I have fixed petsc-dev to never return garbage in this call. Also fixed the other MatMFFD methods to use the compose/query function paradigm
Barry On Mar 29, 2011, at 2:35 AM, Jed Brown wrote: > On Tue, Mar 29, 2011 at 08:07, Kontsantinos Kontzialis <ckontzialis at > lycos.com> wrote: > When should I call MatMFFDGeth in order to get the differencing parameter? > I call it after SNESSolve but I get zero for h. > > I suspect you are calling this function on a matrix of the wrong type (and, > because this function does not check types properly, returning nonsense bytes > that happen to represent a very small floating point value). Try calling > > ierr = SNESGetJacobian(snes,&mat,0,0,0);CHKERRQ(ierr); > ierr = MatMFFDGetH(mat,&h);CHKERRQ(ierr); >
