Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Segmentation test fails on ppc64le during build on Alpine #2288

Closed
mksully22 opened this issue Apr 20, 2018 · 2 comments
Closed

Segmentation test fails on ppc64le during build on Alpine #2288

mksully22 opened this issue Apr 20, 2018 · 2 comments

Comments

@mksully22
Copy link

On Alpine 3.7 building PCL version 1.8.1 using compiler g++-6.4.0-r7 the package builds successfully however during the "make tests" the following test fails:
Test 32 failing
32: Test command: /home/buildozer/aports/testing/pcl/src/pcl-pcl-1.8.1/build/test/segmentation/test_non_linear "/home/buildozer/aports/testing/pcl/src/pcl-pcl-1.8.1/test/noisy_slice_displaced.pcd"
32: Test timeout computed to be: 10000000
32: [==========] Running 1 test from 1 test case.
32: [----------] Global test environment set-up.
32: [----------] 1 test from SACSegmentation
32: [ RUN ] SACSegmentation.Segmentation
32: /home/buildozer/aports/testing/pcl/src/pcl-pcl-1.8.1/test/segmentation/test_non_linear.cpp:76: Failure
32: The difference between static_cast (inliers->indices.size ()) and 3516 is 11, which exceeds 10, where
32: static_cast (inliers->indices.size ()) evaluates to 3505,
32: 3516 evaluates to 3516, and
32: 10 evaluates to 10.
32: [ FAILED ] SACSegmentation.Segmentation (4 ms)
32: [----------] 1 test from SACSegmentation (4 ms total)
32:
32: [----------] Global test environment tear-down
32: [==========] 1 test from 1 test case ran. (4 ms total)
32: [ PASSED ] 0 tests.
32: [ FAILED ] 1 test, listed below:
32: [ FAILED ] SACSegmentation.Segmentation

It fails in the same way for both the ppc64le and aarch64 architectures.

@taketwo
Copy link
Member

taketwo commented Sep 9, 2018

I can confirm this on Ubuntu 16.04 laptop, get exactly the same output. However, the test passes both on Travis and AppVeyor.

@taketwo
Copy link
Member

taketwo commented Sep 9, 2018

The evaluated algorithm is non-deterministic, thus the implemented test assertions are approximate, not exact. I don't think there is any fundamental reason why the chosen threshold for the number of inliers is 10 and not, say, 15. The fitted sphere looks alright and passes the tests for the coefficients. I'd therefore propose to simply increase the epsilon.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants