[3.12] gh-119614: Fix truncation of strings with embedded null characters in Tkinter (GH-120909) (GH-120939)

Now the null character is always represented as \xc0\x80 for
Tcl_NewStringObj().
(cherry picked from commit c38e2f6)

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

Author: miss-islington

Lib/test/test_tcl.py

@@ -73,6 +73,18 @@ def testCall(self):
         tcl.call('set','a','1')
         self.assertEqual(tcl.call('set','a'),'1')
 
+    def test_call_passing_null(self):
+        tcl = self.interp
+        tcl.call('set', 'a', 'a\0b')  # ASCII-only
+        self.assertEqual(tcl.getvar('a'), 'a\x00b')
+        self.assertEqual(tcl.call('set', 'a'), 'a\x00b')
+        self.assertEqual(tcl.eval('set a'), 'a\x00b')
+
+        tcl.call('set', 'a', '\u20ac\0')  # non-ASCII
+        self.assertEqual(tcl.getvar('a'), '\u20ac\x00')
+        self.assertEqual(tcl.call('set', 'a'), '\u20ac\x00')
+        self.assertEqual(tcl.eval('set a'), '\u20ac\x00')
+
     def testCallException(self):
         tcl = self.interp
         self.assertRaises(TclError,tcl.call,'set','a')
@@ -98,6 +110,18 @@ def testSetVar(self):
         tcl.setvar('a','1')
         self.assertEqual(tcl.eval('set a'),'1')
 
+    def test_setvar_passing_null(self):
+        tcl = self.interp
+        tcl.setvar('a', 'a\0b')  # ASCII-only
+        self.assertEqual(tcl.getvar('a'), 'a\x00b')
+        self.assertEqual(tcl.call('set', 'a'), 'a\x00b')
+        self.assertEqual(tcl.eval('set a'), 'a\x00b')
+
+        tcl.setvar('a', '\u20ac\0')  # non-ASCII
+        self.assertEqual(tcl.getvar('a'), '\u20ac\x00')
+        self.assertEqual(tcl.call('set', 'a'), '\u20ac\x00')
+        self.assertEqual(tcl.eval('set a'), '\u20ac\x00')
+
     def testSetVarArray(self):
         tcl = self.interp
         tcl.setvar('a(1)','1')

Lib/test/test_tkinter/test_misc.py

@@ -382,6 +382,15 @@ def test_info_patchlevel(self):
             self.assertEqual(vi.micro, 0)
         self.assertTrue(str(vi).startswith(f'{vi.major}.{vi.minor}'))
 
+    def test_embedded_null(self):
+        widget = tkinter.Entry(self.root)
+        widget.insert(0, 'abc\0def')  # ASCII-only
+        widget.selection_range(0, 'end')
+        self.assertEqual(widget.selection_get(), 'abc\x00def')
+        widget.insert(0, '\u20ac\0')  # non-ASCII
+        widget.selection_range(0, 'end')
+        self.assertEqual(widget.selection_get(), '\u20ac\0abc\x00def')
+
 
 class EventTest(AbstractTkTest, unittest.TestCase):
 

Misc/NEWS.d/next/Library/2024-06-23-17-50-40.gh-issue-119614.vwPGLB.rst

@@ -0,0 +1,2 @@
+Fix truncation of strings with embedded null characters in some internal
+operations in :mod:`tkinter`.

Modules/_tkinter.c

@@ -516,7 +516,7 @@ unicodeFromTclObj(TkappObject *tkapp, Tcl_Obj *value)
         else
             Py_UNREACHABLE();
     }
-#endif
+#endif /* USE_TCL_UNICODE */
     const char *s = Tcl_GetStringFromObj(value, &len);
     return unicodeFromTclStringAndSize(s, len);
 }
@@ -1024,7 +1024,9 @@ AsObj(PyObject *value)
             PyErr_SetString(PyExc_OverflowError, "string is too long");
             return NULL;
         }
-        if (PyUnicode_IS_ASCII(value)) {
+        if (PyUnicode_IS_ASCII(value) &&
+            strlen(PyUnicode_DATA(value)) == (size_t)PyUnicode_GET_LENGTH(value))
+        {
             return Tcl_NewStringObj((const char *)PyUnicode_DATA(value),
                                     (int)size);
         }
@@ -1039,9 +1041,6 @@ AsObj(PyObject *value)
                     "surrogatepass", NATIVE_BYTEORDER);
         else
             Py_UNREACHABLE();
-#else
-        encoded = _PyUnicode_AsUTF8String(value, "surrogateescape");
-#endif
         if (!encoded) {
             return NULL;
         }
@@ -1051,12 +1050,39 @@ AsObj(PyObject *value)
             PyErr_SetString(PyExc_OverflowError, "string is too long");
             return NULL;
         }
-#if USE_TCL_UNICODE
         result = Tcl_NewUnicodeObj((const Tcl_UniChar *)PyBytes_AS_STRING(encoded),
                                    (int)(size / sizeof(Tcl_UniChar)));
 #else
+        encoded = _PyUnicode_AsUTF8String(value, "surrogateescape");
+        if (!encoded) {
+            return NULL;
+        }
+        size = PyBytes_GET_SIZE(encoded);
+        if (strlen(PyBytes_AS_STRING(encoded)) != (size_t)size) {
+            /* The string contains embedded null characters.
+             * Tcl needs a null character to be represented as \xc0\x80 in
+             * the Modified UTF-8 encoding.  Otherwise the string can be
+             * truncated in some internal operations.
+             *
+             * NOTE: stringlib_replace() could be used here, but optimizing
+             * this obscure case isn't worth it unless stringlib_replace()
+             * was already exposed in the C API for other reasons. */
+            Py_SETREF(encoded,
+                      PyObject_CallMethod(encoded, "replace", "y#y#",
+                                          "\0", (Py_ssize_t)1,
+                                          "\xc0\x80", (Py_ssize_t)2));
+            if (!encoded) {
+                return NULL;
+            }
+            size = PyBytes_GET_SIZE(encoded);
+        }
+        if (size > INT_MAX) {
+            Py_DECREF(encoded);
+            PyErr_SetString(PyExc_OverflowError, "string is too long");
+            return NULL;
+        }
         result = Tcl_NewStringObj(PyBytes_AS_STRING(encoded), (int)size);
-#endif
+#endif /* USE_TCL_UNICODE */
         Py_DECREF(encoded);
         return result;
     }