CGAL::solve_quadratic_program falls into an infinite loop

[qwerty10086]

Issue Details

CGAL::solve_quadratic_program falls into an infinite loop when solving a QP, and CGAL::QP_BLAND doesn't help. The code is as follows. If I don't set the bounds of variables, Solution.status() will return unbounded. I'm not sure if this is related to the error.

Source Code

#include <iostream>
#include <CGAL/QP_functions.h>
#include <CGAL/Gmpz.h>

double asDouble(uint64_t vNum)
{
    return *((double*)&vNum);
}

int main()
{
    uint64_t RawMatrixD[3][3] =
    {
        {0x3f7eaa3603f152ce},
        {0xbfc066eaeb06439b,0x4004a227d8c1ebe2},
        {0x3fab5441711b0d76,0xbff05c9bedd40fa2,0x3fda741861e0dba0}
    };
    uint64_t RawVectorC[3] = { 0x3fdd373719889d74,0xc01907af0ffd00a0,0x40073fe9cd265d35 };
    uint64_t RawLower[3] = { 0xc0257ef04e41147c,0xc022cf45ae85fe84,0xc0373b807ab20626 };
    uint64_t RawUpper[3] = { 0xc0239ea83194a030,0xc020eefd91d98a38,0xc0364b5c6c5bcc00 };

    double MatrixD[3][3], VectorC[3], Lower[3], Upper[3];
    for (size_t i = 0; i < 3; ++i)
    {
        for (size_t k = 0; k < 3; ++k)
            MatrixD[i][k] = asDouble(RawMatrixD[i][k]);
        VectorC[i] = asDouble(RawVectorC[i]);
        Lower[i] = asDouble(RawLower[i]);
        Upper[i] = asDouble(RawUpper[i]);
    }

    std::cout << MatrixD[0][0] << std::endl;
    std::cout << MatrixD[1][0] << "  " << MatrixD[1][1] << std::endl;
    std::cout << MatrixD[2][0] << "  " << MatrixD[2][1] << "  " << MatrixD[2][2] << std::endl;
    std::cout << std::endl;
    std::cout << VectorC[0] << "  " << VectorC[1] << "  " << VectorC[2] << std::endl;
    std::cout << std::endl;
    std::cout << Lower[0] << "  " << Lower[1] << "  " << Lower[2] << std::endl;
    std::cout << std::endl;
    std::cout << Upper[0] << "  " << Upper[1] << "  " << Upper[2] << std::endl;
    std::cout << std::endl;

    CGAL::Quadratic_program<double> Qp(CGAL::SMALLER, false, 0.0, false, 0.0);

    Qp.set_d(0, 0, MatrixD[0][0]);
    Qp.set_d(1, 0, MatrixD[1][0]); Qp.set_d(1, 1, MatrixD[1][1]);
    Qp.set_d(2, 0, MatrixD[2][0]); Qp.set_d(2, 1, MatrixD[2][1]); Qp.set_d(2, 2, MatrixD[2][2]);

    Qp.set_c(0, VectorC[0]); Qp.set_c(1, VectorC[1]); Qp.set_c(2, VectorC[2]);

    Qp.set_l(0, true, Lower[0]); Qp.set_l(1, true, Lower[1]); Qp.set_l(2, true, Lower[2]);

    Qp.set_u(0, true, Upper[0]); Qp.set_u(1, true, Upper[1]); Qp.set_u(2, true, Upper[2]);

    CGAL::Quadratic_program_options Options;
    Options.set_verbosity(1);
    Options.set_pricing_strategy(CGAL::QP_BLAND);
    Options.set_auto_validation(true);
    auto Solution = CGAL::solve_quadratic_program(Qp, CGAL::Gmpz(), Options);

    std::cout << Solution;

    return 0;
}

Environment

• Operating system (Windows/Mac/Linux, 32/64 bits): Windows 64bits
• Compiler: MSVC (VS2019)
• Release or debug mode: both
• Specific flags used (if any):
• CGAL version: 5.5.2
• Boost version: 1.82.0
• Other libraries versions if used (Eigen, TBB, etc.):

[afabri]

Sorry for the late reply. No idea if meanwhile you found the mistake you made yourself. You started probably from this example in the User Manual.

But you use Gmpz as exact type whereas you can see that the example uses Gmpzf.

Let me know if we can close this issue.

[qwerty10086]

Thanks for your reply. I did see this example when the problem occured, but I ignored the exact type. Just now I tried to replace Gmpz with Gmpzf and it helped.