I think we should improve the fenics interface to snes to throw an error if the user tries to set SNESVINEWTONSSLS or other methods designed for bound constrained problems when bounds are not set. May you file a bug report, possibly with a test code? I will try to fix it next week.
Best, Corrado Corrado Maurini > Le 4 oct. 2013 à 15:01, Jed Brown <[email protected]> a écrit : > > Jan Blechta <[email protected]> writes: > >> Hi Corrado and FEniCS list, >> >> I'm just looking onto option of using semi-smooth Newton. It seems >> being implemented as SNESVINEWTONSSLS [1] in PETSc. As I understand it, >> it should work without specifying any bounds (letting them +/-INF). > > How are you setting them? Note that there is no point using a > semi-smooth method without any bounds. It's just more complicated (and > much less tested) code without offering anything over the unconstrained > SNES types. > >> Nevertheless it raises >> >> [0]PETSC ERROR: Object is in wrong state! >> [0]PETSC ERROR: This DM is incapable of computing variable bounds.! > > Please send the ENTIRE error message. > >> when calling solve() without bounds. Is it possible bug in >> PETScSNESSolver.cpp? Tested with PETSc 3.4.0. >> >> [1]http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/SNES/SNESVINEWTONSSLS.html >> >> Jan >> _______________________________________________ >> fenics mailing list >> [email protected] >> http://fenicsproject.org/mailman/listinfo/fenics _______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
