I have made some changes to the handling of parameters in NewtonSolver and PETScSNESSolver in order to fix this bug:
https://bitbucket.org/fenics-project/dolfin/issue/10/nonlinearvariationalsolver-does-not-pass The change involves pushing the linear solver parameters down one level from NonlinearVariationalSolver to NewtonSolver and PETScSNESSolver where these parameters are actually used. Please have a look to see if the changes are ok, especially regarding PETScSNESSolver which I have never used myself: https://bitbucket.org/fenics-project/dolfin/branch/logg/fix-issue-10 The SNES demo has been updated and seems to work fine. -- Anders _______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
