[petsc-dev] Fortran callbacks when null

[Barry Smith]

  Jed,

     It seems currently for any optional callbacks from Fortran we have this ugly code

  if (FORTRANNULLFUNCTION(d)) {
      *ierr = TSMonitorSet(*ts,ourmonitor,NULL,NULL);
    } else {
      *ierr = PetscObjectSetFortranCallback((PetscObject)*ts,PETSC_FORTRAN_CALLBACK_CLASS,&_cb.mondestroy,(PetscVoidFunction)d,mctx);
      *ierr = TSMonitorSet(*ts,ourmonitor,NULL,ourmonitordestroy);
    }

This posses multiply uglinesses 

1) the same duplicate  if () ugliness people need to write and maintain for each XXXSetFunctionXXX() routine in PETSc

2) If the user passes in, by mistake something like PETSC_NULL_SOMETHING else as the function we cannot error check for that since FORTRANNULLFUNCTION is inside an if ()

Could we refactor these so that 

a) PetscObjectSetFortranCallback() handles a NULL function (by simply not calling the function in that case) and

b) we replace the if (FORTRANNULLFUNCTION) block monstrosity by something like 

  CHKFORTRANNULLFUNCTION(d);  <<< this does error checking
  *ierr = PetscObjectSetFortranCallback((PetscObject)*ts,PETSC_FORTRAN_CALLBACK_CLASS,&_cb.mondestroy,(PetscVoidFunction)d,mctx); if (*ierr) return;
  *ierr = TSMonitorSet(*ts,ourmonitor,NULL,ourmonitordestroy);

   everywhere?

   If this looks correct to you, just let me know and I'll implement it.

  Barry




> Begin forwarded message:
> 
> From: Praveen C <cpraveen at gmail.com>
> Subject: Re: [petsc-users] Segmentation fault due to TSDestroy
> Date: March 20, 2017 at 7:46:18 AM CDT
> To: petsc-users <petsc-users at mcs.anl.gov>
> 
> It turns out the problem was with this
> 
>    call TSMonitorSet(ts, Monitor, ctx, PETSC_NULL_OBJECT, ierr); CHKERRQ(ierr)
> 
> The correct way is
> 
>    call TSMonitorSet(ts, Monitor, ctx, PETSC_NULL_FUNCTION, ierr); CHKERRQ(ierr)
> 
> Thanks
> praveen
> 
> On Sat, Mar 18, 2017 at 9:30 PM, Satish Balay <balay at mcs.anl.gov <mailto:balay at mcs.anl.gov>> wrote:
> Perhaps there is some memory corruption - you can try runnng the code with valgrind.
> 
> Satish
> 
> On Sat, 18 Mar 2017, Praveen C wrote:
> 
> > Dear all
> >
> > I get a segmentation fault when I call TSDestroy. Without TSDestroy the
> > code runs fine. I have included portion of my code below.
> >
> > subroutine runts(ctx)
> >    use userdata
> >    use comdata
> >    use mtsdata
> >    implicit none
> > #include <petsc/finclude/petsc.h90>
> >    type(tsdata) :: ctx
> >    ! Local variables
> >    integer,parameter :: h = 100 ! File id for history file
> >    TS                :: ts
> >    Vec               :: u
> >    PetscErrorCode    :: ierr
> >    external          :: RHSFunction, Monitor
> >
> >    call VecDuplicate(ctx%p%v_res, u, ierr); CHKERRQ(ierr)
> >
> >    ! Copy initial condition into u
> >    call VecCopy(ctx%p%v_u, u, ierr); CHKERRQ(ierr)
> >
> >    call TSCreate(PETSC_COMM_WORLD, ts, ierr); CHKERRQ(ierr)
> >    call TSSetProblemType(ts, TS_NONLINEAR, ierr); CHKERRQ(ierr)
> >    call TSSetRHSFunction(ts, PETSC_NULL_OBJECT, RHSFunction, ctx, ierr);
> > CHKERRQ(ierr)
> >    call TSSetInitialTimeStep(ts, 0.0, dtg, ierr); CHKERRQ(ierr)
> >    call TSSetType(ts, TSRK, ierr); CHKERRQ(ierr);
> >    call TSSetDuration(ts, itmax, tfinal, ierr); CHKERRQ(ierr);
> >    call TSSetExactFinalTime(ts, TS_EXACTFINALTIME_MATCHSTEP, ierr);
> > CHKERRQ(ierr);
> >    call TSMonitorSet(ts, Monitor, ctx, PETSC_NULL_OBJECT, ierr);
> > CHKERRQ(ierr)
> >    call TSSetSolution(ts, u, ierr); CHKERRQ(ierr)
> >    call TSSetFromOptions(ts, ierr); CHKERRQ(ierr)
> >    call TSSetUp(ts, ierr); CHKERRQ(ierr)
> >
> >    call TSSolve(ts, u, ierr); CHKERRQ(ierr)
> >
> >    call VecCopy(u, ctx%p%v_u, ierr); CHKERRQ(ierr)
> >    call VecDestroy(u, ierr); CHKERRQ(ierr)
> >    call TSDestroy(ts, ierr); CHKERRQ(ierr)
> >
> > end subroutine runts
> >
> > Thanks
> > praveen
> >
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20170321/10f2d5b8/attachment.html>