[Debian-med-packaging] Bug#914817: Bug#914817: camitk: FTBFS on i386

Emmanuel Promayon Emmanuel.Promayon at univ-grenoble-alpes.fr
Fri Nov 30 06:42:04 GMT 2018


X-Debbugs-Cc: Gert Wollny <gw.fossdev at gmail.com>


Thank you Mattia for this report.
I rebuild locally and got the same two tests failing.

After a first investigation, I think Gert is probably right about this "-ffloat-store" or at least about floating point precision.

For instance, for test #46, which complains about a difference between expected output and actual output:

diff .../camitk-build/Testing/Temporary/action-reconstruction-integration-test/output-1.vtk ...
/camitk-build/Testing/Temporary/action-reconstruction-integration-test/2018-11-29T16-54-57/output-1.vtk

gives

3626c3626
< 0.804485 -0.417868 -0.422127 0.224294 -0.343943 -0.911809 -0.567178 -0.335857 -0.752004
---
> 0.804485 -0.417868 -0.422127 0.224294 -0.343943 -0.911809 -0.567178 -0.335857 -0.752003 
3632c3632
< -0.346881 0 -0.937909 -0.150485 -0.0767808 -0.985626 -0.458605 -0.843355 -0.280059
---
> -0.346881 0 -0.937909 -0.150485 -0.0767809 -0.985626 -0.458605 -0.843355 -0.280059 
3635c3635
< 0.13497 -0.236088 -0.962313 -0.487965 -0.150704 -0.859755 -0.484595 -0.281462 -0.828219
---
> 0.13497 -0.236088 -0.962313 -0.487964 -0.150704 -0.859755 -0.484595 -0.281462 -0.828219 

Note the floating point precision difference.

In both cases the vtk class that is used is the marching cube reconstruction algorithm.

May be the "-ffloat-store" cxx flag is not added in VTK (or more specically is just erased somewhere just for the marching cube reconstruction algorithm, as other filters seem to not have this problem?

I will check a little bit further to see if the problem is in camitk or vtk7

Best regards,
Emmanuel



On 27/11/2018 17:48, Mattia Rizzolo wrote:
> Source: camitk
> Version: 4.1.2-1
> Severity: serious
> X-Debbugs-Cc: Gert Wollny <gw.fossdev at gmail.com>
>
> Hi,
>
> so camitk FTBFS on i386; I don't know why it sometimes suceeds on the
> buildds, but for example on the r-b infra it always failed recently.
>
> In particular, it fails two tests:
>
> 47: Test command: /usr/bin/cmake "-DCAMITK_TEST_COMMAND=/build/camitk-4.1.2/camitk-build/bin/camitk-actionstatemachine" "-DCAMITK_TEST_COMMAND_ARG=-f /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test/asm-input.scxml -o /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test -a" "-DCAMITK_TEST_EXPECTED_FILES=output-1.vtk" "-DCAMITK_TEST_OUTPUT_DIR=/build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test" "-DCAMITK_TEST_NAME=action-reconstruction-integration-test" "-P" "/build/camitk-4.1.2/sdk/cmake/modules/macros/camitk/test/CamiTKTestActionStateMachine.cmake"
> 47: Test timeout computed to be: 1800
> 47: QStandardPaths: XDG_RUNTIME_DIR points to non-existing path '/run/user/1000', please create it with 0700 permissions.
> 47: [FAIL]
> 47: CMake Error at /build/camitk-4.1.2/sdk/cmake/modules/macros/camitk/test/CamiTKTestActionStateMachine.cmake:115 (message):
> 47:   action-reconstruction-integration-test: (one or more) output file do(es)
> 47:   not match the corresponding expected file.
> 47:
> 47:   Action state machine regression test called with:
> 47:
> 47:    - CAMITK_TEST_NAME=action-reconstruction-integration-test
> 47:    - CAMITK_TEST_COMMAND=/build/camitk-4.1.2/camitk-build/bin/camitk-actionstatemachine
> 47:    - CAMITK_TEST_COMMAND_ARG=-f /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test/asm-input.scxml -o /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test -a
> 47:    - CAMITK_TEST_EXPECTED_FILES=output-1.vtk
> 47:    - CAMITK_TEST_OUTPUT_DIR=/build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test
> 47:
> 47:   =================================================
> 47:
> 47:   Comparing
> 47:   /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test/2018-11-27T16-12-36/output-1.vtk
> 47:   with
> 47:   /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test/output-1.vtk
> 47:
> 47:
> 47:   Output:
> 47:
> 47:
> 47:
> 47:   Result:
> 47:
> 47:   action-reconstruction-integration-test: comparing output-1.vtk failed:
> 47:
> 47:   Output file
> 47:   /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test/output-1.vtk
> 47:   is not the same as
> 47:   /build/camitk-4.1.2/camitk-build/Testing/Temporary/action-reconstruction-integration-test/2018-11-27T16-12-36/output-1.vtk
> 47:
> 47:
> 47:
> 47:
> 47:
> 47:
>   47/550 Test  #47: action-reconstruction-integration-test ......................***Failed    3.00 sec
>
>
> and similarly for test 341.
>
>
> I've tried the proposed change by Gert, i.e. passing -ffloat-store, but
> it didn't help.
>
>
> _______________________________________________
> Debian-med-packaging mailing list
> Debian-med-packaging at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-med-packaging

-- 
Emmanuel Promayon
Professeur Univ. Grenoble Alpes / Polytech Grenoble
Laboratoire TIMC-IMAG / équipe GMCAO
Responsable du département Technologies de l'Information pour la Santé


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20181130/7c864867/attachment.html>


More information about the Debian-med-packaging mailing list