Lightgbm: Using C_API LGBM_BoosterPredictForMatSingleRow and it runs out Segmentation fault

[BuGTEa]

Environment info

Operating System: Ubuntu 18.04.3

CPU/GPU model: Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz

C++/Python/R version: C++ 11

LightGBM version or commit hash: 8f7199a

Error message

Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

Reproducible examples

int main(int argc, char** argv) {
    BoosterHandle handle;
    int p;
    auto model = LGBM_BoosterCreateFromModelfile("model.txt", &p, &handle);
    if(model != 0){
        cout << "Fail On Loaded！" << endl;
        exit(-1);
    }
    else{
        cout << "Success" << endl;
        cout << p << endl;
    }
    vector<double> test{1, 3466, 3, 4, 1, 3456, -1, 0, 14, 34, 30};
    void* in_p = static_cast<void*>(test.data());
    std::vector<double> out(1, 0);
    auto* out_result = static_cast<double*>(out.data());
    int64_t out_len;
    auto result = LGBM_BoosterPredictForMatSingleRow(&handle, in_p, C_API_DTYPE_FLOAT64, 11, 1, C_API_PREDICT_NORMAL, -1, "", &out_len, out_result);
    return 0;
}

Steps to reproduce

Run That Scripts Using CMakeLists. Model Loaded is success, but LGBM_BoosterPredictForMatSingleRow encounters segmentation fault.

[guolinke]

@imatiach-msft
And @BuGTEa , could you provide the model file (model.txt)?

[BuGTEa]

Here it is. The model.txt is generated by Python API.
model.txt

[guolinke]

you should pass handle directly in LGBM_BoosterPredictForMatSingleRow, not its pointer.