On Aug 13, 2011, at 5:22 PM, Jed Brown wrote:
> On Sat, Aug 13, 2011 at 17:19, Barry Smith <bsmith at mcs.anl.gov> wrote:
>
> Matt,
>
> What did I break? I didn't intend to change any "short" that is used in
> testing.
>
> http://petsc.cs.iit.edu/petsc/petsc-dev/rev/db8ec2efa042
>
> The solver parameters shouldn't be a reproducibility problem because they
> generally aren't computed values, but I'd rather not see all those zeros.
These are NOT monitor routines, they are view routines. I think all the
changes are in order and were needed. Yes we have to fix our testing process so
one doesn't need to constantly change the output files.
The reason you see all those zeros is because of the damn lack of proper
support for formatting options for __float128 precision. I couldn't leave the
%G in some places because with __float128 you get garbage print out. If you
want you can change the %14.12e in the SNES view to %g but then you will need
to update the output file yuck.
Barry
>
