Issue #16218, #16444: Backport improvment on tests for non-ASCII characters

4 files changed, 105 insertions, 10 deletions

diff --git a/Lib/test/support.py b/Lib/test/support.py
index c5640e0..d0a37ea 100644
--- a/Lib/test/support.py
+++ b/Lib/test/support.py
@@ -603,6 +603,49 @@ else:
 # module name.
 TESTFN = "{}_{}_tmp".format(TESTFN, os.getpid())
 
+# FS_NONASCII: non-ASCII character encodable by os.fsencode(),
+# or None if there is no such character.
+FS_NONASCII = None
+for character in (
+    # First try printable and common characters to have a readable filename.
+    # For each character, the encoding list are just example of encodings able
+    # to encode the character (the list is not exhaustive).
+
+    # U+00E6 (Latin Small Letter Ae): cp1252, iso-8859-1
+    '\u00E6',
+    # U+0130 (Latin Capital Letter I With Dot Above): cp1254, iso8859_3
+    '\u0130',
+    # U+0141 (Latin Capital Letter L With Stroke): cp1250, cp1257
+    '\u0141',
+    # U+03C6 (Greek Small Letter Phi): cp1253
+    '\u03C6',
+    # U+041A (Cyrillic Capital Letter Ka): cp1251
+    '\u041A',
+    # U+05D0 (Hebrew Letter Alef): Encodable to cp424
+    '\u05D0',
+    # U+060C (Arabic Comma): cp864, cp1006, iso8859_6, mac_arabic
+    '\u060C',
+    # U+062A (Arabic Letter Teh): cp720
+    '\u062A',
+    # U+0E01 (Thai Character Ko Kai): cp874
+    '\u0E01',
+
+    # Then try more "special" characters. "special" because they may be
+    # interpreted or displayed differently depending on the exact locale
+    # encoding and the font.
+
+    # U+00A0 (No-Break Space)
+    '\u00A0',
+    # U+20AC (Euro Sign)
+    '\u20AC',
+):
+    try:
+        os.fsdecode(os.fsencode(character))
+    except UnicodeError:
+        pass
+    else:
+        FS_NONASCII = character
+        break
 
 # TESTFN_UNICODE is a non-ascii filename
 TESTFN_UNICODE = TESTFN + "-\xe0\xf2\u0258\u0141\u011f"
@@ -647,6 +690,38 @@ elif sys.platform != 'darwin':
         # the byte 0xff. Skip some unicode filename tests.
         pass
 
+# TESTFN_UNDECODABLE is a filename (bytes type) that should *not* be able to be
+# decoded from the filesystem encoding (in strict mode). It can be None if we
+# cannot generate such filename (ex: the latin1 encoding can decode any byte
+# sequence). On UNIX, TESTFN_UNDECODABLE can be decoded by os.fsdecode() thanks
+# to the surrogateescape error handler (PEP 383), but not from the filesystem
+# encoding in strict mode.
+TESTFN_UNDECODABLE = None
+for name in (
+    # b'\xff' is not decodable by os.fsdecode() with code page 932. Windows
+    # accepts it to create a file or a directory, or don't accept to enter to
+    # such directory (when the bytes name is used). So test b'\xe7' first: it is
+    # not decodable from cp932.
+    b'\xe7w\xf0',
+    # undecodable from ASCII, UTF-8
+    b'\xff',
+    # undecodable from iso8859-3, iso8859-6, iso8859-7, cp424, iso8859-8, cp856
+    # and cp857
+    b'\xae\xd5'
+    # undecodable from UTF-8 (UNIX and Mac OS X)
+    b'\xed\xb2\x80', b'\xed\xb4\x80',
+):
+    try:
+        name.decode(TESTFN_ENCODING)
+    except UnicodeDecodeError:
+        TESTFN_UNDECODABLE = os.fsencode(TESTFN) + name
+        break
+
+if FS_NONASCII:
+    TESTFN_NONASCII = TESTFN + '-' + FS_NONASCII
+else:
+    TESTFN_NONASCII = None
+
 # Save the initial cwd
 SAVEDCWD = os.getcwd()
 
diff --git a/Lib/test/test_cmd_line.py b/Lib/test/test_cmd_line.py
index 2b0c6e2..f617c2f 100644
--- a/Lib/test/test_cmd_line.py
+++ b/Lib/test/test_cmd_line.py
@@ -93,15 +93,15 @@ class CmdLineTest(unittest.TestCase):
         # All good if execution is successful
         assert_python_ok('-c', 'pass')
 
-    @unittest.skipIf(sys.getfilesystemencoding() == 'ascii',
-                     'need a filesystem encoding different than ASCII')
+    @unittest.skipUnless(test.support.FS_NONASCII, 'need support.FS_NONASCII')
     def test_non_ascii(self):
         # Test handling of non-ascii data
         if test.support.verbose:
             import locale
             print('locale encoding = %s, filesystem encoding = %s'
                   % (locale.getpreferredencoding(), sys.getfilesystemencoding()))
-        command = "assert(ord('\xe9') == 0xe9)"
+        command = ("assert(ord(%r) == %s)"
+                   % (test.support.FS_NONASCII, ord(test.support.FS_NONASCII)))
         assert_python_ok('-c', command)
 
     # On Windows, pass bytes to subprocess doesn't test how Python decodes the
diff --git a/Lib/test/test_cmd_line_script.py b/Lib/test/test_cmd_line_script.py
index 6e097e3..f066204 100644
--- a/Lib/test/test_cmd_line_script.py
+++ b/Lib/test/test_cmd_line_script.py
@@ -363,14 +363,30 @@ class CmdLineTest(unittest.TestCase):
             self.assertTrue(text[1].startswith('  File '))
             self.assertTrue(text[3].startswith('NameError'))
 
-    def test_non_utf8(self):
+    def test_non_ascii(self):
+        # Mac OS X denies the creation of a file with an invalid UTF-8 name.
+        # Windows allows to create a name with an arbitrary bytes name, but
+        # Python cannot a undecodable bytes argument to a subprocess.
+        #if (support.TESTFN_UNDECODABLE
+        #and sys.platform not in ('win32', 'darwin')):
+        #    name = os.fsdecode(support.TESTFN_UNDECODABLE)
+        #elif support.TESTFN_NONASCII:
+        if support.TESTFN_NONASCII:
+            name = support.TESTFN_NONASCII
+        else:
+            self.skipTest("need support.TESTFN_NONASCII")
+
         # Issue #16218
-        with temp_dir() as script_dir:
-            script_name = _make_test_script(script_dir,
-                    '\udcf1\udcea\udcf0\udce8\udcef\udcf2')
-            self._check_script(script_name, script_name, script_name,
-                               script_dir, None,
-                               importlib.machinery.SourceFileLoader)
+        source = 'print(ascii(__file__))\n'
+        script_name = _make_test_script(os.curdir, name, source)
+        self.addCleanup(support.unlink, script_name)
+        rc, stdout, stderr = assert_python_ok(script_name)
+        self.assertEqual(
+            ascii(script_name),
+            stdout.rstrip().decode('ascii'),
+            'stdout=%r stderr=%r' % (stdout, stderr))
+        self.assertEqual(0, rc)
+
 
 def test_main():
     support.run_unittest(CmdLineTest)
diff --git a/Lib/test/test_os.py b/Lib/test/test_os.py
index 7d6b377..13a2b38 100644
--- a/Lib/test/test_os.py
+++ b/Lib/test/test_os.py
@@ -1243,6 +1243,8 @@ if sys.platform != 'win32':
         def setUp(self):
             if support.TESTFN_UNENCODABLE:
                 self.dir = support.TESTFN_UNENCODABLE
+            elif support.TESTFN_NONASCII:
+                self.dir = support.TESTFN_NONASCII
             else:
                 self.dir = support.TESTFN
             self.bdir = os.fsencode(self.dir)
@@ -1257,6 +1259,8 @@ if sys.platform != 'win32':
             add_filename(support.TESTFN_UNICODE)
             if support.TESTFN_UNENCODABLE:
                 add_filename(support.TESTFN_UNENCODABLE)
+            if support.TESTFN_NONASCII:
+                add_filename(support.TESTFN_NONASCII)
             if not bytesfn:
                 self.skipTest("couldn't create any non-ascii filename")