- Fix for lone surrogates, utf8 and certain encode error handlers.

This commit is contained in:
Toshio Kuratomi 2010-08-20 15:07:00 -04:00
parent db6a64a5ec
commit a9aa92da40
2 changed files with 186 additions and 1 deletions

View File

@ -0,0 +1,175 @@
Index: Objects/unicodeobject.c
===================================================================
--- Objects/unicodeobject.c (revision 80382)
+++ Objects/unicodeobject.c (revision 80383)
@@ -159,6 +159,12 @@
const Py_UNICODE *unicode, Py_ssize_t size, PyObject **exceptionObject,
Py_ssize_t startpos, Py_ssize_t endpos, Py_ssize_t *newpos);
+static void raise_encode_exception(PyObject **exceptionObject,
+ const char *encoding,
+ const Py_UNICODE *unicode, Py_ssize_t size,
+ Py_ssize_t startpos, Py_ssize_t endpos,
+ const char *reason);
+
/* Same for linebreaks */
static unsigned char ascii_linebreak[] = {
0, 0, 0, 0, 0, 0, 0, 0,
@@ -2461,61 +2467,88 @@
/* Encode Latin-1 */
*p++ = (char)(0xc0 | (ch >> 6));
*p++ = (char)(0x80 | (ch & 0x3f));
- }
- else {
- /* Encode UCS2 Unicode ordinals */
- if (ch < 0x10000) {
+ } else if (0xD800 <= ch && ch <= 0xDFFF) {
#ifndef Py_UNICODE_WIDE
- /* Special case: check for high surrogate */
- if (0xD800 <= ch && ch <= 0xDBFF && i != size) {
- Py_UCS4 ch2 = s[i];
- /* Check for low surrogate and combine the two to
- form a UCS4 value */
- if (0xDC00 <= ch2 && ch2 <= 0xDFFF) {
- ch = ((ch - 0xD800) << 10 | (ch2 - 0xDC00)) + 0x10000;
- i++;
- goto encodeUCS4;
- }
- /* Fall through: handles isolated high surrogates */
- }
+ /* Special case: check for high and low surrogate */
+ if (ch <= 0xDBFF && i != size && 0xDC00 <= s[i] && s[i] <= 0xDFFF) {
+ Py_UCS4 ch2 = s[i];
+ /* Combine the two surrogates to form a UCS4 value */
+ ch = ((ch - 0xD800) << 10 | (ch2 - 0xDC00)) + 0x10000;
+ i++;
+
+ /* Encode UCS4 Unicode ordinals */
+ *p++ = (char)(0xf0 | (ch >> 18));
+ *p++ = (char)(0x80 | ((ch >> 12) & 0x3f));
+ *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
+ *p++ = (char)(0x80 | (ch & 0x3f));
+
#endif
- if (ch >= 0xd800 && ch <= 0xdfff) {
- Py_ssize_t newpos;
- PyObject *rep;
- char *prep;
- int k;
- rep = unicode_encode_call_errorhandler
- (errors, &errorHandler, "utf-8", "surrogates not allowed",
- s, size, &exc, i-1, i, &newpos);
- if (!rep)
+ } else {
+ Py_ssize_t newpos;
+ PyObject *rep;
+ Py_ssize_t repsize, k;
+ rep = unicode_encode_call_errorhandler
+ (errors, &errorHandler, "utf-8", "surrogates not allowed",
+ s, size, &exc, i-1, i, &newpos);
+ if (!rep)
+ goto error;
+
+ if (PyBytes_Check(rep))
+ repsize = PyBytes_GET_SIZE(rep);
+ else
+ repsize = PyUnicode_GET_SIZE(rep);
+
+ if (repsize > 4) {
+ Py_ssize_t offset;
+
+ if (result == NULL)
+ offset = p - stackbuf;
+ else
+ offset = p - PyBytes_AS_STRING(result);
+
+ if (nallocated > PY_SSIZE_T_MAX - repsize + 4) {
+ /* integer overflow */
+ PyErr_NoMemory();
goto error;
- /* Implementation limitations: only support error handler that return
- bytes, and only support up to four replacement bytes. */
- if (!PyBytes_Check(rep)) {
- PyErr_SetString(PyExc_TypeError, "error handler should have returned bytes");
- Py_DECREF(rep);
- goto error;
}
- if (PyBytes_Size(rep) > 4) {
- PyErr_SetString(PyExc_TypeError, "error handler returned too many bytes");
- Py_DECREF(rep);
- goto error;
+ nallocated += repsize - 4;
+ if (result != NULL) {
+ if (_PyBytes_Resize(&result, nallocated) < 0)
+ goto error;
+ } else {
+ result = PyBytes_FromStringAndSize(NULL, nallocated);
+ if (result == NULL)
+ goto error;
+ Py_MEMCPY(PyBytes_AS_STRING(result), stackbuf, offset);
}
- prep = PyBytes_AsString(rep);
- for(k = PyBytes_Size(rep); k > 0; k--)
+ p = PyBytes_AS_STRING(result) + offset;
+ }
+
+ if (PyBytes_Check(rep)) {
+ char *prep = PyBytes_AS_STRING(rep);
+ for(k = repsize; k > 0; k--)
*p++ = *prep++;
- Py_DECREF(rep);
- continue;
-
+ } else /* rep is unicode */ {
+ Py_UNICODE *prep = PyUnicode_AS_UNICODE(rep);
+ Py_UNICODE c;
+
+ for(k=0; k<repsize; k++) {
+ c = prep[k];
+ if (0x80 <= c) {
+ raise_encode_exception(&exc, "utf-8", s, size,
+ i-1, i, "surrogates not allowed");
+ goto error;
+ }
+ *p++ = (char)prep[k];
+ }
}
- *p++ = (char)(0xe0 | (ch >> 12));
- *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
- *p++ = (char)(0x80 | (ch & 0x3f));
- continue;
+ Py_DECREF(rep);
}
-#ifndef Py_UNICODE_WIDE
- encodeUCS4:
-#endif
+ } else if (ch < 0x10000) {
+ *p++ = (char)(0xe0 | (ch >> 12));
+ *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
+ *p++ = (char)(0x80 | (ch & 0x3f));
+ } else /* ch >= 0x10000 */ {
/* Encode UCS4 Unicode ordinals */
*p++ = (char)(0xf0 | (ch >> 18));
*p++ = (char)(0x80 | ((ch >> 12) & 0x3f));
Index: Lib/test/test_codecs.py
===================================================================
--- Lib/test/test_codecs.py (revision 80382)
+++ Lib/test/test_codecs.py (revision 80383)
@@ -571,6 +571,16 @@
def test_lone_surrogates(self):
self.assertRaises(UnicodeEncodeError, "\ud800".encode, "utf-8")
self.assertRaises(UnicodeDecodeError, b"\xed\xa0\x80".decode, "utf-8")
+ self.assertEqual("[\uDC80]".encode("utf-8", "backslashreplace"),
+ b'[\\udc80]')
+ self.assertEqual("[\uDC80]".encode("utf-8", "xmlcharrefreplace"),
+ b'[&#56448;]')
+ self.assertEqual("[\uDC80]".encode("utf-8", "surrogateescape"),
+ b'[\x80]')
+ self.assertEqual("[\uDC80]".encode("utf-8", "ignore"),
+ b'[]')
+ self.assertEqual("[\uDC80]".encode("utf-8", "replace"),
+ b'[?]')
def test_surrogatepass_handler(self):
self.assertEquals("abc\ud800def".encode("utf-8", "surrogatepass"),

View File

@ -40,7 +40,7 @@
Summary: Version 3 of the Python programming language aka Python 3000
Name: python3
Version: %{pybasever}.2
Release: 12%{?dist}
Release: 13%{?dist}
License: Python
Group: Development/Languages
Source: http://python.org/ftp/python/%{version}/Python-%{version}.tar.bz2
@ -224,6 +224,11 @@ Patch109: python-3.1.2-CVE-2008-5983.patch
# Sent upstream as http://bugs.python.org/issue9054
Patch110: python-3.1.2-fix-expat-issue9054.patch
# Fix encoding to utf8 when lone surrogates are present and error handler is
# set to ignore, replace, or others that return a unicode str.
# http://bugs.python.org/issue8092
Patch111: python3-r80382-lone-surrogate-and-utf8-error-handler.patch
BuildRoot: %{_tmppath}/%{name}-%{version}-root
BuildRequires: readline-devel, openssl-devel, gmp-devel
BuildRequires: ncurses-devel, gdbm-devel, zlib-devel, expat-devel
@ -384,6 +389,8 @@ rm -r Modules/zlib || exit 1
%patch110 -p0 -b .fix-expat-issue9054
%patch111 -p0 -b .surrogate-utf8
# Currently (2010-01-15), http://docs.python.org/library is for 2.6, and there
# are many differences between 2.6 and the Python 3 library.
#
@ -1064,6 +1071,9 @@ rm -fr %{buildroot}
%changelog
* Fri Aug 20 2010 Toshio Kuratomi <toshio@fedoraproject.org> - 3.1.2-13
- Fix for lone surrogates, utf8 and certain encode error handlers.
* Fri Jul 2 2010 David Malcolm <dmalcolm@redhat.com> - 3.1.2-12
- rebuild