[distributed] Crash when trying to use default PG after creating new PG

[xmfan]

🐛 Describe the bug

Not sure if I'm doing something dumb, but I couldn't find docs on it and even LLMs were puzzled:

Repro:

# CRASH=1 torchrun --nproc_per_node=8 try_async_pg.py

import os
import torch
import torch.distributed as dist

rank = int(os.environ["RANK"])
world_size = int(os.environ["WORLD_SIZE"])
device = torch.device("cuda", int(rank))
torch.cuda.set_device(device)

dist.init_process_group(backend="nccl", device_id=device)
pg2 = torch.distributed.new_group(backend="nccl", device_id=device)

crash = bool(int(os.environ["CRASH"]))
if crash:
    dist.barrier()
else:
    dist.barrier(group=pg2)
    dist.barrier()

dist.destroy_process_group()

Error:

(/home/xmfan/core/a/pytorch-env) [16:41:24] ~/core/a/modded-nanogpt (ca) > CRASH=1 torchrun --nproc_per_node=8 try_async_pg.py 
WARNING:torch.distributed.run:
*****************************************
Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
*****************************************
NCCL version 2.25.1+cuda12.4
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620074 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620075 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620076 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620077 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620078 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620079 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 2620081 closing signal SIGTERM
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -11) local_rank: 6 (pid: 2620080) of binary: /home/xmfan/core/a/pytorch-env/bin/python
Traceback (most recent call last):
  File "/home/xmfan/core/a/pytorch-env/bin/torchrun", line 33, in <module>
    sys.exit(load_entry_point('torch', 'console_scripts', 'torchrun')())
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/xmfan/core/a/pytorch/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
    return f(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^
  File "/home/xmfan/core/a/pytorch/torch/distributed/run.py", line 892, in main
    run(args)
  File "/home/xmfan/core/a/pytorch/torch/distributed/run.py", line 883, in run
    elastic_launch(
  File "/home/xmfan/core/a/pytorch/torch/distributed/launcher/api.py", line 139, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/xmfan/core/a/pytorch/torch/distributed/launcher/api.py", line 270, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
=========================================================
try_async_pg.py FAILED
---------------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
---------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2025-04-02_16:41:45
  host      : devvm062.dkl0.facebook.com
  rank      : 6 (local_rank: 6)
  exitcode  : -11 (pid: 2620080)
  error_file: <N/A>
  traceback : Signal 11 (SIGSEGV) received by PID 2620080
=========================================================

Versions

Collecting environment information...
PyTorch version: 2.8.0a0+git78300c8
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A

OS: CentOS Stream 9 (x86_64)
GCC version: (GCC) 11.5.0 20240719 (Red Hat 11.5.0-5)
Clang version: 19.1.7 (CentOS 19.1.7-1.el9)
CMake version: version 3.31.4
Libc version: glibc-2.34

Python version: 3.12.2 | packaged by conda-forge | (main, Feb 16 2024, 20:50:58) [GCC 12.3.0] (64-bit runtime)
Python platform: Linux-6.4.3-0_fbk14_hardened_2601_gcd42476b84e9-x86_64-with-glibc2.34
Is CUDA available: True
CUDA runtime version: 12.4.131
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA H100
GPU 1: NVIDIA H100
GPU 2: NVIDIA H100
GPU 3: NVIDIA H100
GPU 4: NVIDIA H100
GPU 5: NVIDIA H100
GPU 6: NVIDIA H100
GPU 7: NVIDIA H100

Nvidia driver version: 550.90.07
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 52 bits physical, 57 bits virtual
Byte Order: Little Endian
CPU(s): 368
On-line CPU(s) list: 0-367
Vendor ID: AuthenticAMD
Model name: AMD EPYC 9654 96-Core Processor
CPU family: 25
Model: 17
Thread(s) per core: 1
Core(s) per socket: 368
Socket(s): 1
Stepping: 1
BogoMIPS: 4792.80
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm rep_good nopl cpuid extd_apicid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw topoext perfctr_core invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx512_bf16 clzero xsaveerptr wbnoinvd arat npt lbrv nrip_save tsc_scale vmcb_clean pausefilter pfthreshold v_vmsave_vmload vgif vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid fsrm flush_l1d arch_capabilities
Virtualization: AMD-V
Hypervisor vendor: KVM
Virtualization type: full
L1d cache: 23 MiB (368 instances)
L1i cache: 23 MiB (368 instances)
L2 cache: 184 MiB (368 instances)
L3 cache: 16 MiB (1 instance)
NUMA node(s): 1
NUMA node0 CPU(s): 0-367
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Retbleed: Not affected
Vulnerability Spec store bypass: Vulnerable
Vulnerability Spectre v1: Vulnerable: __user pointer sanitization and usercopy barriers only; no swapgs barriers
Vulnerability Spectre v2: Vulnerable, IBPB: disabled, STIBP: disabled, PBRSB-eIBRS: Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] bert_pytorch==0.0.1a4
[pip3] flake8==6.1.0
[pip3] flake8-bugbear==23.3.23
[pip3] flake8-comprehensions==3.15.0
[pip3] flake8-executable==2.1.3
[pip3] flake8-logging-format==0.9.0
[pip3] flake8-pyi==23.3.1
[pip3] flake8-simplify==0.19.3
[pip3] functorch==1.14.0a0+b71aa0b
[pip3] mypy==1.14.0
[pip3] mypy-extensions==1.0.0
[pip3] numpy==2.1.0
[pip3] onnx==1.17.0
[pip3] optree==0.13.0
[pip3] pytorch-labs-segment-anything-fast==0.2
[pip3] pytorch-triton==3.2.0+git4b3bb1f8
[pip3] torch==2.8.0a0+git78300c8
[pip3] torch_geometric==2.4.0
[pip3] torchao==0.8.0
[pip3] torchaudio==2.6.0a0+c670ad8
[pip3] torchdata==0.12.0a0+d155220
[pip3] torchdata==0.12.0a0+d155220
[pip3] torchmetrics==1.0.3
[pip3] torchmultimodal==0.1.0b0
[pip3] torchpippy==0.2.0+1bcb2bf
[pip3] torchpippy==0.2.0+1bcb2bf
[pip3] torchrec==1.1.0
[pip3] torchtext==0.17.0a0+bde7ecd
[pip3] torchtitan==0.0.2
[pip3] torchvision==0.22.0a0+d462da2
[conda] bert-pytorch 0.0.1a4 dev_0
[conda] blas 1.0 mkl
[conda] functorch 1.14.0a0+b71aa0b pypi_0 pypi
[conda] magma-cuda116 2.6.1 1 pytorch
[conda] mkl 2023.1.0 h213fc3f_46344
[conda] mkl-include 2023.1.0 h06a4308_46344
[conda] mkl-service 2.4.0 py312h5eee18b_2
[conda] mkl_fft 1.3.11 py312h5eee18b_0
[conda] mkl_random 1.2.8 py312h526ad5a_0
[conda] numpy 2.1.0 pypi_0 pypi
[conda] optree 0.13.0 pypi_0 pypi
[conda] pytorch-labs-segment-anything-fast 0.2 pypi_0 pypi
[conda] pytorch-triton 3.2.0+git4b3bb1f8 pypi_0 pypi
[conda] torch 2.8.0a0+git78300c8 dev_0
[conda] torch-geometric 2.4.0 pypi_0 pypi
[conda] torchao 0.8.0 pypi_0 pypi
[conda] torchaudio 2.6.0a0+c670ad8 dev_0
[conda] torchdata 0.12.0a0+d155220 pypi_0 pypi
[conda] torchfix 0.4.0 pypi_0 pypi
[conda] torchmetrics 1.0.3 pypi_0 pypi
[conda] torchmultimodal 0.1.0b0 pypi_0 pypi
[conda] torchpippy 0.2.0+1bcb2bf pypi_0 pypi
[conda] torchrec 1.1.0 pypi_0 pypi
[conda] torchtext 0.17.0a0+bde7ecd dev_0
[conda] torchtitan 0.0.2 pypi_0 pypi
[conda] torchvision 0.22.0a0+d462da2 dev_0

cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k

[mori360]

The code works on my server. Could I know which line breaks in your running?

[xmfan]

The dist.barrier() is causing the crash. Wondering if there's additional logs you need and how I could turn them on?

[mori360]

Could you try run with TORCH_DISTRIBUTED_DEBUG=DETAIL?
It's wired that your code just works on my side.

[xmfan]

I can repro consistently using local build, but I can't repro using nightly downloaded from pytorch.org. I suspect my environment is in a bad state, but I still want to debug this in case it's an actual bug with distributed post-crash cleanup

WARNING:torch.distributed.run:
*****************************************
Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
*****************************************
[Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
[Gloo] Rank [Gloo] Rank 76 is connected to  is connected to 77 peer ranks.  peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 77

[rank4]:[W402 22:21:57.514984780 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 4]  using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank3]:[W402 22:21:57.514985051 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank2]:[W402 22:21:57.515030411 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank5]:[W402 22:21:57.515023870 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 5]  using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank6]:[W402 22:21:57.515042309 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 6]  using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank7]:[W402 22:21:57.515088839 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 7]  using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank1]:[W402 22:21:57.515080096 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
[rank0]:[W402 22:21:57.515107858 ProcessGroupNCCL.cpp:4776] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
NCCL version 2.25.1+cuda12.4
[W402 22:22:01.788149026 ProcessGroupNCCL.cpp:1501] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
(/home/xmfan/core/a/pytorch-env) [22:22:03] ~/core/a/modded-nanogpt (ca) > 

[wconstab]

2 comments.
1 if you can get a stack trace from gdb it might be helpful
2 is your logs show gloo is being initialized, which is surprising given your repro script appears to only initialize nccl- maybe something to look into further

[xmfan]

i added a dist.breakpoint() after the first PG initialization and attached gdb to rank 0

Here's the gdb backtrace:

0x00007fcef20fde2c in read () from /lib64/libc.so.6
(gdb) c
Continuing.
[New Thread 0x7fcc467dd640 (LWP 1668407)]
[New Thread 0x7fcc45fdc640 (LWP 1668408)]
[Thread 0x7fcc45fdc640 (LWP 1668408) exited]
[Thread 0x7fcc467dd640 (LWP 1668407) exited]
[New Thread 0x7fcc45fdc640 (LWP 1668463)]
[New Thread 0x7fcc3922a640 (LWP 1668483)]
[New Thread 0x7fcc467dd640 (LWP 1668489)]
[New Thread 0x7fcc38a29640 (LWP 1668493)]
[New Thread 0x7fcc11fff640 (LWP 1668567)]
[New Thread 0x7fcc117fe640 (LWP 1668580)]
[Thread 0x7fcc38a29640 (LWP 1668493) exited]
[Thread 0x7fcc467dd640 (LWP 1668489) exited]
[New Thread 0x7fcc467dd640 (LWP 1668590)]
[Thread 0x7fcc467dd640 (LWP 1668590) exited]

Thread 1 "python" received signal SIGSEGV, Segmentation fault.
0x00007fcecdda08fc in ncclCommSetAsyncError (comm=comm@entry=0x1, nextState=nextState@entry=ncclInProgress) at init.cc:1804
1804      __atomic_store_n(&comm->asyncResult, nextState, __ATOMIC_RELEASE);
(gdb) bt
#0  0x00007fcecdda08fc in ncclCommSetAsyncError (comm=comm@entry=0x1, nextState=nextState@entry=ncclInProgress) at init.cc:1804
#1  0x00007fcecdd9ca3d in ncclGroupEndInternal (simInfo=simInfo@entry=0x0) at group.cc:559
#2  0x00007fcecdda6a7d in ncclCommFinalize (comm=0x557a2d637f00) at init.cc:1957
#3  0x00007fcecb21d6fd in c10d::NCCLComm::finalize() () from /home/xmfan/core/a/pytorch/torch/lib/libtorch_cuda.so
#4  0x00007fcecb2566a9 in c10d::ProcessGroupNCCL::shutdown() () from /home/xmfan/core/a/pytorch/torch/lib/libtorch_cuda.so
#5  0x00007fcee10989fd in c10d::ProcessGroup::shutdown() () from /home/xmfan/core/a/pytorch/torch/lib/libtorch_cpu.so
#6  0x00007fcee966c2c4 in pybind11::cpp_function::initialize<pybind11::cpp_function::initialize<void, c10d::ProcessGroup, , pybind11::name, pybind11::is_method, pybind11::sibling, pybind11::call_guard<pybind11::gil_scoped_release>, char [27]>(void (c10d::ProcessGroup::*)(), pybind11::name const&, pybind11::is_method const&, pybind11::sibling const&, pybind11::call_guard<pybind11::gil_scoped_release> const&, char const (&) [27])::{lambda(c10d::ProcessGroup*)#1}, void, c10d::ProcessGroup*, pybind11::name, pybind11::is_method, pybind11::sibling, pybind11::call_guard<pybind11::gil_scoped_release>, char [27]>(pybind11::cpp_function::initialize<void, c10d::ProcessGroup, , pybind11::name, pybind11::is_method, pybind11::sibling, pybind11::call_guard<pybind11::gil_scoped_release>, char [27]>(void (c10d::ProcessGroup::*)(), pybind11::name const&, pybind11::is_method const&, pybind11::sibling const&, pybind11::call_guard<pybind11::gil_scoped_release> const&, char const (&) [27])::{lambda(c10d::ProcessGroup*)#1}&&, void (*)(c10d::ProcessGroup*), pybind11::name const&, pybind11::is_method const&, pybind11::sibling const&, pybind11::call_guard<pybind11::gil_scoped_release> const&, char const (&) [27])::{lambda(pybind11::detail::function_call&)#3}::_FUN(pybind11::detail::function_call) ()
   from /home/xmfan/core/a/pytorch/torch/lib/libtorch_python.so
#7  0x00007fcee8d803cd in pybind11::cpp_function::dispatcher(_object*, _object*, _object*) () from /home/xmfan/core/a/pytorch/torch/lib/libtorch_python.so
#8  0x0000557a252b44a1 in cfunction_call (func=0x7fce7366bf10, args=0x7fce712d4220, kwargs=0x0) at /usr/local/src/conda/python-3.12.2/Objects/methodobject.c:537
#9  0x0000557a252919ab in _PyObject_MakeTpCall (tstate=0x557a2572d3c8 <_PyRuntime+459656>, callable=0x7fce7366bf10, args=<optimized out>, nargs=1, keywords=0x0)
    at /usr/local/src/conda/python-3.12.2/Include/object.h:705
#10 0x0000557a25195a7e in _PyEval_EvalFrameDefault (tstate=<optimized out>, frame=0x7fcef247d098, throwflag=<optimized out>) at Python/bytecodes.c:2706
#11 0x0000557a2535ccae in PyEval_EvalCode (co=0x557a25867540, globals=<optimized out>, locals=0x7fcef232dc80) at /usr/local/src/conda/python-3.12.2/Python/ceval.c:578
#12 0x0000557a2538045a in run_eval_code_obj (tstate=0x557a2572d3c8 <_PyRuntime+459656>, co=0x557a25867540, globals=0x7fcef232dc80, locals=0x7fcef232dc80) at /usr/local/src/conda/python-3.12.2/Python/pythonrun.c:1722
#13 0x0000557a2537bacb in run_mod (mod=<optimized out>, filename=0x7fcef1df42d0, globals=0x7fcef232dc80, locals=0x7fcef232dc80, flags=0x7ffcb9766540, arena=0x7fcef224d830)
    at /usr/local/src/conda/python-3.12.2/Python/pythonrun.c:1743
#14 0x0000557a25392230 in pyrun_file (fp=0x557a2572f490, filename=0x7fcef1df42d0, start=<optimized out>, globals=0x7fcef232dc80, locals=0x7fcef232dc80, closeit=1, flags=0x7ffcb9766540)
    at /usr/local/src/conda/python-3.12.2/Python/pythonrun.c:1643
#15 0x0000557a25391e0e in _PyRun_SimpleFileObject (fp=0x557a2572f490, filename=0x7fcef1df42d0, closeit=1, flags=0x7ffcb9766540) at /usr/local/src/conda/python-3.12.2/Python/pythonrun.c:433
#16 0x0000557a25391684 in _PyRun_AnyFileObject (fp=0x557a2572f490, filename=0x7fcef1df42d0, closeit=1, flags=0x7ffcb9766540) at /usr/local/src/conda/python-3.12.2/Python/pythonrun.c:78
#17 0x0000557a2538b6b2 in pymain_run_file_obj (skip_source_first_line=0, filename=0x7fcef1df42d0, program_name=0x7fcef1df43f0) at /usr/local/src/conda/python-3.12.2/Modules/main.c:360
#18 pymain_run_file (config=0x557a256cffa8 <_PyRuntime+77672>) at /usr/local/src/conda/python-3.12.2/Modules/main.c:379
#19 pymain_run_python (exitcode=0x7ffcb9766514) at /usr/local/src/conda/python-3.12.2/Modules/main.c:629
#20 Py_RunMain () at /usr/local/src/conda/python-3.12.2/Modules/main.c:709
#21 0x0000557a25347217 in Py_BytesMain (argc=<optimized out>, argv=<optimized out>) at /usr/local/src/conda/python-3.12.2/Modules/main.c:763
#22 0x00007fcef20295d0 in __libc_start_call_main () from /lib64/libc.so.6
#23 0x00007fcef2029680 in __libc_start_main_impl () from /lib64/libc.so.6
#24 0x0000557a25347095 in _start () at /usr/local/src/conda/python-3.12.2/Parser/parser.c:35620
(gdb)