BUG: No reconstruction of the normal field on JDIR if one of the boudaries is set to 'axis'

[aborderies]

Describe the issue:

Hi all,
I think there is a problem with the reconstruction of the normal field in ghost cells in the J direction, in spherical coordinates when the axis boundary is used.

idefix/src/fluid/boundary/boundary.hpp

Lines 464 to 492 in ede2e1f

	if(dir==JDIR) {
	nstart = data->beg[JDIR]-1;
	nend = data->end[JDIR];
	if(!this->haveAxis) {
	idefix_for("ReconstructBX2s",0,data->np_tot[KDIR],0,data->np_tot[IDIR],
	KOKKOS_LAMBDA (int k, int i) {
	if(reconstructLeft) {
	for(int j = nstart ; j>=0 ; j-- ) {
	Vs(BX2s,k,j,i) = 1.0 / Ax2(k,j,i) * ( Ax2(k,j+1,i)*Vs(BX2s,k,j+1,i)
	+(D_EXPAND( Ax1(k,j,i+1) * Vs(BX1s,k,j,i+1) - Ax1(k,j,i) * Vs(BX1s,k,j,i) ,
	,
	+ Ax3(k+1,j,i) * Vs(BX3s,k+1,j,i) - Ax3(k,j,i) * Vs(BX3s,k,j,i) )));
	}
	}
	if(reconstructRight) {
	for(int j = nend ; j<nx2 ; j++ ) {
	Vs(BX2s,k,j+1,i) = 1.0 / Ax2(k,j+1,i) * ( Ax2(k,j,i)*Vs(BX2s,k,j,i)
	-(D_EXPAND( Ax1(k,j,i+1) * Vs(BX1s,k,j,i+1) - Ax1(k,j,i) * Vs(BX1s,k,j,i) ,
	,
	+ Ax3(k+1,j,i) * Vs(BX3s,k+1,j,i) - Ax3(k,j,i) * Vs(BX3s,k,j,i) )));
	}
	}
	}
	);
	} else {
	// We have an axis, ask myAxis to do that job for us
	axis->ReconstructBx2s();
	}
	}

If only one of the two boudaries is set to axis, and the other is not (reflective or outflow for example), because of line 467,

if(!this->haveAxis) {

Idefix only reconstructs the axis boundary and not the other side. This causes my code to crash when I use a grid with $\theta$ ranging from 0$ to $\pi/2$ (top half of the domain) with a reflective condition at $\theta = \pi/2$ and with NIMHD resistivities that depend on the norm of the magnetic field.

Error message:

runtime information:

Idefix V2.2.01

[glesur]

Thanks for spotting this @aborderies . A fix is underway.

[glesur]

can you check that the branch https://github.com/idefix-code/idefix/tree/fixAxis fix your problem?

[aborderies]

Hi @glesur, thanks for your reply ! It seems that the branch fix the problem !

[glesur]

fixed and merged by #345