Re: octave-image tests

[Carnë Draug]

On 31 July 2015 at 23:03, Orion Poplawski <address@hidden> wrote:
>
>>> pkg load image
>>> runtests('/usr/share/octave/packages/image-2.4.0');
> Processing files in /usr/share/octave/packages/image-2.4.0:
>
>
>   bwmorph.m .............................................. PASS   11/13   
> FAIL 2
>
>>> test('/usr/share/octave/packages/image-2.4.0/bwmorph.m','verbose');
>
> ***** test
>  in  = logical ([0  1  0  0  0
>                  1  0  0  1  0
>                  1  0  1  0  0
>                  1  1  1  1  1
>                  1  1  1  1  1]);
>
>  out = logical ([0  1  0  0   0
>                  1  0  0  1  0
>                  1  0  1  0  0
>                  1  1  0  1  1
>                  1  1  1  1  1]);
>  assert (bwmorph (in, "remove"), out);
>
>  out = logical ([0  1  0  0  0
>                  1  0  0  1  0
>                  1  0  1  0  0
>                  1  1  0  1  1
>                  1  1  1  1  1]);
>  assert (bwmorph (in, "remove", Inf), out);
>
>  ## tests for spur are failing (matlab incompatible)
>  out = logical ([0  1  0  0  0
>                  1  0  0  0  0
>                  1  0  1  0  0
>                  1  1  1  1  1
>                  1  1  1  1  1]);
>  assert (bwmorph (in, "spur"), out);
>
>  out = logical ([0  1  0  0  0
>                  1  0  0  0  0
>                  1  0  0  0  0
>                  1  1  1  1  1
>                  1  1  1  1  1]);
>  assert (bwmorph (in, "spur", Inf), out);
> !!!!! test failed
> ASSERT errors for:  assert (bwmorph (in, "spur"),out)
>
>   Location  |  Observed  |  Expected  |  Reason
>    (1,2)          0            1         Abs err 1 exceeds tol 0
>
>
>   cp2tform.m ............................................. PASS    6/7    
> FAIL 1
>
>>> test('/usr/share/octave/packages/image-2.4.0/cp2tform.m','verbose');
>>>>>> /usr/share/octave/packages/image-2.4.0/cp2tform.m
> ***** xtest
>  npt = 100000;
>  [crw, cap] = coords (npt);
>  ttype = 'polynomial';
>  ord = 4;
>  T = cp2tform (crw, cap, ttype, ord);
>  crw2 = tforminv (T, cap);
>  finalerr = norm (crw - crw2)/npt;
>  assert (finalerr < eps, "norm = %3.2e ( > eps)", finalerr)
> !!!!! known failure
> norm = 5.31e-16 ( > eps)
> PASSES 6 out of 7 tests (1 expected failure)
>
> (hmm, can this get marked XFAIL?)

You mean the grayscale error below or the bwmorph and cp2tform failures above?

>   grayslice.m ............................................ PASS    7/8    
> FAIL 1
>
> I can't reproduce this failure running it directly.

That test uses random input and only happens sometimes.  If you test it in
a loop, running directly, you should see it.  I need to take a closer at it,
see if I can fix it.

>
> These all appear to be expected failures:
> [...]

Yes, they are expected in the sense that I know the bug is there.  But they
are not expected in the sense that it should be the user expected return
value.  I prefer to not mislead anyone (specially myself), that the bug is
not there by marking it as an xtest.

And thank you for packaging this for CentOS 6.

Carnë