quadl test request for Matlab

[Rik]

On 24-Aug-2011, Andreas Weber wrote:
> | Hello list,
> | 
> | one test in quadl fails since more than 3 weeks (I guess since the first
> | code sprint).
> | 
> | octave:2> test quadl
> |   ***** assert ( quadl (@(x) sin (2 + 3*x).^2, 0, 10, 0.3, []),
> |         (60 + sin(4) - sin(64))/12, -0.3)
> | !!!!! test failed
> | assert (quadl (@(x) sin (2 + 3 * x) .^ 2, 0, 10, 0.3, []),(60 + sin (4)
> | - sin (64)) / 12,-0.3) expected
> |  4.8603
> | but got
> |  8.1963
> | maximum relative error 0.686392 exceeds tolerance 0.3
> | 
> | Jordi told me in #octave that he is aware of that
>
> Yes, there seem to be two reports about this problem in the bug tracker:
>
> https://savannah.gnu.org/bugs/?33863
> https://savannah.gnu.org/bugs/?33792
>
> | so perhaps it should
> | be marked as "expected failure" until the problem is fixed?
>
> I suggested doing that in my comment to bug #33792 but Rik was
> concerned that it would be forgotten if we mark it as expected.
I was concerned.  I looked a little bit at the issue and it seems that the
algorithm itself is not robust when the tolerance is large.

In order to verify that could someone with ML access check what the
following code produces?

quadl (@(x) sin (2 + 3*x).^2, 0, 10, 0.3, [])

If it is the quadl algorithm itself, rather than just our implementation of
it, which is bad then we can go ahead and remove the test completely.

--Rik