Re: make check failure from current sources

[Jaroslav Hajek]

On Sun, Nov 2, 2008 at 9:35 AM, Thomas Treichl <address@hidden> wrote:
> Ben Abbott schrieb:
>>
>> No. The problem remains. I'll be unusually busy for the next several days
>> (at the least), but will try to dig a bit deeper into this problem.
>>
>> If anyone else is building the developer's sources on Mac OSX, please let
>> me/us know if you see this problem as well.
>
> Hi Ben,
>
> the problem is there (I also already tried Jaroslav's changeset too) - and
> if I comment out the first test of dblquad.m and then run 'make check' again
> then there appear some new failures that have not been there on my machine
> before. I cut the messages out of the fntests.log file:
>
>>>>>> processing
>>>>>> /Users/Thomas/Development/octave/scripts/geometry/griddata3.m
>  ***** testif HAVE_QHULL
>  rand('state', 0);
>  x = 2 * rand(1000, 1) - 1;
>  y = 2 * rand(1000, 1) - 1;
>  z = 2 * rand(1000, 1) - 1;
>  v = x.^2 + y.^2 + z.^2;
>  [xi, yi, zi] = meshgrid (-0.8:0.2:0.8);
>  ##vi = reshape (griddatan([x(:), y(:), z(:)], v, [xi(:), yi(:), zi(:)],
> 'linear'), size (xi));
>  vi = griddata3 (x, y, z, v, xi, yi, zi, 'linear');
>  vv = vi - xi.^2 - yi.^2 - zi.^2;
>  assert (max(abs(vv(:))), 0, 0.1)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.  ***** testif HAVE_QHULL
>  rand('state', 0);
>  x = 2 * rand(1000, 1) - 1;
>  y = 2 * rand(1000, 1) - 1;
>  z = 2 * rand(1000, 1) - 1;
>  v = x.^2 + y.^2 + z.^2;
>  [xi, yi, zi] = meshgrid (-0.8:0.2:0.8);
>  ##vi = reshape (griddatan([x(:), y(:), z(:)], v, [xi(:), yi(:), zi(:)],
> 'linear'), size (xi));
>  vi = griddata3 (x, y, z, v, xi, yi, zi, 'nearest');
>  vv = vi - xi.^2 - yi.^2 - zi.^2;
>  assert (max(abs(vv(:))), 0, 0.1)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.
>>>>>> processing
>>>>>> /Users/Thomas/Development/octave/scripts/geometry/griddatan.m
>  ***** testif HAVE_QHULL
>  [xx,yy] = meshgrid(linspace(-1,1,32));
>  xi = [xx(:), yy(:)];
>  x = (2 * rand(100,2) - 1);
>  x = [x;1,1;1,-1;-1,-1;-1,1];
>  y = sin(2*(sum(x.^2,2)));
>  zz = griddatan(x,y,xi,'linear');
>  zz2 = griddata(x(:,1),x(:,2),y,xi(:,1),xi(:,2),'linear');
>  assert (zz, zz2, 1e-10)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.
>>>>>> processing /Users/Thomas/Development/octave/scripts/geometry/rectint.m
>>>>>> processing
>>>>>> /Users/Thomas/Development/octave/scripts/geometry/tsearchn.m
>  ***** test
>  [idx, p] = tsearchn (x,tri,[-1,-1]);
>  assert (idx, 1)
>  assert (p, [1,0,0], 1e-12)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.shared variables {
>  x =
>
>    -1  -1
>    -1   1
>     1  -1
>
>  tri =
>
>     1   2   3
>
> }
>  ***** test
>  [idx, p] = tsearchn (x,tri,[-1,1]);
>  assert (idx, 1)
>  assert (p, [0,1,0], 1e-12)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.shared variables {
>  x =
>
>    -1  -1
>    -1   1
>     1  -1
>
>  tri =
>
>     1   2   3
>
> }
>  ***** test
>  [idx, p] = tsearchn (x,tri,[1,-1]);
>  assert (idx, 1)
>  assert (p, [0,0,1], 1e-12)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.shared variables {
>  x =
>
>    -1  -1
>    -1   1
>     1  -1
>
>  tri =
>
>     1   2   3
>
> }
>  ***** test
>  [idx, p] = tsearchn (x,tri,[-1/3,-1/3]);
>  assert (idx, 1)
>  assert (p, [1/3,1/3,1/3], 1e-12)
> !!!!! test failed
> A(I): Index exceeds matrix dimension.shared variables {
>  x =
>
>    -1  -1
>    -1   1
>     1  -1
>
>  tri =
>
>     1   2   3
>
> }
>>>>>> processing
>>>>>> /Users/Thomas/Development/octave/scripts/miscellaneous/fullfile.m
>  ***** assert (fullfile (""), "")
> !!!!! test failed
> A(I): Index exceeds matrix dimension.shared variables {
>  fs = /
>  fsx = /x
>  xfs = x/
>  fsxfs = /x/
>  xfsy = x/y
> }
>
>
>
>
>

Hi Thomas,

did you also have this one when you tested?
http://hg.savannah.gnu.org/hgweb/octave/rev/f7d44b6a74df

If yes, then I'm out of ideas what can be wrong. If you are willing to
digg into this, I could send you some patches with debug traceprints
that may help reveal where the problem happens.

regards


-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz