[SciPy-Dev] Powell failure on MingW windows build - any insights?

[Matthew Brett]

Hi,

On Tue, Apr 22, 2014 at 1:08 PM, Pauli Virtanen <pav at iki.fi> wrote:
> 22.04.2014 22:51, Matthew Brett kirjoitti:
> [clip]
>> It seems to me the iteration path test is not a good test.  It's
>> likely that these kind of tiny differences will lead to divergent
>> paths, and testing that steps 34 though 38 are the same seems
>> arbitrary and difficult to justify.
>>
>> I think that test should be removed, leaving the check for the same
>> solution and roughly the same number of iterations.  Is that
>> reasonable?
>
> The point of those tests is to safeguard against unintended changes
> when refactoring the codes. An error added e.g. in line search code
> can still make the code converge, and just checking iteration counts
> in an easy test does not necessarily reveal regressions.
>
> Having the check that far (34 evals) in the tail may however be
> excessive. The other routines check earlier only after a few
> iterations, so if you can see how much of the trace is reproducible,
> that should be OK.

The course is the same through iteration 14 - is that acceptable as a length?

Cheers,

Matthew