core dump when link libonnxruntime_providers_shared.so

[kuramawzw123]

Describe the issue

I Build with cuda-11.8 and cudnn 8.2 on centos7.9.
build shell :
./build.sh --skip_submodule_sync
--cuda_home $CUDA_HOME
--cudnn_home $CUDNN_HOME
--use_cuda
--config $CONFIG
--build_shared_lib
--parallel 32
--cmake_extra_defines ONNXRUNTIME_VERSION=$(cat ./VERSION_NUMBER)
--build_dir $BUILD_DIR
--cmake_extra_defines 'CMAKE_CUDA_ARCHITECTURES=80'
--skip_tests
for debug core dump, I build with Debug mode, --config Debug, then in build dir, generate dynamice lib,
then I write simple test with link onnxruntime lib(-L$(ONNX_LIB) -lonnxruntime
-lcustom_op_library
-lonnxruntime_providers_shared
-lonnxruntime_providers_cuda)
but programe got core dump when starts , evne not enter to main function, core dump when init global variable,
here is my core dump info
#0 0x00007fffa6467060 in __static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535)
at /home/wzw/work/ort_1.16.3/onnxruntime/onnxruntime/core/providers/shared_library/provider_bridge_provider.cc:88
#1 0x00007fffa64670d8 in _GLOBAL__sub_I_provider_bridge_provider.cc(void) ()
at /home/wzw/work/ort_1.16.3/onnxruntime/onnxruntime/core/providers/shared_library/provider_bridge_provider.cc:755
#2 0x00007ffff7dea9c3 in _dl_init_internal () from /lib64/ld-linux-x86-64.so.2
#3 0x00007ffff7ddc17a in _dl_start_user () from /lib64/ld-linux-x86-64.so.2
#4 0x0000000000000001 in ?? ()
#5 0x00007fffffffd429 in ?? ()
#6 0x0000000000000000 in ?? ()

https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/core/providers/shared_library/provider_bridge_provider.cc#L88.
ProviderHostCPU& g_host_cpu = g_host->GetProviderHostCPU();
g_host is nullptrt,
becasue Provider_GetHost() function just return gbloal var g_host, where g_host init nullptr
(https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/core/providers/shared/common.cc#L8)

To reproduce

#include
#include <onnxruntime_cxx_api.h>
int main() {
std::cout << "onnx runtime test\n" << "\n";
return 0;
}
g++ -o test test.cxx -D USE_CUDA=1 -g -I./onnxruntime -L./onnxruntime --lonnxruntime
-lcustom_op_library
-lonnxruntime_providers_shared
-lonnxruntime_providers_cuda

Urgency

No response

Platform

Linux

OS Version

CentOS Linux release 7.9.2009 (Core)

ONNX Runtime Installation

Built from Source

ONNX Runtime Version or Commit ID

1.16.3

ONNX Runtime API

C++

Architecture

X64

Execution Provider

CUDA

Execution Provider Library Version

cuda 11.8

Tasks

Beta Give feedback

No tasks being tracked yet.

[snnn]

You only need to link to onnxruntime. Please remove the following things from the link command.

-lcustom_op_library
-lonnxruntime_providers_shared
-lonnxruntime_providers_cuda