1;
     return NULL;
-#endif   /* __APPLE__ */
+#endif   /* __APPLE__ or __ANDROID__ */
 }
 
 /* Encode a wide character string to the locale encoding with the
@@ -424,7 +424,7 @@ oom:
 char*
 Py_EncodeLocale(const wchar_t *text, size_t *error_pos)
 {
-#ifdef __APPLE__
+#if defined(__APPLE__) || defined(__ANDROID__)
     Py_ssize_t len;
     PyObject *unicode, *bytes = NULL;
     char *cpath;
@@ -522,7 +522,7 @@ Py_EncodeLocale(const wchar_t *text, size_t *error_pos)
         bytes = result;
     }
     return result;
-#endif   /* __APPLE__ */
+#endif   /* __APPLE__ or __ANDROID__ */
 }