[3.12] gh-52551: Fix encoding issues in strftime() (GH-125193) (GH-125657) (GH-125661)

Fix time.strftime(), the strftime() method and formatting of the
datetime classes datetime, date and time.

* Characters not encodable in the current locale are now acceptable in
  the format string.
* Surrogate pairs and sequence of surrogatescape-encoded bytes are no
  longer recombinated.
* Embedded null character no longer terminates the format string.

This fixes also gh-78662 and gh-124531.

(cherry picked from commit 08ccbb9)
(cherry picked from commit ad3eac1)

Co-authored-by: Serhiy Storchaka <[email protected]>

Author: miss-islington

Lib/test/datetimetester.py

@@ -2816,11 +2816,32 @@ def test_more_strftime(self):
             self.assertEqual(t.strftime("%z"), "-0200" + z)
             self.assertEqual(t.strftime("%:z"), "-02:00:" + z)
 
-        # bpo-34482: Check that surrogates don't cause a crash.
-        try:
-            t.strftime('%y\ud800%m %H\ud800%M')
-        except UnicodeEncodeError:
-            pass
+    def test_strftime_special(self):
+        t = self.theclass(2004, 12, 31, 6, 22, 33, 47)
+        s1 = t.strftime('%c')
+        s2 = t.strftime('%B')
+        # gh-52551, gh-78662: Unicode strings should pass through strftime,
+        # independently from locale.
+        self.assertEqual(t.strftime('\U0001f40d'), '\U0001f40d')
+        self.assertEqual(t.strftime('\U0001f4bb%c\U0001f40d%B'), f'\U0001f4bb{s1}\U0001f40d{s2}')
+        self.assertEqual(t.strftime('%c\U0001f4bb%B\U0001f40d'), f'{s1}\U0001f4bb{s2}\U0001f40d')
+        # Lone surrogates should pass through.
+        self.assertEqual(t.strftime('\ud83d'), '\ud83d')
+        self.assertEqual(t.strftime('\udc0d'), '\udc0d')
+        self.assertEqual(t.strftime('\ud83d%c\udc0d%B'), f'\ud83d{s1}\udc0d{s2}')
+        self.assertEqual(t.strftime('%c\ud83d%B\udc0d'), f'{s1}\ud83d{s2}\udc0d')
+        self.assertEqual(t.strftime('%c\udc0d%B\ud83d'), f'{s1}\udc0d{s2}\ud83d')
+        # Surrogate pairs should not recombine.
+        self.assertEqual(t.strftime('\ud83d\udc0d'), '\ud83d\udc0d')
+        self.assertEqual(t.strftime('%c\ud83d\udc0d%B'), f'{s1}\ud83d\udc0d{s2}')
+        # Surrogate-escaped bytes should not recombine.
+        self.assertEqual(t.strftime('\udcf0\udc9f\udc90\udc8d'), '\udcf0\udc9f\udc90\udc8d')
+        self.assertEqual(t.strftime('%c\udcf0\udc9f\udc90\udc8d%B'), f'{s1}\udcf0\udc9f\udc90\udc8d{s2}')
+        # gh-124531: The null character should not terminate the format string.
+        self.assertEqual(t.strftime('\0'), '\0')
+        self.assertEqual(t.strftime('\0'*1000), '\0'*1000)
+        self.assertEqual(t.strftime('\0%c\0%B'), f'\0{s1}\0{s2}')
+        self.assertEqual(t.strftime('%c\0%B\0'), f'{s1}\0{s2}\0')
 
     def test_extract(self):
         dt = self.theclass(2002, 3, 4, 18, 45, 3, 1234)
@@ -3573,6 +3594,33 @@ def test_strftime(self):
         # gh-85432: The parameter was named "fmt" in the pure-Python impl.
         t.strftime(format="%f")
 
+    def test_strftime_special(self):
+        t = self.theclass(1, 2, 3, 4)
+        s1 = t.strftime('%I%p%Z')
+        s2 = t.strftime('%X')
+        # gh-52551, gh-78662: Unicode strings should pass through strftime,
+        # independently from locale.
+        self.assertEqual(t.strftime('\U0001f40d'), '\U0001f40d')
+        self.assertEqual(t.strftime('\U0001f4bb%I%p%Z\U0001f40d%X'), f'\U0001f4bb{s1}\U0001f40d{s2}')
+        self.assertEqual(t.strftime('%I%p%Z\U0001f4bb%X\U0001f40d'), f'{s1}\U0001f4bb{s2}\U0001f40d')
+        # Lone surrogates should pass through.
+        self.assertEqual(t.strftime('\ud83d'), '\ud83d')
+        self.assertEqual(t.strftime('\udc0d'), '\udc0d')
+        self.assertEqual(t.strftime('\ud83d%I%p%Z\udc0d%X'), f'\ud83d{s1}\udc0d{s2}')
+        self.assertEqual(t.strftime('%I%p%Z\ud83d%X\udc0d'), f'{s1}\ud83d{s2}\udc0d')
+        self.assertEqual(t.strftime('%I%p%Z\udc0d%X\ud83d'), f'{s1}\udc0d{s2}\ud83d')
+        # Surrogate pairs should not recombine.
+        self.assertEqual(t.strftime('\ud83d\udc0d'), '\ud83d\udc0d')
+        self.assertEqual(t.strftime('%I%p%Z\ud83d\udc0d%X'), f'{s1}\ud83d\udc0d{s2}')
+        # Surrogate-escaped bytes should not recombine.
+        self.assertEqual(t.strftime('\udcf0\udc9f\udc90\udc8d'), '\udcf0\udc9f\udc90\udc8d')
+        self.assertEqual(t.strftime('%I%p%Z\udcf0\udc9f\udc90\udc8d%X'), f'{s1}\udcf0\udc9f\udc90\udc8d{s2}')
+        # gh-124531: The null character should not terminate the format string.
+        self.assertEqual(t.strftime('\0'), '\0')
+        self.assertEqual(t.strftime('\0'*1000), '\0'*1000)
+        self.assertEqual(t.strftime('\0%I%p%Z\0%X'), f'\0{s1}\0{s2}')
+        self.assertEqual(t.strftime('%I%p%Z\0%X\0'), f'{s1}\0{s2}\0')
+
     def test_format(self):
         t = self.theclass(1, 2, 3, 4)
         self.assertEqual(t.__format__(''), str(t))
@@ -4002,9 +4050,8 @@ def tzname(self, dt): return self.tz
         self.assertRaises(TypeError, t.strftime, "%Z")
 
         # Issue #6697:
-        if '_Fast' in self.__class__.__name__:
-            Badtzname.tz = '\ud800'
-            self.assertRaises(ValueError, t.strftime, "%Z")
+        Badtzname.tz = '\ud800'
+        self.assertEqual(t.strftime("%Z"), '\ud800')
 
     def test_hash_edge_cases(self):
         # Offsets that overflow a basic time.

Lib/test/test_time.py

@@ -177,8 +177,33 @@ def test_strftime(self):
                 self.fail('conversion specifier: %r failed.' % format)
 
         self.assertRaises(TypeError, time.strftime, b'%S', tt)
-        # embedded null character
-        self.assertRaises(ValueError, time.strftime, '%S\0', tt)
+
+    def test_strftime_special(self):
+        tt = time.gmtime(self.t)
+        s1 = time.strftime('%c', tt)
+        s2 = time.strftime('%B', tt)
+        # gh-52551, gh-78662: Unicode strings should pass through strftime,
+        # independently from locale.
+        self.assertEqual(time.strftime('\U0001f40d', tt), '\U0001f40d')
+        self.assertEqual(time.strftime('\U0001f4bb%c\U0001f40d%B', tt), f'\U0001f4bb{s1}\U0001f40d{s2}')
+        self.assertEqual(time.strftime('%c\U0001f4bb%B\U0001f40d', tt), f'{s1}\U0001f4bb{s2}\U0001f40d')
+        # Lone surrogates should pass through.
+        self.assertEqual(time.strftime('\ud83d', tt), '\ud83d')
+        self.assertEqual(time.strftime('\udc0d', tt), '\udc0d')
+        self.assertEqual(time.strftime('\ud83d%c\udc0d%B', tt), f'\ud83d{s1}\udc0d{s2}')
+        self.assertEqual(time.strftime('%c\ud83d%B\udc0d', tt), f'{s1}\ud83d{s2}\udc0d')
+        self.assertEqual(time.strftime('%c\udc0d%B\ud83d', tt), f'{s1}\udc0d{s2}\ud83d')
+        # Surrogate pairs should not recombine.
+        self.assertEqual(time.strftime('\ud83d\udc0d', tt), '\ud83d\udc0d')
+        self.assertEqual(time.strftime('%c\ud83d\udc0d%B', tt), f'{s1}\ud83d\udc0d{s2}')
+        # Surrogate-escaped bytes should not recombine.
+        self.assertEqual(time.strftime('\udcf0\udc9f\udc90\udc8d', tt), '\udcf0\udc9f\udc90\udc8d')
+        self.assertEqual(time.strftime('%c\udcf0\udc9f\udc90\udc8d%B', tt), f'{s1}\udcf0\udc9f\udc90\udc8d{s2}')
+        # gh-124531: The null character should not terminate the format string.
+        self.assertEqual(time.strftime('\0', tt), '\0')
+        self.assertEqual(time.strftime('\0'*1000, tt), '\0'*1000)
+        self.assertEqual(time.strftime('\0%c\0%B', tt), f'\0{s1}\0{s2}')
+        self.assertEqual(time.strftime('%c\0%B\0', tt), f'{s1}\0{s2}\0')
 
     def _bounds_checking(self, func):
         # Make sure that strftime() checks the bounds of the various parts

Misc/NEWS.d/next/Library/2024-10-09-17-07-33.gh-issue-52551.PBakSY.rst

@@ -0,0 +1,8 @@
+Fix encoding issues in :func:`time.strftime`, the
+:meth:`~datetime.datetime.strftime` method of the :mod:`datetime` classes
+:class:`~datetime.datetime`, :class:`~datetime.date` and
+:class:`~datetime.time` and formatting of these classes. Characters not
+encodable in the current locale are now acceptable in the format string.
+Surrogate pairs and sequence of surrogatescape-encoded bytes are no longer
+recombinated. Embedded null character no longer terminates the format
+string.

Modules/_datetimemodule.c

@@ -1501,7 +1501,7 @@ make_somezreplacement(PyObject *object, char *sep, PyObject *tzinfoarg)
     PyObject *tzinfo = get_tzinfo_member(object);
 
     if (tzinfo == Py_None || tzinfo == NULL) {
-        return PyBytes_FromStringAndSize(NULL, 0);
+        return PyUnicode_FromStringAndSize(NULL, 0);
     }
 
     assert(tzinfoarg != NULL);
@@ -1512,7 +1512,7 @@ make_somezreplacement(PyObject *object, char *sep, PyObject *tzinfoarg)
                          tzinfoarg) < 0)
         return NULL;
 
-    return PyBytes_FromStringAndSize(buf, strlen(buf));
+    return PyUnicode_FromString(buf);
 }
 
 static PyObject *
@@ -1569,7 +1569,7 @@ make_freplacement(PyObject *object)
     else
         sprintf(freplacement, "%06d", 0);
 
-    return PyBytes_FromStringAndSize(freplacement, strlen(freplacement));
+    return PyUnicode_FromString(freplacement);
 }
 
 /* I sure don't want to reproduce the strftime code from the time module,
@@ -1590,159 +1590,124 @@ wrap_strftime(PyObject *object, PyObject *format, PyObject *timetuple,
     PyObject *Zreplacement = NULL;      /* py string, replacement for %Z */
     PyObject *freplacement = NULL;      /* py string, replacement for %f */
 
-    const char *pin;            /* pointer to next char in input format */
-    Py_ssize_t flen;            /* length of input format */
-    char ch;                    /* next char in input format */
-
-    PyObject *newfmt = NULL;            /* py string, the output format */
-    char *pnew;         /* pointer to available byte in output format */
-    size_t totalnew;            /* number bytes total in output format buffer,
-                               exclusive of trailing \0 */
-    size_t usednew;     /* number bytes used so far in output format buffer */
-
-    const char *ptoappend;      /* ptr to string to append to output buffer */
-    Py_ssize_t ntoappend;       /* # of bytes to append to output buffer */
-
     assert(object && format && timetuple);
     assert(PyUnicode_Check(format));
-    /* Convert the input format to a C string and size */
-    pin = PyUnicode_AsUTF8AndSize(format, &flen);
-    if (!pin)
+
+    PyObject *strftime = _PyImport_GetModuleAttrString("time", "strftime");
+    if (strftime == NULL) {
         return NULL;
+    }
 
     /* Scan the input format, looking for %z/%Z/%f escapes, building
      * a new format.  Since computing the replacements for those codes
      * is expensive, don't unless they're actually used.
      */
-    if (flen > INT_MAX - 1) {
-        PyErr_NoMemory();
-        goto Done;
-    }
 
-    totalnew = flen + 1;        /* realistic if no %z/%Z */
-    newfmt = PyBytes_FromStringAndSize(NULL, totalnew);
-    if (newfmt == NULL) goto Done;
-    pnew = PyBytes_AsString(newfmt);
-    usednew = 0;
-
-    while ((ch = *pin++) != '\0') {
-        if (ch != '%') {
-            ptoappend = pin - 1;
-            ntoappend = 1;
+    _PyUnicodeWriter writer;
+    _PyUnicodeWriter_Init(&writer);
+    writer.overallocate = 1;
+
+    Py_ssize_t flen = PyUnicode_GET_LENGTH(format);
+    Py_ssize_t i = 0;
+    Py_ssize_t start = 0;
+    Py_ssize_t end = 0;
+    while (i != flen) {
+        i = PyUnicode_FindChar(format, '%', i, flen, 1);
+        if (i < 0) {
+            assert(!PyErr_Occurred());
+            break;
         }
-        else if ((ch = *pin++) == '\0') {
-        /* Null byte follows %, copy only '%'.
-         *
-         * Back the pin up one char so that we catch the null check
-         * the next time through the loop.*/
-            pin--;
-            ptoappend = pin - 1;
-            ntoappend = 1;
+        end = i;
+        i++;
+        if (i == flen) {
+            break;
         }
+        Py_UCS4 ch = PyUnicode_READ_CHAR(format, i);
+        i++;
         /* A % has been seen and ch is the character after it. */
-        else if (ch == 'z') {
+        PyObject *replacement = NULL;
+        if (ch == 'z') {
             /* %z -> +HHMM */
             if (zreplacement == NULL) {
                 zreplacement = make_somezreplacement(object, "", tzinfoarg);
                 if (zreplacement == NULL)
-                    goto Done;
+                    goto Error;
             }
-            assert(zreplacement != NULL);
-            assert(PyBytes_Check(zreplacement));
-            ptoappend = PyBytes_AS_STRING(zreplacement);
-            ntoappend = PyBytes_GET_SIZE(zreplacement);
+            replacement = zreplacement;
         }
-        else if (ch == ':' && *pin == 'z' && pin++) {
+        else if (ch == ':' && i < flen && PyUnicode_READ_CHAR(format, i) == 'z') {
             /* %:z -> +HH:MM */
+            i++;
             if (colonzreplacement == NULL) {
                 colonzreplacement = make_somezreplacement(object, ":", tzinfoarg);
                 if (colonzreplacement == NULL)
-                    goto Done;
+                    goto Error;
             }
-            assert(colonzreplacement != NULL);
-            assert(PyBytes_Check(colonzreplacement));
-            ptoappend = PyBytes_AS_STRING(colonzreplacement);
-            ntoappend = PyBytes_GET_SIZE(colonzreplacement);
+            replacement = colonzreplacement;
         }
         else if (ch == 'Z') {
             /* format tzname */
             if (Zreplacement == NULL) {
                 Zreplacement = make_Zreplacement(object,
                                                  tzinfoarg);
                 if (Zreplacement == NULL)
-                    goto Done;
+                    goto Error;
             }
-            assert(Zreplacement != NULL);
-            assert(PyUnicode_Check(Zreplacement));
-            ptoappend = PyUnicode_AsUTF8AndSize(Zreplacement,
-                                                  &ntoappend);
-            if (ptoappend == NULL)
-                goto Done;
+            replacement = Zreplacement;
         }
         else if (ch == 'f') {
             /* format microseconds */
             if (freplacement == NULL) {
                 freplacement = make_freplacement(object);
                 if (freplacement == NULL)
-                    goto Done;
+                    goto Error;
             }
-            assert(freplacement != NULL);
-            assert(PyBytes_Check(freplacement));
-            ptoappend = PyBytes_AS_STRING(freplacement);
-            ntoappend = PyBytes_GET_SIZE(freplacement);
+            replacement = freplacement;
         }
         else {
             /* percent followed by something else */
-            ptoappend = pin - 2;
-            ntoappend = 2;
-        }
-
-        /* Append the ntoappend chars starting at ptoappend to
-         * the new format.
-         */
-        if (ntoappend == 0)
             continue;
-        assert(ptoappend != NULL);
-        assert(ntoappend > 0);
-        while (usednew + ntoappend > totalnew) {
-            if (totalnew > (PY_SSIZE_T_MAX >> 1)) { /* overflow */
-                PyErr_NoMemory();
-                goto Done;
-            }
-            totalnew <<= 1;
-            if (_PyBytes_Resize(&newfmt, totalnew) < 0)
-                goto Done;
-            pnew = PyBytes_AsString(newfmt) + usednew;
         }
-        memcpy(pnew, ptoappend, ntoappend);
-        pnew += ntoappend;
-        usednew += ntoappend;
-        assert(usednew <= totalnew);
+        assert(replacement != NULL);
+        assert(PyUnicode_Check(replacement));
+        if (_PyUnicodeWriter_WriteSubstring(&writer, format, start, end) < 0) {
+            goto Error;
+        }
+        start = i;
+        if (_PyUnicodeWriter_WriteStr(&writer, replacement) < 0) {
+            goto Error;
+        }
     }  /* end while() */
 
-    if (_PyBytes_Resize(&newfmt, usednew) < 0)
-        goto Done;
-    {
-        PyObject *format;
-        PyObject *strftime = _PyImport_GetModuleAttrString("time", "strftime");
-
-        if (strftime == NULL)
+    PyObject *newformat;
+    if (start == 0) {
+        _PyUnicodeWriter_Dealloc(&writer);
+        newformat = Py_NewRef(format);
+    }
+    else {
+        if (_PyUnicodeWriter_WriteSubstring(&writer, format, start, flen) < 0) {
+            goto Error;
+        }
+        newformat = _PyUnicodeWriter_Finish(&writer);
+        if (newformat == NULL) {
             goto Done;
-        format = PyUnicode_FromString(PyBytes_AS_STRING(newfmt));
-        if (format != NULL) {
-            result = PyObject_CallFunctionObjArgs(strftime,
-                                                   format, timetuple, NULL);
-            Py_DECREF(format);
         }
-        Py_DECREF(strftime);
     }
+    result = PyObject_CallFunctionObjArgs(strftime,
+                                          newformat, timetuple, NULL);
+    Py_DECREF(newformat);
+
  Done:
     Py_XDECREF(freplacement);
     Py_XDECREF(zreplacement);
     Py_XDECREF(colonzreplacement);
     Py_XDECREF(Zreplacement);
-    Py_XDECREF(newfmt);
+    Py_XDECREF(strftime);
     return result;
+
+ Error:
+    _PyUnicodeWriter_Dealloc(&writer);
+    goto Done;
 }
 
 /* ---------------------------------------------------------------------------