Test failures for 0.98.4 on CentOS 7

[verdurin]

I am packaging GEMMA for EasyBuild, to aid installation on HPC clusters.
In fact, I am updating a previous version.

It builds fine on a Fedora 33 node.

However, I have seen test failures on two different CentOS 7 nodes.

Here are the failing tests:

testBslmm2
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 1940
## number of analyzed individuals = 757
## number of covariates = 1
## number of phenotypes = 1
## number of total SNPs/var        =    12226
## number of analyzed SNPs         =    10775
Calculating Relatedness Matrix ...
                                                   0%
====                                               8%
========                                           16%
============                                       25%
================                                   33%
====================                               41%
========================                           49%
============================                       57%
================================                   65%
====================================               74%
========================================           82%
============================================       90%
=================================================  98%
================================================== 100%

**** INFO: Done.
ASSERT:expected:<579.50> but was:<579.66>

testCenteredRelatednessMatrixKFullLOCO1
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 1940
## number of analyzed individuals = 1410
## number of covariates = 1
## number of phenotypes = 1
## leave one chromosome out (LOCO) =        1
## number of total SNPs/var        =    12226
## number of SNPS for K            =    11182
## number of SNPS for GWAS         =     1044
## number of analyzed SNPs         =    10768
Calculating Relatedness Matrix ...
                                                   0%
====                                               8%
========                                           16%
============                                       25%
================                                   33%
====================                               41%
========================                           49%
============================                       57%
================================                   65%
====================================               74%
========================================           82%
============================================       90%
=================================================  98%
================================================== 100%

**** INFO: Done.
ASSERT:expected:<2246.49> but was:<2246.57>

testUnivariateLinearMixedModelFullLOCO1
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 1940
## number of analyzed individuals = 1410
## number of covariates = 1
## number of phenotypes = 1
## leave one chromosome out (LOCO) =        1
## number of total SNPs/var        =    12226
## number of SNPS for K            =    11182
## number of SNPS for GWAS         =     1044
## number of analyzed SNPs         =    10768
Start Eigen-Decomposition...
pve estimate =0.593114
se(pve) =0.0332587
                                                   0%
====                                               8%
========                                           16%
============                                       25%
================                                   33%
====================                               41%
========================                           49%
============================                       57%
================================                   65%
====================================               74%
========================================           82%
============================================       90%
=================================================  98%
================================================== 100%
================================================== 100%

**** INFO: Done.
## total computation time = 0.081 min
ASSERT:expected:<267507852.03> but was:<267507851.98>
testCenteredRelatednessMatrixK
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 1940
## number of analyzed individuals = 1410
## number of covariates = 1
## number of phenotypes = 1
## number of total SNPs/var        =    12226
## number of analyzed SNPs         =    10768
Calculating Relatedness Matrix ...
                                                   0%
====                                               8%
========                                           16%
============                                       25%
================                                   33%
====================                               41%
========================                           49%
============================                       57%
================================                   65%
====================================               74%
========================================           82%
============================================       90%
=================================================  98%
================================================== 100%

**** INFO: Done.
ASSERT:expected:<1119.60> but was:<1119.64>
testUnivariateLinearMixedModel
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 1940
## number of analyzed individuals = 1410
## number of covariates = 1
## number of phenotypes = 1
## number of total SNPs/var        =    12226
## number of analyzed SNPs         =    10768
Start Eigen-Decomposition...
pve estimate =0.608801
se(pve) =0.032774
                                                   0%
====                                               8%
========                                           16%
============                                       25%
================                                   33%
====================                               41%
========================                           49%
============================                       57%
================================                   65%
====================================               74%
========================================           82%
============================================       90%
=================================================  98%
================================================== 100%
================================================== 100%

**** INFO: Done.
## total computation time = 0.216833 min
ASSERT:expected:<4038540440.81> but was:<4038540440.86>
testLinearMixedModelPhenotypes
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 1940
## number of analyzed individuals = 1197
## number of covariates = 1
## number of phenotypes = 2
## number of total SNPs/var        =    12226
## number of analyzed SNPs         =    10758
Start Eigen-Decomposition...
REMLE estimate for Vg in the null model:
1.3940
-0.2267 2.0817
se(Vg):
0.1567
0.1363  0.2359
REMLE estimate for Ve in the null model:
0.3489
0.0491  0.4144
se(Ve):
0.0206
0.0166  0.0267
REMLE likelihood = -2855.1664
MLE estimate for Vg in the null model:
1.3959
-0.2267 2.0854
se(Vg):
0.1568
0.1365  0.2361
MLE estimate for Ve in the null model:
0.3483
0.0490  0.4136
se(Ve):
0.0206
0.0166  0.0266
MLE likelihood = -2856.0280
                                                   0%
====                                               8%
========                                           16%
============                                       25%
================                                   33%
====================                               41%
========================                           49%
============================                       57%
================================                   65%
====================================               74%
========================================           82%
============================================       90%
=================================================  98%
================================================== 100%

**** INFO: Done.
ASSERT:expected:<4029037056.58> but was:<4029037056.54>
testPlinkStandardRelatednessMatrixK
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
## number of total individuals = 427
## number of analyzed individuals = 427
## number of covariates = 1
## number of phenotypes = 1
## number of total SNPs/var        =   358499
## number of analyzed SNPs         =   273349
Calculating Relatedness Matrix ...
                                                   0%
                                                   0%
                                                   1%
                                                   1%
<snip>
================================================== 100%

**** INFO: Done.
ASSERT:expected:<-358.05> but was:<-358.07>
testPlinkLinearMixedModelCovariates
GEMMA 0.98.4 (2021-01-29) by Xiang Zhou and team (C) 2012-2021
GSL_RNG_SEED=10
GSL random generator type: mt19937; seed = 10 (option -1); first value = 3312796937
Reading Files ...
**** INFO: no intercept term is found in the cvt file: a column of 1s is added.
## number of total individuals = 427
## number of analyzed individuals = 427
## number of covariates = 2
## number of phenotypes = 1
## number of total SNPs/var        =   358499
## number of analyzed SNPs         =   223242
Start Eigen-Decomposition...
pve estimate =0.365545
se(pve) =0.47481
                                                   0%
                                                   0%
                                                   1%
                                                   1%
<snip>
================================================== 100%

**** INFO: Done.
ASSERT:expected:<89757159113.94> but was:<89757159113.77>

#
# Test report
#
tests passed: 24
tests failed: 8
tests total:  32
success rate: 75%

The easyconfig for the GEMMA build is at:

https://github.com/easybuilders/easybuild-easyconfigs/blob/272f5294fed562dfa4a57e1eb8e640fcba39b0d8/easybuild/easyconfigs/g/GEMMA/GEMMA-0.98.4-foss-2020a.eb

with dependencies on GSL/2.6 and EIGEN/3.3.7. GCC/9.3.0, zlib/1.2.11, and OpenBLAS/0.3.9 are also in use.

In both cases, the failing builds are on OpenStack instances, running CentOS 7.9.

This didn't look quite the same as the Debian failures reported elsewhere.

Check list:

1. I have found an issue with GEMMA
2. I have searched for it on the issue tracker (incl. closed issues)
3. I have searched for it on the mailing list
4. I have tried the latest release of GEMMA
5. I have read and agreed to below code of conduct

[pjotrp]

Yes, sorry. These are round-off errors which were fixed for Debian. I'll change the test system soon.