bpo-9263: Dump Python object on GC assertion failure (GH-10062)

Changes:

* Add _PyObject_AssertFailed() function.
* Add _PyObject_ASSERT() and _PyObject_ASSERT_WITH_MSG() macros.
* gc_decref(): replace assert() with _PyObject_ASSERT_WITH_MSG() to
  dump the faulty object if the assertion fails.

_PyObject_AssertFailed() calls:

* _PyMem_DumpTraceback(): try to log the traceback where the object
  memory has been allocated if tracemalloc is enabled.
* _PyObject_Dump(): log repr(obj).
* Py_FatalError(): log the current Python traceback.

_PyObject_AssertFailed() uses _PyObject_IsFreed() heuristic to check
if the object memory has been freed by a debug hook on Python memory
allocators.

Initial patch written by David Malcolm.

Co-Authored-By: David Malcolm <dmalcolm@redhat.com>

1 files changed, 66 insertions, 3 deletions

diff --git a/Lib/test/test_gc.py b/Lib/test/test_gc.py
index 8d806db..16b2242 100644
--- a/Lib/test/test_gc.py
+++ b/Lib/test/test_gc.py
@@ -1,14 +1,17 @@
 import unittest
 from test.support import (verbose, refcount_test, run_unittest,
                           strip_python_stderr, cpython_only, start_threads,
-                          temp_dir, requires_type_collecting, TESTFN, unlink)
+                          temp_dir, requires_type_collecting, TESTFN, unlink,
+                          import_module)
 from test.support.script_helper import assert_python_ok, make_script
 
+import gc
 import sys
+import sysconfig
+import textwrap
+import threading
 import time
-import gc
 import weakref
-import threading
 
 try:
     from _testcapi import with_tp_del
@@ -62,6 +65,14 @@ class Uncollectable(object):
     def __tp_del__(self):
         pass
 
+if sysconfig.get_config_vars().get('PY_CFLAGS', ''):
+    BUILD_WITH_NDEBUG = ('-DNDEBUG' in sysconfig.get_config_vars()['PY_CFLAGS'])
+else:
+    # Usually, sys.gettotalrefcount() is only present if Python has been
+    # compiled in debug mode. If it's missing, expect that Python has
+    # been released in release mode: with NDEBUG defined.
+    BUILD_WITH_NDEBUG = (not hasattr(sys, 'gettotalrefcount'))
+
 ### Tests
 ###############################################################################
 
@@ -878,6 +889,58 @@ class GCCallbackTests(unittest.TestCase):
         self.assertEqual(len(gc.garbage), 0)
 
 
+    @unittest.skipIf(BUILD_WITH_NDEBUG,
+                     'built with -NDEBUG')
+    def test_refcount_errors(self):
+        self.preclean()
+        # Verify the "handling" of objects with broken refcounts
+
+        # Skip the test if ctypes is not available
+        import_module("ctypes")
+
+        import subprocess
+        code = textwrap.dedent('''
+            from test.support import gc_collect, SuppressCrashReport
+
+            a = [1, 2, 3]
+            b = [a]
+
+            # Avoid coredump when Py_FatalError() calls abort()
+            SuppressCrashReport().__enter__()
+
+            # Simulate the refcount of "a" being too low (compared to the
+            # references held on it by live data), but keeping it above zero
+            # (to avoid deallocating it):
+            import ctypes
+            ctypes.pythonapi.Py_DecRef(ctypes.py_object(a))
+
+            # The garbage collector should now have a fatal error
+            # when it reaches the broken object
+            gc_collect()
+        ''')
+        p = subprocess.Popen([sys.executable, "-c", code],
+                             stdout=subprocess.PIPE,
+                             stderr=subprocess.PIPE)
+        stdout, stderr = p.communicate()
+        p.stdout.close()
+        p.stderr.close()
+        # Verify that stderr has a useful error message:
+        self.assertRegex(stderr,
+            br'gcmodule\.c:[0-9]+: gc_decref: Assertion "gc_get_refs\(g\) > 0" failed.')
+        self.assertRegex(stderr,
+            br'refcount is too small')
+        self.assertRegex(stderr,
+            br'object  : \[1, 2, 3\]')
+        self.assertRegex(stderr,
+            br'type    : list')
+        self.assertRegex(stderr,
+            br'refcount: 1')
+        # "address : 0x7fb5062efc18"
+        # "address : 7FB5062EFC18"
+        self.assertRegex(stderr,
+            br'address : [0-9a-fA-Fx]+')
+
+
 class GCTogglingTests(unittest.TestCase):
     def setUp(self):
         gc.enable()