[Insight-developers] ITK 3.12 : Dashboard State : Remaining Issues.
Luis Ibanez
luis.ibanez at kitware.com
Fri Feb 27 11:27:37 EST 2009
Gaëtan Lehmann wrote:
>
> Hi Luis,
>
> Le 27 févr. 09 à 15:27, Luis Ibanez a écrit :
>
>> 17) pixel2.jouy.inra.fr
>> opensolaris-gcc3.4 ctest2.4-patch 8
>>
>> Many warnings from Tcl wrapping...
>> (??)
>>
>
> These warnings are not specific to Tcl. The problem comes from
> opensolaris, according to
>
> http://opensolaris.org/os/community/on/flag-days/pages/2008120801/
>
> Pixel2 runs svn_107 however, so, according to the link above, the
> warning should have disappeared.
> We'll try to investigate that in the next weeks.
>
>>
>> 18 Failing tests
>>
>> 14 Related to the KdTree
>> 3 Related to OptImageToImage
>> 1 ImportContainerTest
>
>
> The code should be valid anyway, so I guess those errors comes from a
> real problem in ITK on opensolaris.
>
> Gaëtan
>
>
------------
Hi Gaetan,
Thanks for looking at the build on opensolaris.
From the test output (below),
it looks like we are dealing with a numerical precision problem.
It may be just a matter of relaxing the tolerance used for comparing
the positions of points.
Currently we are using:
if( min_dist < result_dist )
maybe we should drop a bit of tolerance like:
if( min_dist + vnl_math::eps < result_dist )
I'm not sure we want to do this today,
just before the release...
Luis
-------- OUTPUT ------------------
Test output
Problem found
Query point [0.989496, -0.443952]
Reported closest point [0.858836, -0.494721] distance 0.140176
Actual closest point [0.858836, -0.494721] distance 0.140176
Test FAILED.
Problem found
Query point [1.08861, -1.21899]
Reported closest point [1.49566, -1.19854] distance 0.407558
Actual closest point [1.49566, -1.19854] distance 0.407558
Test FAILED.
Problem found
Query point [-1.59813, -0.346713]
Reported closest point [-1.49922, -0.375315] distance 0.102966
Actual closest point [-1.49922, -0.375315] distance 0.102966
Test FAILED.
3 out of 100 points failed to find the correct closest point.
--------------
More information about the Insight-developers
mailing list