> On Feb 6, 2015, at 9:28 PM, Jed Brown <j...@jedbrown.org> wrote: > > Barry Smith <bsm...@mcs.anl.gov> writes: >> Nonsense. They could always put the initial conditions in an >> appropriate vector and the callback just be a vec copy, then you get >> back the effect of TSSet"Initial"Solution() > > How is that simpler or clearer than passing it as an argument to > TSSolve? I didn't say it's capital-H "Hard", just that ugly and > needless complexity.
I am not proposing removing the solution from TSSolve() argument list, far from it. I am proposing removing the TSSetSolution(), keeping the solution in TSSolve() as the standard approach, but also allowing a callback for when the user wants the DM passed to TS to be in charge of creating all the vectors.