Some thorns seem to incorrectly use CCTK_GFINDEX4D

[Gabriele Bozzola]

A student in my group found the following problem in ID_converter_ILGRMHD:

If you compile Einstein Toolkit in debug mode, the code crashes. This is due to the function CCTK_GFINDEX4D that is used in some thorns (such as ID_converter_ILGRMHD). The function is used to unroll a vector index into a linear one:

int CCTK_GFIndex4D (const cGH *GH, int i, int j, int k, int l)
{
#ifdef CCTK_DEBUG
  if (i < 0 || i >= GH->cctk_lsh[0] ||
      j < 0 || j >= GH->cctk_lsh[1] ||
      k < 0 || k >= GH->cctk_lsh[2] ||
      l < 0 || l >= GH->cctk_lsh[3])
  {
    CCTK_VError (__LINE__, __FILE__, "Cactus",
                 "Grid function index out of bounds.  i=%d j=%d k=%d l=%d cctk_lsh=[%d,%d,%d,%d]",
                 i, j, k, l, GH->cctk_lsh[0], GH->cctk_lsh[1], GH->cctk_lsh[2], GH->cctk_lsh[3]);
  }
#endif
  return (i + GH->cctk_ash[0]*(j + GH->cctk_ash[1]*(k + GH->cctk_ash[2] * l)));
  }

The code crashes when CCTK_DEBUG because cctk_lsh[3] is 0, so the conditional in line 7 is not satisfied.

To reproduce, run magnetizedTOV.parin IllinoisGRMHD with CCTK_DEBUG enabled.

[Roland Haas]

Still? Grr. I thought that had been fixed in #2582 but clearly this was missed and also missed in #2394.
Is this the only place or are you aware of others? What should be used is CCTK_VECTGFINDEX3D (curernlty at http://einsteintoolkit.org/referencemanual/ReferenceManual.html#x1-226000A2)

[Gabriele Bozzola]

Let me double check what version we are running, maybe it was fixed and I rushed into reporting it!

[Roland Haas]

Oh, it still is in the current version (master, git hash 0cb4d23 "Baikal & BaikalVacuum codegen: Add pip install sympy==1.11 to the automated codegen instructions" of wvuthorns). We (@Zach Etienne and I) just missed it in the last round of removing CCTK_GFINDEX4D and also missed it when handling a slightly different issue in ID_converter_ILGRMHD that actually showed lines with CCTK_GFINDEX4D in the ticket description.

[Gabriele Bozzola]

Yes, we were running an older version and sorry for opening a duplicated ticket. The problem is now only in wvuthorns_diagnostics

particle_tracerET/src/compute_particles_four_velocities.C:          const double vU[3] = { vel[CCTK_GFINDEX4D(cctkGH,i,j,k,0)],
particle_tracerET/src/compute_particles_four_velocities.C:                                 vel[CCTK_GFINDEX4D(cctkGH,i,j,k,1)],
particle_tracerET/src/compute_particles_four_velocities.C:                                 vel[CCTK_GFINDEX4D(cctkGH,i,j,k,2)] };
particle_tracerET/src/convert_to_MHD_3velocity.C:       double ETvx = vel[CCTK_GFINDEX4D(cctkGH,i,j,k,0)];
particle_tracerET/src/convert_to_MHD_3velocity.C:       double ETvy = vel[CCTK_GFINDEX4D(cctkGH,i,j,k,1)];
particle_tracerET/src/convert_to_MHD_3velocity.C:       double ETvz = vel[CCTK_GFINDEX4D(cctkGH,i,j,k,2)];
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,0)] = -(y[index] + 0.5*dY)*A_b*pow(max(Pressure_at_Ax_stagger-P_cut,0.0),n_s);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          if(!have_two_NSs_along_x_axis) Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)] =  (xL + 0.5*dX)*A_b*pow(max(Pressure_at_Ay_stagger-P_cut,0.0),n_s);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:            if(r1<=r_NS1) Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)] =  (x1 + 0.5*dX)*A_b*pow(max(Pressure_at_Ay_stagger-P_cut,0.0),n_s);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:            if(r2<=r_NS2) Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)] =  (x2 + 0.5*dX)*A_b*pow(max(Pressure_at_Ay_stagger-P_cut,0.0),n_s);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:            if(r1>r_NS1 && r2>r_NS2) Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)] = 0.0; // No external B-field.
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,0)] = - (yL + 0.5 * dY) * (Ap1 + Ap2);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)] = (x1 + 0.5 * dX) * Ap1 + (x2 + 0.5 * dX) * Ap2;
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,2)] =  0.0;
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,0)] = -y[index]*A_b*pow(max(PL-P_cut,0.0),n_s);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)] =  x[index]*A_b*pow(max(PL-P_cut,0.0),n_s);
Seed_Magnetic_Fields_BNS/src/Seed_Magnetic_Fields.C:          Avec[CCTK_GFINDEX4D(cctkGH,i,j,k,2)] =  0.0;
smallbPoynET/src/compute_bi_b2_Poyn_fluxET.C:   double ETvx = vel[CCTK_GFINDEX4D(cctkGH,i,j,k,0)];
smallbPoynET/src/compute_bi_b2_Poyn_fluxET.C:   double ETvy = vel[CCTK_GFINDEX4D(cctkGH,i,j,k,1)];
smallbPoynET/src/compute_bi_b2_Poyn_fluxET.C:   double ETvz = vel[CCTK_GFINDEX4D(cctkGH,i,j,k,2)];
smallbPoynET/src/compute_bi_b2_Poyn_fluxET.C:   double Bx_center = Bvec[CCTK_GFINDEX4D(cctkGH,i,j,k,0)];
smallbPoynET/src/compute_bi_b2_Poyn_fluxET.C:   double By_center = Bvec[CCTK_GFINDEX4D(cctkGH,i,j,k,1)];
smallbPoynET/src/compute_bi_b2_Poyn_fluxET.C:   double Bz_center = Bvec[CCTK_GFINDEX4D(cctkGH,i,j,k,2)];

‌

EDIT:

‌

Oh, it still is in the current version (master, git hash 0cb4d23 "Baikal & BaikalVacuum codegen: Add pip install sympy==1.11 to the automated codegen instructions" of wvuthorns). We (@Zach Etienne and I) just missed it in the last round of removing CCTK_GFINDEX4D and also missed it when handling a slightly different issue in ID_converter_ILGRMHD that actually showed lines with CCTK_GFINDEX4D in the ticket description.

The message above was referring to the master of the repo

[Roland Haas]

yup. I had not updated my own repo. Sorry about the confusion. particle_tracerET is actually a newly submitted module in the Schwarzschild release. Anyhow, assigned to @Zach Etienne .

[Samuel Cupp]

I pushed a commit to wvuthorns_diagnostics to change the CCTK_GFINDEX4D to CCTK_VECTGFINDEX3D for Seed_Magnetic_Fields_BNS, particle_tracerET, and smallbPoynET. If there are any issues, let me know.

[Samuel Cupp]

I had missed one file in particle_tracerET, but I have pushed that as well now. I am closing the ticket since greping for CCTK_GFINDEX4D gets nothing in wvuthorns and wvuthorns_diagnostics.