[ITK-dev] SimpleITK 0.9.1 tests failure

[Dmitry Mikhirev]

Hello Brad!

I'm sorry for long delay, I was a bit busy..

Here are test results:

https://open.cdash.org/viewTest.php?onlyfailed&buildid=4137704

Please ignore java tests, they most probably failed because of
misconfigured environment. The same tests passed successfully during
package build.

What do you think about BasicFilters ad Python tests results? Are they
acceptable, or is it better to try fixing them (e. g. use another FFT
library)?

В Sun, 6 Dec 2015 20:42:26 -0500
Bradley Lowekamp <brad at lowekamp.net> пишет:

> Hello again,
> 
> I don’t think these are critical, but they should be investigated.
> 
> Missing numpy should have produced a warning during the build
> process. If you have it installed in the system python it should get
> copied over to the test environment.
> 
> There other failures are with the following algorithms:
> InverseDeconvolutionImageFilter
> WienerDeconvolutionImageFilter
> TikhonovDeconvolutionImageFilter 
> MaskedFFTNormalizedCorrelationImageFilter
> 
> The all are failing with tolerance issues. Meaning that the results
> are alittle different. These algorithms all use FFTs in the
> computation. What FFT library is being used? FFTW? system VXL? ITK’s
> VXL?
> 
> The tests produce images for output including a difference image.
> These should be inspected to see if it’s just noise or patterned
> which may indicate something like a boundary condition issue.
> 
> If you do a "ctest -D ExperimentalTest -D ExperimentalSubmit” the
> results will be submitted to CDash so they can be better shared.
> 
> Good luck,
> Brad
> 


-- 
Regards,
       Dmitry Mikhirev