[petsc-dev] [petsc-maint] Complementarity problems: vinewton?sls or tao_?s?ls

[Jason Sarich]

Hi Patrick,

I did try out the patch, thanks. I'll push it into next, but I want to
tweak it a bit. I'd feel more comfortable if we caught only the specific
error that petsc4py is throwing when getsubmatrix isn't implemented. And
the patch breaks my ex54mf test problem, there's a logic hole somewhere.

Jason


On Tue, Mar 24, 2015 at 1:25 PM, Patrick Farrell <
patrick.farrell at maths.ox.ac.uk> wrote:

> On 12/03/15 19:42, Patrick Farrell wrote:
> > I know this suggestion isn't the cleanest, but could Jason's code
> > try mat->ops->getsubmatrix, and if that fails fall back to the
> > shell matrix code?
>
> Given that (a) no one objected and (b) it solves the problem, I've
> implemented
> this suggestion in pefarrell/feature-submatrixfree-fallback of
> https://bitbucket.org/pefarrell/petsc.
>
> I didn't make a pull request because for some reason the pull request shows
> tonnes of changes I had nothing to do with (I branched off Jason's branch,
> made the fix and pushed, but bitbucket seems to be confused). Jason, could
> you
> please review/cherry-pick 13006bad66cf92787ea508665d0b39f5c9b8d46f ?
>
> I'd really appreciate getting all of this stuff into next as soon as
> possible.
> It's been floating around in branches for months and I fear bitrot is
> setting
> in. My code has depended on this for months and it all works fine.
>
> Thanks,
>
> Patrick
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20150330/9dc0c683/attachment.html>