[Insight-developers] gcc 4.3 test failures

[Bill Lorensen]

Niels,

The values should be identical. They are derived from the same source
image (random) and filter. The only difference is an extra
StreamingImageFilter. There are about 10 tests failing with the same
issue. I think it is a buggy compiler. We saw a similar issue with
itkRegionGrow2DTest a week ago. In that case, the numbers were
integral and the equality test failed. After adding a bunch of extra
output, the test started passing. This is the first release of gcc
4.3.0. There is a just released 4.3.1 which we have not tried yet.

Bill

On Sat, Jun 14, 2008 at 10:48 AM, Niels Dekker <niels-xtk at xs4all.nl> wrote:
> Bill Lorensen wrote:
>>
>> Luis,
>> I added some additional output to one of the gcc4.3 failing tests.
>> http://www.cdash.org/CDash/testDetails.php?test=4748940&build=99510
>> As you can see from the output, the test seems to fail even though the
>> two variables are equal. Have you had a chance to upgrade to the new
>> gcc 4.3?
>
> cdash.org/CDash/testDetails.php?test=4748940&build=99510 says:
>
>  Compare stand-alone and streamer outputs
>  it1.Get() != it2.Get(): 0.207653 != 0.207653
>  Test failed.
>
> I guess the difference between the two is extremely small. You could try to
> increase the precision of std::cout, by using std::setprecision(), within
> itkCurvatureFlowTest.cxx
> http://www.cplusplus.com/reference/iostream/manipulators/setprecision.html
>
> And maybe display the difference as well: it1.Get()- it2.Get().
>
> Just my 2 cents,
>
> Kind regards,
> --
> Niels Dekker
> http://www.xs4all.nl/~nd/dekkerware
> Scientific programmer at LKEB, Leiden University Medical Center
>