[petsc-dev] Bug in PCSetUp_FieldSplit/PCReset_FieldSplit

[Sander Arens]

Barry,

The branch resolves the difficulties. Sorry for the sloppy code, it was
just a quick fix to illustrate something. Had I known you were going to use
it as a patch I would have tested it more thoroughly.

Thanks,
Sander

On 26 November 2016 at 06:02, Barry Smith <bsmith at mcs.anl.gov> wrote:

>
>    Sander,
>
>     Thanks for the patch, I have put it into the branch
> barry/Sander/fix-pcreset-fieldsplit and confirmed that it does
> support reset (after fixing some bugs) and merged it into next for
> testing. Can you try it and see if it resolves the difficulties you were
> having?
>
>    Barry
>
> Before submitting patches you should try the test codes with -malloc_test
> or under valgrind, you would have found some memory leaks due to bugs in
> your patch and from those found the bugs in the patch you sent.
>
>
> > On Nov 22, 2016, at 3:10 AM, Sander Arens <Sander.Arens at UGent.be> wrote:
> >
> > It works now for the particular case when there's a SNESReset before a
> SNESSolve, but when you put a SNESReset and SNESSolve again after the first
> SNESSolve it breaks again (I noticed this first  playing around with
> TSArkimex when it broke at this line).
> >
> > Destroying the fieldsplit linked list in PCReset_FieldSplit seemed to
> work (see diff file)...
> >
> > On 21 November 2016 at 21:02, Matthew Knepley <knepley at gmail.com> wrote:
> > Okay, try out the fix in next.
> >
> >   Thanks,
> >
> >     Matt
> >
> > On Thu, Nov 17, 2016 at 3:11 PM, Matthew Knepley <knepley at gmail.com>
> wrote:
> > On Thu, Nov 17, 2016 at 1:22 PM, Sander Arens <Sander.Arens at ugent.be>
> wrote:
> > I think there's a bug in PCSetUp_FieldSplit or PCReset_FieldSplit. I can
> get this by putting a SNESReset before the SNESSolve in snes ex77 and
> running ./config/builder2.py check src/snes/examples/tutorials/ex77.c.
> Can somebody reproduce this?
> >
> > You are right. Fixing...
> >
> >    Matt
> >
> > Thanks,
> > Sander
> >
> >
> >
> > --
> > What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> > -- Norbert Wiener
> >
> >
> >
> > --
> > What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> > -- Norbert Wiener
> >
> > <fieldsplit.diff>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20161128/07e26fc7/attachment.html>