-J does not work on the main branch ('b in QMR contains NaNs') with MF as of cf9f3d6f

[MiCurry]

Running -J on the main branch results in an error when running the COMPUTE_J task with: b in QMR contains NaNs after the forward is computed. Normally this type of error occurs when an solnVector is not correctly distributed or transferred to a worker task.

Running git bisect I was able to determine that this bug was introduced in cf9f3d6. This error does not happen before that commit, and also does not occur on the classic branch.

I'll spend some time today to determine the cause, but I'm putting this issue here in case I am not able to resolve it quickly.

If this issue is still open, and users need to run -J, they can do so on the classic branch.

[MiCurry]

Doing some digging, by the time we call QMR in ForwardSolver, eSol is completely zero. In the classic branch it has a number of values:

ModEM/f90/3D_MT/FWD/EMsolve3D.f90

Line 309 in 40cd775

Call QMR(b, eSol, KSSiter)

Still haven't found a solution yet, but it looks like eSol is set in the SdivCorr. I might table working on this for now, and focus on finishing up the rest of the EsolnManager and return to this afterwards.