This bugzilla service is closed. All entries have been migrated to https://gitlab.com/libeigen/eigen
Bug 673 - Test failures
Summary: Test failures
Status: CONFIRMED
Alias: None
Product: Eigen
Classification: Unclassified
Component: Tests (show other bugs)
Version: 3.2
Hardware: All All
: Normal Failed Unit Test
Assignee: Nobody
URL:
Whiteboard:
Keywords:
Depends on: 715
Blocks:
  Show dependency treegraph
 
Reported: 2013-10-08 22:37 UTC by Orion Poplawski
Modified: 2019-12-04 12:40 UTC (History)
4 users (show)



Attachments

Description Orion Poplawski 2013-10-08 22:37:06 UTC
We are seeing the following test failures on Fedora with 3.2 on 32-bit:

fedora-18-i386/result/build.log
346/651 Test #346: schur_complex_2 ..................***Failed    0.08 sec
556/651 Test #556: cholmod_support_1 ................***Failed    2.95 sec
557/651 Test #557: cholmod_support_2 ................***Failed   12.33 sec
633/651 Test #633: gmres_1 ..........................***Failed    0.28 sec
635/651 Test #635: minres_1 .........................***Failed    0.01 sec
637/651 Test #637: levenberg_marquardt ..............***Failed    0.01 sec
650/651 Test #650: bdcsvd_2 .........................***Failed    0.01 sec
fedora-19-i386/result/build.log
177/651 Test #177: ref_1 ............................***Failed    0.16 sec
556/651 Test #556: cholmod_support_1 ................***Failed    3.63 sec
557/651 Test #557: cholmod_support_2 ................***Failed   22.74 sec
570/651 Test #570: NonLinearOptimization ............***Failed    0.09 sec
585/651 Test #585: matrix_function_1 ................***Failed    0.03 sec
635/651 Test #635: minres_1 .........................***Failed    0.05 sec
637/651 Test #637: levenberg_marquardt ..............***Failed    0.04 sec
650/651 Test #650: bdcsvd_2 .........................***Failed    0.01 sec
fedora-20-i386/result/build.log
177/653 Test #177: ref_1 ............................***Failed    0.10 sec
556/653 Test #556: cholmod_support_1 ................***Failed    2.96 sec
557/653 Test #557: cholmod_support_2 ................***Failed   16.85 sec
572/653 Test #572: NonLinearOptimization ............***Failed    0.02 sec
635/653 Test #635: gmres_1 ..........................***Failed    0.22 sec
637/653 Test #637: minres_1 .........................***Failed    0.01 sec
639/653 Test #639: levenberg_marquardt ..............***Failed    0.01 sec
652/653 Test #652: bdcsvd_2 .........................***Failed    0.01 sec
fedora-rawhide-i386/result/build.log
177/653 Test #177: ref_1 ............................***Failed    0.10 sec
556/653 Test #556: cholmod_support_1 ................***Failed    4.90 sec
557/653 Test #557: cholmod_support_2 ................***Failed   29.86 sec
572/653 Test #572: NonLinearOptimization ............***Failed    0.13 sec
601/653 Test #601: matrix_power_8 ...................***Failed    0.06 sec
636/653 Test #636: gmres_2 ..........................***Failed    1.09 sec
637/653 Test #637: minres_1 .........................***Failed    0.30 sec
639/653 Test #639: levenberg_marquardt ..............***Failed    0.18 sec
652/653 Test #652: bdcsvd_2 .........................***Failed    0.08 sec

I'm trying to track down a very strange issue with matrix multiplication in gdl (which is using eigen) and figured that looking at the eigen tests might be a good place to start.  Anything particularly concerning here?
Comment 1 Christoph Hertzberg 2013-10-15 19:37:16 UTC
I don't think any of these are related to matrix multiplication. It might be that gdl also uses Ref<> which then indirectly causes problems. And Ref definitely should not fail (I think all others are in unsupported, so some of them might still fail occasionally).
Comment 2 Gael Guennebaud 2013-10-28 18:01:19 UTC
Indeed, these failures cannot explain a problem related to matrix multiplication. Nevertheless, I cannot reproduce the few "official" ones that are failing.
Could you show the output of:

./ref_1
./schur_complex_2
./cholmod_support_1
./cholmod_support_2

thanks.
Comment 3 Benoit Steiner 2014-01-30 21:26:43 UTC
I have also noticed that the gmres_1 test fails for seed 1391110414 using the latest version of the code base:
./unsupported/test/gmres_1 s1391110414
Initializing random number generator with seed 1391110414
Repeating each test 10 times
gmres_1: /usr/local/google/home/bsteiner/EigenLatest/eigen/Eigen/src/Core/Assign.h:520: Derived& Eigen::DenseBase<Derived>::lazyAssign(const Eigen::DenseBase<OtherDerived>&) [with OtherDerived = Eigen::Matrix<double, -0x00000000000000001, 1>, Derived = Eigen::Block<Eigen::Matrix<double, -0x00000000000000001, -0x00000000000000001>, -0x00000000000000001, -0x00000000000000001, false>]: Assertion `rows() == other.rows() && cols() == other.cols()' failed.
Aborted (core dumped)
Comment 4 Orion Poplawski 2014-02-05 16:57:30 UTC
Sorry for the extreme delay:

$ ./ref_1
Initializing random number generator with seed 1391551109
Repeating each test 10 times

    actual   = 0.5835
    expected = 0.5835

Test ref_matrix(Matrix<float, 1, 1>()) failed in /home/orion/fedora/eigen3/eigen-eigen-ffa86ffb5570/test/ref.cpp (83)
    test_is_equal(m1, m2)
Stack:
  - ref_matrix(Matrix<float, 1, 1>())
  - ref
  - Seed: 1391551109

Aborted

$ ./schur_complex_2
Initializing random number generator with seed 1391551177
Repeating each test 10 times

(so maybe that is working now/sometimes)

$ ./cholmod_support_1
Initializing random number generator with seed 1391551229
Repeating each test 10 times
Test test_cholmod_T<double>() failed in /home/orion/fedora/eigen3/eigen-eigen-ffa86ffb5570/test/sparse_solver.h (39)
    x.isApprox(refX,test_precision<Scalar>())
Stack:
  - test_cholmod_T<double>()
  - cholmod_support
  - Seed: 1391551229

Aborted

$ ./cholmod_support_2
Initializing random number generator with seed 1391551235
Repeating each test 10 times
Test test_cholmod_T<std::complex<double> >() failed in /home/orion/fedora/eigen3/eigen-eigen-ffa86ffb5570/test/sparse_solver.h (39)
    x.isApprox(refX,test_precision<Scalar>())
Stack:
  - test_cholmod_T<std::complex<double> >()
  - cholmod_support
  - Seed: 1391551235

Aborted
Comment 5 Christoph Hertzberg 2014-06-19 14:46:47 UTC
The test_is_equal for Ref could actually fail when using x87-math if one matrix is still partially on the fpu-stack and not stored to memory before comparison.
A fix would be to replace the test by is_approx, or to hide the comparison into a non-inlined function.
Bug 715 duplicates some of the other failures.
Comment 6 Nobody 2019-12-04 12:40:41 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to gitlab.com's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.com/libeigen/eigen/issues/673.

Note You need to log in before you can comment on or make changes to this bug.