Issue #16416: On Mac OS X, operating system data are now always
encoded/decoded to/from UTF-8/surrogateescape, instead of the locale encoding
(which may be ASCII if no locale environment variable is set), to avoid
inconsistencies with os.fsencode() and os.fsdecode() functions which are
already using UTF-8/surrogateescape.
diff --git a/Python/fileutils.c b/Python/fileutils.c
index c563eaa..cba6696 100644
--- a/Python/fileutils.c
+++ b/Python/fileutils.c
@@ -3,6 +3,10 @@
# include <windows.h>
#endif
+#ifdef __APPLE__
+extern wchar_t* _Py_DecodeUTF8_surrogateescape(const char *s, Py_ssize_t size);
+#endif
+
#ifdef HAVE_STAT
/* Decode a byte string from the locale encoding with the
@@ -23,6 +27,17 @@
wchar_t*
_Py_char2wchar(const char* arg, size_t *size)
{
+#ifdef __APPLE__
+ wchar_t *wstr;
+ wstr = _Py_DecodeUTF8_surrogateescape(arg, strlen(arg));
+ if (size != NULL) {
+ if (wstr != NULL)
+ *size = wcslen(wstr);
+ else
+ *size = (size_t)-1;
+ }
+ return wstr;
+#else
wchar_t *res;
#ifdef HAVE_BROKEN_MBSTOWCS
/* Some platforms have a broken implementation of
@@ -107,7 +122,7 @@
argsize -= converted;
out++;
}
-#else
+#else /* HAVE_MBRTOWC */
/* Cannot use C locale for escaping; manually escape as if charset
is ASCII (i.e. escape all bytes > 128. This will still roundtrip
correctly in the locale's charset, which must be an ASCII superset. */
@@ -121,13 +136,14 @@
else
*out++ = 0xdc00 + *in++;
*out = 0;
-#endif
+#endif /* HAVE_MBRTOWC */
if (size != NULL)
*size = out - res;
return res;
oom:
fprintf(stderr, "out of memory\n");
return NULL;
+#endif /* __APPLE__ */
}
/* Encode a (wide) character string to the locale encoding with the
@@ -144,14 +160,42 @@
char*
_Py_wchar2char(const wchar_t *text, size_t *error_pos)
{
+#ifdef __APPLE__
+ Py_ssize_t len;
+ PyObject *unicode, *bytes = NULL;
+ char *cpath;
+
+ unicode = PyUnicode_FromWideChar(text, wcslen(text));
+ if (unicode == NULL)
+ return NULL;
+
+ bytes = _PyUnicode_AsUTF8String(unicode, "surrogateescape");
+ Py_DECREF(unicode);
+ if (bytes == NULL) {
+ PyErr_Clear();
+ if (error_pos != NULL)
+ *error_pos = (size_t)-1;
+ return NULL;
+ }
+
+ len = PyBytes_GET_SIZE(bytes);
+ cpath = PyMem_Malloc(len+1);
+ if (cpath == NULL) {
+ PyErr_Clear();
+ Py_DECREF(bytes);
+ if (error_pos != NULL)
+ *error_pos = (size_t)-1;
+ return NULL;
+ }
+ memcpy(cpath, PyBytes_AsString(bytes), len + 1);
+ Py_DECREF(bytes);
+ return cpath;
+#else /* __APPLE__ */
const size_t len = wcslen(text);
char *result = NULL, *bytes = NULL;
size_t i, size, converted;
wchar_t c, buf[2];
- if (error_pos != NULL)
- *error_pos = (size_t)-1;
-
/* The function works in two steps:
1. compute the length of the output buffer in bytes (size)
2. outputs the bytes */
@@ -198,11 +242,15 @@
size += 1; /* nul byte at the end */
result = PyMem_Malloc(size);
- if (result == NULL)
+ if (result == NULL) {
+ if (error_pos != NULL)
+ *error_pos = (size_t)-1;
return NULL;
+ }
bytes = result;
}
return result;
+#endif /* __APPLE__ */
}
/* In principle, this should use HAVE__WSTAT, and _wstat