Re: quadl failure

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Re: quadl failure

Rik-5
On 12/15/2011 10:00 AM, [hidden email] wrote:
> assert (quadl (@(x) sin (3 * x) .* cosh (x) .* sinh (x), 10,
> 15),2.588424538641647e+10,-9e-15) expected
>  2.5884e+10
> but got
>  2.5884e+10
> maximum relative error 1.04636e-14 exceeds tolerance 9e-15
> octave-3.5.90:4>
This bug has been reported on the MinGW platform here
(https://savannah.gnu.org/bugs/?34942) and also here
(https://savannah.gnu.org/bugs/?33863).  The suggestion was just to
increase the tolerance on the assert comparison.  I haven't done that
immediately because I was concerned that the increase in tolerance (22%)
might be masking something fundamental.

But, if the problem is spreading to other operating systems then perhaps
the easiest approach is to increase the tolerance.  Like the reporter, I
also run on Ubuntu 10.04 and I never experience a failure.  I thought the
trouble might be relating to truncating intermediate floating point results
on the MinGW platform.

--Rik