bpo-36389: _PyObject_CheckConsistency() available in release mode (GH-16612)

bpo-36389, bpo-38376: The _PyObject_CheckConsistency() function is
now also available in release mode. For example, it can be used to
debug a crash in the visit_decref() function of the GC.

Modify the following functions to also work in release mode:

* _PyDict_CheckConsistency()
* _PyObject_CheckConsistency()
* _PyType_CheckConsistency()
* _PyUnicode_CheckConsistency()

Other changes:

* _PyMem_IsPtrFreed(ptr) now also returns 1 if ptr is NULL
  (equals to 0).
* _PyBytesWriter_CheckConsistency() now returns 1 and is only used
  with assert().
* Reorder _PyObject_Dump() to write safe fields first, and only
  attempt to render repr() at the end.

Author: vstinner

Include/cpython/unicodeobject.h

@@ -246,6 +246,10 @@ typedef struct {
     } data;                     /* Canonical, smallest-form Unicode buffer */
 } PyUnicodeObject;
 
+PyAPI_FUNC(int) _PyUnicode_CheckConsistency(
+    PyObject *op,
+    int check_content);
+
 /* Fast access macros */
 #define PyUnicode_WSTR_LENGTH(op) \
     (PyUnicode_IS_COMPACT_ASCII(op) ?                  \

Include/internal/pycore_object.h

@@ -11,7 +11,6 @@ extern "C" {
 #include "pycore_pystate.h"   /* _PyRuntime.gc */
 
 PyAPI_FUNC(int) _PyType_CheckConsistency(PyTypeObject *type);
-PyAPI_FUNC(int) _PyUnicode_CheckConsistency(PyObject *op, int check_content);
 PyAPI_FUNC(int) _PyDict_CheckConsistency(PyObject *mp, int check_content);
 
 /* Tell the GC to track this object.

Include/internal/pycore_pymem.h

@@ -155,8 +155,9 @@ PyAPI_FUNC(int) _PyMem_SetDefaultAllocator(
     PyMemAllocatorEx *old_alloc);
 
 /* Heuristic checking if a pointer value is newly allocated
-   (uninitialized) or newly freed. The pointer is not dereferenced, only the
-   pointer value is checked.
+   (uninitialized), newly freed or NULL (is equal to zero).
+
+   The pointer is not dereferenced, only the pointer value is checked.
 
    The heuristic relies on the debug hooks on Python memory allocators which
    fills newly allocated memory with CLEANBYTE (0xCD) and newly freed memory
@@ -166,11 +167,13 @@ static inline int _PyMem_IsPtrFreed(void *ptr)
 {
     uintptr_t value = (uintptr_t)ptr;
 #if SIZEOF_VOID_P == 8
-    return (value == (uintptr_t)0xCDCDCDCDCDCDCDCD
+    return (value == 0
+            || value == (uintptr_t)0xCDCDCDCDCDCDCDCD
             || value == (uintptr_t)0xDDDDDDDDDDDDDDDD
             || value == (uintptr_t)0xFDFDFDFDFDFDFDFD);
 #elif SIZEOF_VOID_P == 4
-    return (value == (uintptr_t)0xCDCDCDCD
+    return (value == 0
+            || value == (uintptr_t)0xCDCDCDCD
             || value == (uintptr_t)0xDDDDDDDD
             || value == (uintptr_t)0xFDFDFDFD);
 #else

Include/unicodeobject.h

@@ -1032,16 +1032,6 @@ PyAPI_FUNC(int) PyUnicode_IsIdentifier(PyObject *s);
 
 /* === Characters Type APIs =============================================== */
 
-#if defined(Py_DEBUG) && !defined(Py_LIMITED_API)
-PyAPI_FUNC(int) _PyUnicode_CheckConsistency(
-    PyObject *op,
-    int check_content);
-#elif !defined(NDEBUG)
-/* For asserts that call _PyUnicode_CheckConsistency(), which would
- * otherwise be a problem when building with asserts but without Py_DEBUG. */
-#define _PyUnicode_CheckConsistency(op, check_content) PyUnicode_Check(op)
-#endif
-
 #ifndef Py_LIMITED_API
 #  define Py_CPYTHON_UNICODEOBJECT_H
 #  include  "cpython/unicodeobject.h"

Lib/test/test_capi.py

@@ -719,6 +719,9 @@ def check_pyobject_is_freed(self, func_name):
         ''')
         assert_python_ok('-c', code, PYTHONMALLOC=self.PYTHONMALLOC)
 
+    def test_pyobject_null_is_freed(self):
+        self.check_pyobject_is_freed('check_pyobject_null_is_freed')
+
     def test_pyobject_uninitialized_is_freed(self):
         self.check_pyobject_is_freed('check_pyobject_uninitialized_is_freed')
 

Lib/test/test_gc.py

@@ -985,16 +985,19 @@ def test_refcount_errors(self):
             br'gcmodule\.c:[0-9]+: gc_decref: Assertion "gc_get_refs\(g\) > 0" failed.')
         self.assertRegex(stderr,
             br'refcount is too small')
+        # "address : 0x7fb5062efc18"
+        # "address : 7FB5062EFC18"
+        address_regex = br'[0-9a-fA-Fx]+'
         self.assertRegex(stderr,
-            br'object  : \[1, 2, 3\]')
+            br'object address  : ' + address_regex)
         self.assertRegex(stderr,
-            br'type    : list')
+            br'object refcount : 1')
         self.assertRegex(stderr,
-            br'refcount: 1')
-        # "address : 0x7fb5062efc18"
-        # "address : 7FB5062EFC18"
+            br'object type     : ' + address_regex)
+        self.assertRegex(stderr,
+            br'object type name: list')
         self.assertRegex(stderr,
-            br'address : [0-9a-fA-Fx]+')
+            br'object repr     : \[1, 2, 3\]')
 
 
 class GCTogglingTests(unittest.TestCase):

Misc/NEWS.d/next/C API/2019-10-07-17-15-09.bpo-36389.hFX_jD.rst

@@ -0,0 +1,3 @@
+The ``_PyObject_CheckConsistency()`` function is now also available in release
+mode. For example, it can be used to debug a crash in the ``visit_decref()``
+function of the GC.

Modules/_testcapimodule.c

@@ -4611,6 +4611,14 @@ test_pyobject_is_freed(const char *test_name, PyObject *op)
 }
 
 
+static PyObject*
+check_pyobject_null_is_freed(PyObject *self, PyObject *Py_UNUSED(args))
+{
+    PyObject *op = NULL;
+    return test_pyobject_is_freed("check_pyobject_null_is_freed", op);
+}
+
+
 static PyObject*
 check_pyobject_uninitialized_is_freed(PyObject *self, PyObject *Py_UNUSED(args))
 {
@@ -5475,6 +5483,7 @@ static PyMethodDef TestMethods[] = {
     {"pymem_api_misuse", pymem_api_misuse, METH_NOARGS},
     {"pymem_malloc_without_gil", pymem_malloc_without_gil, METH_NOARGS},
     {"pymem_getallocatorsname", test_pymem_getallocatorsname, METH_NOARGS},
+    {"check_pyobject_null_is_freed", check_pyobject_null_is_freed, METH_NOARGS},
     {"check_pyobject_uninitialized_is_freed", check_pyobject_uninitialized_is_freed, METH_NOARGS},
     {"check_pyobject_forbidden_bytes_is_freed", check_pyobject_forbidden_bytes_is_freed, METH_NOARGS},
     {"check_pyobject_freed_is_freed", check_pyobject_freed_is_freed, METH_NOARGS},

Modules/gcmodule.c

@@ -375,7 +375,6 @@ update_refs(PyGC_Head *containers)
 static int
 visit_decref(PyObject *op, void *data)
 {
-    assert(op != NULL);
     _PyObject_ASSERT(op, !_PyObject_IsFreed(op));
 
     if (PyObject_IS_GC(op)) {

Objects/bytesobject.c

@@ -3206,10 +3206,10 @@ _PyBytesWriter_GetSize(_PyBytesWriter *writer, char *str)
     return str - start;
 }
 
-Py_LOCAL_INLINE(void)
+#ifndef NDEBUG
+Py_LOCAL_INLINE(int)
 _PyBytesWriter_CheckConsistency(_PyBytesWriter *writer, char *str)
 {
-#ifdef Py_DEBUG
     char *start, *end;
 
     if (writer->use_small_buffer) {
@@ -3239,15 +3239,16 @@ _PyBytesWriter_CheckConsistency(_PyBytesWriter *writer, char *str)
     end = start + writer->allocated;
     assert(str != NULL);
     assert(start <= str && str <= end);
-#endif
+    return 1;
 }
+#endif
 
 void*
 _PyBytesWriter_Resize(_PyBytesWriter *writer, void *str, Py_ssize_t size)
 {
     Py_ssize_t allocated, pos;
 
-    _PyBytesWriter_CheckConsistency(writer, str);
+    assert(_PyBytesWriter_CheckConsistency(writer, str));
     assert(writer->allocated < size);
 
     allocated = size;
@@ -3303,7 +3304,7 @@ _PyBytesWriter_Resize(_PyBytesWriter *writer, void *str, Py_ssize_t size)
     writer->allocated = allocated;
 
     str = _PyBytesWriter_AsString(writer) + pos;
-    _PyBytesWriter_CheckConsistency(writer, str);
+    assert(_PyBytesWriter_CheckConsistency(writer, str));
     return str;
 
 error:
@@ -3316,7 +3317,7 @@ _PyBytesWriter_Prepare(_PyBytesWriter *writer, void *str, Py_ssize_t size)
 {
     Py_ssize_t new_min_size;
 
-    _PyBytesWriter_CheckConsistency(writer, str);
+    assert(_PyBytesWriter_CheckConsistency(writer, str));
     assert(size >= 0);
 
     if (size == 0) {
@@ -3377,7 +3378,7 @@ _PyBytesWriter_Finish(_PyBytesWriter *writer, void *str)
     Py_ssize_t size;
     PyObject *result;
 
-    _PyBytesWriter_CheckConsistency(writer, str);
+    assert(_PyBytesWriter_CheckConsistency(writer, str));
 
     size = _PyBytesWriter_GetSize(writer, str);
     if (size == 0 && !writer->use_bytearray) {