Set krylov linear solver paramters in newton solver

variational formulation
12

Looks like old DOLFIN intercepts a PETSc error and then throws its own should the KSP iterations exceed their limit. Quick solution is to handle it with python:

try:
    custom_solver.solve(problem, u.vector())
except Exception as e:
    # Handle the error within your time stepping algorithm gracefully...

Consider using a PETScSNESSolver and its underlying SNES and KSP objects to get more information about the lack of convergence. Also consider foregoing DOLFIN’s implementations of Newton and Krylov solvers altogether and just use a SNES and KSP directly.