Re: octave 5.1.90 make test "fail"

[Kai Torben Ohlhus]

On 1/18/20 9:40 AM, Christian Himpe wrote:
> Dear Developers,
> 
> Testing Octave 5.1.90 via make test, I got a FAIL for
> "camlookat.m"; from the "fntest.log":
> 
>>>>>> processing 
>>>>>> /home/ch/BUILA/octave-5.1.90/scripts/plot/appearance/camlookat.m
> ***** test
>  hf = figure ("visible", "off");
>  unwind_protect
>    [x y z] = sphere ();
>    h1 = surf (x + 1, y, z);
>    hold on
>    h2 = surf (x - 1, y + 2, z + 1);
>    dir = camtarget () - campos ();
>    dir /= norm (dir);
>    camlookat (h1);
>    dir2 = camtarget () - campos ();
>    dir2 /= norm (dir2);
>    assert (dir, dir2, -2*eps);
>    camlookat (h2);
>    dir2 = camtarget () - campos ();
>    dir2 /= norm (dir2);
>    assert (dir, dir2, -2*eps)
>    camlookat ([h1 h2]);
>    dir2 = camtarget () - campos ();
>    dir2 /= norm (dir2);
>    assert (dir, dir2, -2*eps);
>  unwind_protect_cleanup
>    close (hf);
>  end_unwind_protect
> !!!!! test failed
> ASSERT errors for:  assert (dir,dir2,-2 * eps)
> 
>   Location  |  Observed  |  Expected  |  Reason
>     (2)        0.72803      0.72803      Rel err 4.5749e-16 exceeds tol
>   4.4409e-16 by 1e-17
> 
> 
> Maybe the relative error needs to be raised by a small bit here?
> 
> Best
> 
> Christian
> 

Thank you for reporting this.  I guess you are using some Linux, right?
 Is there anything specific about your system setup what might cause
this tiny difference?

Does this also happen for 5.1.0 or the current development version?
Does this test always fail when explicitly call it several times from
the GUI and CLI by

   test camlookat

I think 2*eps or 4*eps is just cosmetic and we should not change for
5.2.0.  But if it occurs in Octave 6, can you please file a bug report
about it?

Thanks,
Kai