test_sys_settrace segfaults on main

[JelleZijlstra]

Crash report

On clean main:

% ./python.exe -m test -R 3:3 test_sys_settrace
Raised RLIMIT_NOFILE: 256 -> 1024
0:00:00 load avg: 2.44 Run tests sequentially
0:00:00 load avg: 2.44 [1/1] test_sys_settrace
beginning 6 repetitions
123456
...Fatal Python error: Segmentation fault

Current thread 0x00000001fd5ade00 (most recent call first):
  File "/Users/jelle/py/cpython/Lib/test/test_sys_settrace.py", line 1912 in trace
  File "/Users/jelle/py/cpython/Lib/test/test_sys_settrace.py", line 2162 in test_jump_backwards_into_while_block
  File "/Users/jelle/py/cpython/Lib/test/test_sys_settrace.py", line 1975 in run_test
  File "/Users/jelle/py/cpython/Lib/test/test_sys_settrace.py", line 2008 in test
  File "/Users/jelle/py/cpython/Lib/unittest/case.py", line 589 in _callTestMethod
  File "/Users/jelle/py/cpython/Lib/unittest/case.py", line 634 in run
  File "/Users/jelle/py/cpython/Lib/unittest/case.py", line 690 in __call__
  File "/Users/jelle/py/cpython/Lib/unittest/suite.py", line 122 in run
  File "/Users/jelle/py/cpython/Lib/unittest/suite.py", line 84 in __call__
  File "/Users/jelle/py/cpython/Lib/unittest/suite.py", line 122 in run
  File "/Users/jelle/py/cpython/Lib/unittest/suite.py", line 84 in __call__
  File "/Users/jelle/py/cpython/Lib/unittest/suite.py", line 122 in run
  File "/Users/jelle/py/cpython/Lib/unittest/suite.py", line 84 in __call__
  File "/Users/jelle/py/cpython/Lib/test/support/testresult.py", line 146 in run
  File "/Users/jelle/py/cpython/Lib/test/support/__init__.py", line 1141 in _run_suite
  File "/Users/jelle/py/cpython/Lib/test/support/__init__.py", line 1268 in run_unittest
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 393 in run_unittest
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 447 in test_func
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/refleak.py", line 91 in dash_R
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 405 in regrtest_runner
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 451 in _load_run_test
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 491 in _runtest_env_changed_exc
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 348 in _runtest
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/runtest.py", line 375 in runtest
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 462 in run_test
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 495 in run_tests_sequentially
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 632 in run_tests
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 855 in action_run_tests
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 883 in _main
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 822 in main
  File "/Users/jelle/py/cpython/Lib/test/libregrtest/main.py", line 891 in main
  File "/Users/jelle/py/cpython/Lib/test/__main__.py", line 2 in <module>
  File "/Users/jelle/py/cpython/Lib/runpy.py", line 88 in _run_code
  File "/Users/jelle/py/cpython/Lib/runpy.py", line 198 in _run_module_as_main

Extension modules: _testcapi (total: 1)
zsh: segmentation fault  ./python.exe -m test -R 3:3 test_sys_settrace

I saw this in buildbots on my PR #109123 but it also reproduces on main for me locally.

Stack trace:

Process 47388 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x8)
    frame #0: 0x000000010008c140 python.exe`code_richcompare(self=0x00000001025453f0, other=0x00000001025453f0, op=3) at codeobject.c:1808:39 [opt]
   1805	
   1806	        if (co_code == ENTER_EXECUTOR) {
   1807	            const int exec_index = co_arg;
-> 1808	            _PyExecutorObject *exec = co->co_executors->executors[exec_index];
   1809	            co_code = exec->vm_data.opcode;
   1810	            co_arg = exec->vm_data.oparg;
   1811	        }
Target 0: (python.exe) stopped.
warning: python.exe was compiled with optimization - stepping may behave oddly; variables may not be available.
(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x8)
  * frame #0: 0x000000010008c140 python.exe`code_richcompare(self=0x00000001025453f0, other=0x00000001025453f0, op=3) at codeobject.c:1808:39 [opt]
    frame #1: 0x00000001000d2b14 python.exe`do_richcompare(tstate=0x00000001004ea308, v=0x00000001025453f0, w=0x00000001025453f0, op=3) at object.c:787:15 [opt]
    frame #2: 0x00000001000d2a64 python.exe`PyObject_RichCompare(v=0x00000001025453f0, w=0x00000001025453f0, op=3) at object.c:836:21 [opt]
    frame #3: 0x0000000100168354 python.exe`_PyEval_EvalFrameDefault(tstate=<unavailable>, frame=<unavailable>, throwflag=<unavailable>) at generated_cases.c.h:2757:19 [opt]
    frame #4: 0x000000010015e0b4 python.exe`_PyEval_EvalFrame(tstate=<unavailable>, frame=<unavailable>, throwflag=<unavailable>) at pycore_ceval.h:107:16 [opt]
    frame #5: 0x000000010015e04c python.exe`_PyEval_Vector(tstate=0x00000001004ea308, func=0x00000001036c2bd0, locals=0x0000000000000000, args=0x000000016fdfc340, argcount=4, kwnames=0x0000000000000000) at ceval.c:1632:12 [opt]
    frame #6: 0x0000000100086758 python.exe`_PyFunction_Vectorcall(func=0x00000001036c2bd0, stack=0x000000016fdfc340, nargsf=<unavailable>, kwnames=0x0000000000000000) at call.c:0:45 [opt]
    frame #7: 0x00000001000894d8 python.exe`_PyObject_VectorcallTstate(tstate=0x00000001004ea308, callable=0x00000001036c2bd0, args=0x000000016fdfc340, nargsf=4, kwnames=0x0000000000000000) at pycore_call.h:187:11 [opt]
    frame #8: 0x000000010008898c python.exe`method_vectorcall(method=<unavailable>, args=0x000000016fdfc3f0, nargsf=<unavailable>, kwnames=0x0000000000000000) at classobject.c:92:18 [opt]
    frame #9: 0x00000001001dfbc0 python.exe`_PyObject_VectorcallTstate(tstate=0x00000001004ea308, callable=0x000000010372cc50, args=0x000000016fdfc3f0, nargsf=3, kwnames=0x0000000000000000) at pycore_call.h:187:11 [opt]
    frame #10: 0x00000001001e1c34 python.exe`call_trampoline(tstate=0x00000001004ea308, callback=0x000000010372cc50, frame=0x0000000103650ef0, what=2, arg=<unavailable>) at sysmodule.c:1005:24 [opt]
    frame #11: 0x00000001001e1d8c python.exe`trace_trampoline(self=<unavailable>, frame=0x0000000103650ef0, what=2, arg=0x0000000100451758) at sysmodule.c:1042:24 [opt]
    frame #12: 0x00000001001b4c24 python.exe`_Py_call_instrumentation_line(tstate=0x00000001004ea308, frame=<unavailable>, instr=<unavailable>, prev=<unavailable>) at instrumentation.c:1173:27 [opt]

The problem is that co->co_executors is NULL. This line blames to @corona10's PR #108165.

[corona10]

Would you like to check that #109107 will fix this issue?

[corona10]

With fix, I check that it will solve the issue

 ./python.exe -m test -R 3:3 test_sys_settrace
Raised RLIMIT_NOFILE: 256 -> 1024
0:00:00 load avg: 4.02 Run tests sequentially
0:00:00 load avg: 4.02 [1/1] test_sys_settrace
beginning 6 repetitions
123456
......

== Tests result: SUCCESS ==

1 test OK.

Total duration: 21.9 sec
Total tests: run=439 skipped=1
Total test files: run=1/1
Result: SUCCESS

I think that we can close the issue as duplicated.

[JelleZijlstra]

Thanks, this is a duplicate of #109052 so I'll close this issue. I'm compiling the PR now.

[corona10]

For clarification, the PR #108165 was verified that passed the test at that moment because I always check the leak test in the local environment before I submit the PR :)

It looks like the issue is raised with multiple causes, and the first triggered commit was 6f3c138 and this commit is not a root cause.
This is the same conclusion with @vstinner

I investigated the first commit with git bisecting :)