Fixing finite part of triangle configuration with real masses and m1=p1, m2=m3=p2=0

[scarrazza]

This PR fixes a bug in the triangle computation when m1=p1 and m2=m3=p2=0 in real mass regime.
We thank @JacekHoleczek for pointing out this issue.

Bug description

The following configuration:

  double mu2 = 1;
  vector<double> p = { 3.0380490000000004e+00 , 0.0000000000000000e+00 , 3.2368307443778059e+00 };
  vector<double> m = { 3.0380490000000004e+00 , 0.0000000000000000e+00 , 0.0000000000000000e+00 };
  vector<complex> res(3);
  ql::Triangle<complex,double> tr;
  tr.integral(res, mu2, m, p);

generates the wrong output for the finite part:

eps0 = ( -1.2330490063749579e+01 , 2.5532491289035740e+01 )
eps1 = (  1.0922308743166440e+01 , 1.5804231235735950e+01 )
eps2 = (  2.5153215229347099e+00 , 0.0000000000000000e+00 )

Expected behaviour

This PR fixes the issue. The final output is:

eps0 = ( -1.2330490063749521e+01 , 6.9628548600292163e+01 )
eps1 = (  1.0922308743166440e+01 , 1.5804231235735934e+01 )
eps2 = (  2.5153215229347099e+00 , 0.0000000000000000e+00 )

@JacekHoleczek could you please test this branch and let me know if this fixes the issue you have reported?

[scarrazza]

This also closes #10 .

[JacekHoleczek]

Thanks, I can confirm that these patches fix the two worst cases that I had (tested on Ubuntu 20.04 / gcc 9.3.0 and CentOS 7 / gcc 4.8.5).

[scarrazza]

Thanks for testing.