bpo-36346: Make using the legacy Unicode C API optional (GH-21437)

Add compile time option USE_UNICODE_WCHAR_CACHE. Setting it to 0
makes the interpreter not using the wchar_t cache and the legacy Unicode C API.

Author: serhiy-storchaka

Include/cpython/unicodeobject.h

@@ -11,7 +11,9 @@
 
 /* --- Internal Unicode Operations ---------------------------------------- */
 
-#define USE_UNICODE_WCHAR_CACHE 1
+#ifndef USE_UNICODE_WCHAR_CACHE
+#  define USE_UNICODE_WCHAR_CACHE 1
+#endif /* USE_UNICODE_WCHAR_CACHE */
 
 /* Since splitting on whitespace is an important use case, and
    whitespace in most situations is solely ASCII whitespace, we

Lib/test/support/__init__.py

@@ -36,6 +36,11 @@
 from .testresult import get_test_runner
 
 
+try:
+    from _testcapi import unicode_legacy_string
+except ImportError:
+    unicode_legacy_string = None
+
 __all__ = [
     # globals
     "PIPE_MAX_SIZE", "verbose", "max_memuse", "use_resources", "failfast",
@@ -426,6 +431,9 @@ def requires_lzma(reason='requires lzma'):
         lzma = None
     return unittest.skipUnless(lzma, reason)
 
+requires_legacy_unicode_capi = unittest.skipUnless(unicode_legacy_string,
+                        'requires legacy Unicode C API')
+
 is_jython = sys.platform.startswith('java')
 
 is_android = hasattr(sys, 'getandroidapilevel')

Lib/test/test_csv.py

@@ -250,9 +250,9 @@ def test_writerows_errors(self):
             self.assertRaises(OSError, writer.writerows, BadIterable())
 
     @support.cpython_only
+    @support.requires_legacy_unicode_capi
     def test_writerows_legacy_strings(self):
         import _testcapi
-
         c = _testcapi.unicode_legacy_string('a')
         with TemporaryFile("w+", newline='') as fileobj:
             writer = csv.writer(fileobj)

Lib/test/test_decimal.py

@@ -33,7 +33,8 @@
 import numbers
 import locale
 from test.support import (run_unittest, run_doctest, is_resource_enabled,
-                          requires_IEEE_754, requires_docstrings)
+                          requires_IEEE_754, requires_docstrings,
+                          requires_legacy_unicode_capi)
 from test.support import (TestFailed,
                           run_with_locale, cpython_only)
 from test.support.import_helper import import_fresh_module
@@ -582,6 +583,7 @@ def test_explicit_from_string(self):
             self.assertRaises(InvalidOperation, Decimal, "1_2_\u00003")
 
     @cpython_only
+    @requires_legacy_unicode_capi
     def test_from_legacy_strings(self):
         import _testcapi
         Decimal = self.decimal.Decimal
@@ -2817,6 +2819,7 @@ def test_none_args(self):
                                               Overflow])
 
     @cpython_only
+    @requires_legacy_unicode_capi
     def test_from_legacy_strings(self):
         import _testcapi
         c = self.decimal.Context()

Lib/test/test_getargs2.py

@@ -976,6 +976,7 @@ def test_et_hash(self):
         buf = bytearray()
         self.assertRaises(ValueError, getargs_et_hash, 'abc\xe9', 'latin1', buf)
 
+    @support.requires_legacy_unicode_capi
     def test_u(self):
         from _testcapi import getargs_u
         self.assertEqual(getargs_u('abc\xe9'), 'abc\xe9')
@@ -985,6 +986,7 @@ def test_u(self):
         self.assertRaises(TypeError, getargs_u, memoryview(b'memoryview'))
         self.assertRaises(TypeError, getargs_u, None)
 
+    @support.requires_legacy_unicode_capi
     def test_u_hash(self):
         from _testcapi import getargs_u_hash
         self.assertEqual(getargs_u_hash('abc\xe9'), 'abc\xe9')
@@ -994,6 +996,7 @@ def test_u_hash(self):
         self.assertRaises(TypeError, getargs_u_hash, memoryview(b'memoryview'))
         self.assertRaises(TypeError, getargs_u_hash, None)
 
+    @support.requires_legacy_unicode_capi
     def test_Z(self):
         from _testcapi import getargs_Z
         self.assertEqual(getargs_Z('abc\xe9'), 'abc\xe9')
@@ -1003,6 +1006,7 @@ def test_Z(self):
         self.assertRaises(TypeError, getargs_Z, memoryview(b'memoryview'))
         self.assertIsNone(getargs_Z(None))
 
+    @support.requires_legacy_unicode_capi
     def test_Z_hash(self):
         from _testcapi import getargs_Z_hash
         self.assertEqual(getargs_Z_hash('abc\xe9'), 'abc\xe9')

Lib/test/test_unicode.py

@@ -723,6 +723,7 @@ def test_isidentifier(self):
         self.assertFalse("0".isidentifier())
 
     @support.cpython_only
+    @support.requires_legacy_unicode_capi
     def test_isidentifier_legacy(self):
         import _testcapi
         u = '𝖀𝖓𝖎𝖈𝖔𝖉𝖊'
@@ -2350,6 +2351,7 @@ def test_getnewargs(self):
         self.assertEqual(len(args), 1)
 
     @support.cpython_only
+    @support.requires_legacy_unicode_capi
     def test_resize(self):
         from _testcapi import getargs_u
         for length in range(1, 100, 7):
@@ -2920,6 +2922,7 @@ def test_copycharacters(self):
         self.assertRaises(SystemError, unicode_copycharacters, s, 0, b'', 0, 0)
 
     @support.cpython_only
+    @support.requires_legacy_unicode_capi
     def test_encode_decimal(self):
         from _testcapi import unicode_encodedecimal
         self.assertEqual(unicode_encodedecimal('123'),
@@ -2936,6 +2939,7 @@ def test_encode_decimal(self):
             unicode_encodedecimal, "123\u20ac", "replace")
 
     @support.cpython_only
+    @support.requires_legacy_unicode_capi
     def test_transform_decimal(self):
         from _testcapi import unicode_transformdecimaltoascii as transform_decimal
         self.assertEqual(transform_decimal('123'),

Modules/_io/fileio.c

@@ -270,7 +270,14 @@ _io_FileIO___init___impl(fileio *self, PyObject *nameobj, const char *mode,
         if (!PyUnicode_FSDecoder(nameobj, &stringobj)) {
             return -1;
         }
+#if USE_UNICODE_WCHAR_CACHE
+_Py_COMP_DIAG_PUSH
+_Py_COMP_DIAG_IGNORE_DEPR_DECLS
         widename = PyUnicode_AsUnicode(stringobj);
+_Py_COMP_DIAG_POP
+#else /* USE_UNICODE_WCHAR_CACHE */
+        widename = PyUnicode_AsWideCharString(stringobj, NULL);
+#endif /* USE_UNICODE_WCHAR_CACHE */
         if (widename == NULL)
             return -1;
 #else
@@ -491,6 +498,11 @@ _io_FileIO___init___impl(fileio *self, PyObject *nameobj, const char *mode,
         internal_close(self);
 
  done:
+#ifdef MS_WINDOWS
+#if !USE_UNICODE_WCHAR_CACHE
+    PyMem_Free(widename);
+#endif /* USE_UNICODE_WCHAR_CACHE */
+#endif
     Py_CLEAR(stringobj);
     return ret;
 }

Modules/_testcapimodule.c

@@ -1668,6 +1668,7 @@ parse_tuple_and_keywords(PyObject *self, PyObject *args)
 
 static volatile int x;
 
+#if USE_UNICODE_WCHAR_CACHE
 /* Ignore use of deprecated APIs */
 _Py_COMP_DIAG_PUSH
 _Py_COMP_DIAG_IGNORE_DEPR_DECLS
@@ -1772,6 +1773,8 @@ test_Z_code(PyObject *self, PyObject *Py_UNUSED(ignored))
     Py_DECREF(tuple);
     Py_RETURN_NONE;
 }
+_Py_COMP_DIAG_POP
+#endif /* USE_UNICODE_WCHAR_CACHE */
 
 static PyObject *
 test_widechar(PyObject *self, PyObject *Py_UNUSED(ignored))
@@ -1824,6 +1827,10 @@ test_widechar(PyObject *self, PyObject *Py_UNUSED(ignored))
         return raiseTestError("test_widechar",
                               "PyUnicode_FromWideChar(L\"\\U00110000\", 1) didn't fail");
 
+#if USE_UNICODE_WCHAR_CACHE
+/* Ignore use of deprecated APIs */
+_Py_COMP_DIAG_PUSH
+_Py_COMP_DIAG_IGNORE_DEPR_DECLS
     wide = PyUnicode_FromUnicode(invalid, 1);
     if (wide == NULL)
         PyErr_Clear();
@@ -1844,11 +1851,12 @@ test_widechar(PyObject *self, PyObject *Py_UNUSED(ignored))
         return raiseTestError("test_widechar",
                               "PyUnicode_Ready() didn't fail");
     }
+_Py_COMP_DIAG_POP
+#endif /* USE_UNICODE_WCHAR_CACHE */
 #endif
 
     Py_RETURN_NONE;
 }
-_Py_COMP_DIAG_POP
 
 static PyObject *
 unicode_aswidechar(PyObject *self, PyObject *args)
@@ -2024,6 +2032,7 @@ unicode_copycharacters(PyObject *self, PyObject *args)
     return Py_BuildValue("(Nn)", to_copy, copied);
 }
 
+#if USE_UNICODE_WCHAR_CACHE
 /* Ignore use of deprecated APIs */
 _Py_COMP_DIAG_PUSH
 _Py_COMP_DIAG_IGNORE_DEPR_DECLS
@@ -2096,6 +2105,7 @@ unicode_legacy_string(PyObject *self, PyObject *args)
     return u;
 }
 _Py_COMP_DIAG_POP
+#endif /* USE_UNICODE_WCHAR_CACHE */
 
 static PyObject *
 getargs_w_star(PyObject *self, PyObject *args)
@@ -5398,8 +5408,10 @@ static PyMethodDef TestMethods[] = {
     {"codec_incrementaldecoder",
      (PyCFunction)codec_incrementaldecoder,                      METH_VARARGS},
     {"test_s_code",             test_s_code,                     METH_NOARGS},
+#if USE_UNICODE_WCHAR_CACHE
     {"test_u_code",             test_u_code,                     METH_NOARGS},
     {"test_Z_code",             test_Z_code,                     METH_NOARGS},
+#endif /* USE_UNICODE_WCHAR_CACHE */
     {"test_widechar",           test_widechar,                   METH_NOARGS},
     {"unicode_aswidechar",      unicode_aswidechar,              METH_VARARGS},
     {"unicode_aswidecharstring",unicode_aswidecharstring,        METH_VARARGS},
@@ -5408,9 +5420,11 @@ static PyMethodDef TestMethods[] = {
     {"unicode_asutf8andsize",   unicode_asutf8andsize,           METH_VARARGS},
     {"unicode_findchar",        unicode_findchar,                METH_VARARGS},
     {"unicode_copycharacters",  unicode_copycharacters,          METH_VARARGS},
+#if USE_UNICODE_WCHAR_CACHE
     {"unicode_encodedecimal",   unicode_encodedecimal,           METH_VARARGS},
     {"unicode_transformdecimaltoascii", unicode_transformdecimaltoascii, METH_VARARGS},
     {"unicode_legacy_string",   unicode_legacy_string,           METH_VARARGS},
+#endif /* USE_UNICODE_WCHAR_CACHE */
     {"_test_thread_state",      test_thread_state,               METH_VARARGS},
     {"_pending_threadfunc",     pending_threadfunc,              METH_VARARGS},
 #ifdef HAVE_GETTIMEOFDAY

Modules/_winapi.c

@@ -164,10 +164,11 @@ create_converter('LPCVOID', '" F_POINTER "')
 create_converter('BOOL', 'i') # F_BOOL used previously (always 'i')
 create_converter('DWORD', 'k') # F_DWORD is always "k" (which is much shorter)
 create_converter('LPCTSTR', 's')
-create_converter('LPCWSTR', 'u')
-create_converter('LPWSTR', 'u')
 create_converter('UINT', 'I') # F_UINT used previously (always 'I')
 
+class LPCWSTR_converter(Py_UNICODE_converter):
+    type = 'LPCWSTR'
+
 class HANDLE_return_converter(CReturnConverter):
     type = 'HANDLE'
 
@@ -197,7 +198,7 @@ class LPVOID_return_converter(CReturnConverter):
         data.return_conversion.append(
             'return_value = HANDLE_TO_PYNUM(_return_value);\n')
 [python start generated code]*/
-/*[python end generated code: output=da39a3ee5e6b4b0d input=79464c61a31ae932]*/
+/*[python end generated code: output=da39a3ee5e6b4b0d input=011ee0c3a2244bfe]*/
 
 #include "clinic/_winapi.c.h"
 
@@ -520,15 +521,15 @@ _winapi_CreateFileMapping_impl(PyObject *module, HANDLE file_handle,
 /*[clinic input]
 _winapi.CreateJunction
 
-    src_path: LPWSTR
-    dst_path: LPWSTR
+    src_path: LPCWSTR
+    dst_path: LPCWSTR
     /
 [clinic start generated code]*/
 
 static PyObject *
-_winapi_CreateJunction_impl(PyObject *module, LPWSTR src_path,
-                            LPWSTR dst_path)
-/*[clinic end generated code: output=66b7eb746e1dfa25 input=8cd1f9964b6e3d36]*/
+_winapi_CreateJunction_impl(PyObject *module, LPCWSTR src_path,
+                            LPCWSTR dst_path)
+/*[clinic end generated code: output=44b3f5e9bbcc4271 input=963d29b44b9384a7]*/
 {
     /* Privilege adjustment */
     HANDLE token = NULL;