unit_test_diabatization fails on ppc64le
junghans opened this issue · comments
From https://koji.fedoraproject.org/koji/taskinfo?taskID=97058246:
249/253 Test #252: unit_test_diabatization ............................***Failed 12.47 sec
Running 3 test cases...
... ... 2023-2-3 18:16:3 Data was created with basis set def2-svp
... ... 2023-2-3 18:16:3 Using RI
... ... 2023-2-3 18:16:4 Data was created with basis set def2-svp/builddir/build/BUILD/votca-2023-rc.1/xtp/src/tests/test_diabatization.cc(91): �[1;31;49merror: in "diabatization_test/ER_coupling_test": difference{0.00394939} between angle_ref{0.7063121313715891} and angle{0.70910163025498574} exceeds 0.0001%�[0;39;49m
/builddir/build/BUILD/votca-2023-rc.1/xtp/src/tests/test_diabatization.cc(108): �[1;31;49merror: in "diabatization_test/ER_coupling_test": difference{1.18737e-05} between diabatic_H_ref(0, 0){0.22035608577958943} and diabatic_H(0, 0){0.22035870221444354} exceeds 1e-06%�[0;39;49m
/builddir/build/BUILD/votca-2023-rc.1/xtp/src/tests/test_diabatization.cc(109): �[1;31;49merror: in "diabatization_test/ER_coupling_test": difference{0.000874308} between diabatic_H_ref(0, 1){0.00046877322591792892} and diabatic_H(0, 1){0.00046918307805322856} exceeds 1e-06%�[0;39;49m
/builddir/build/BUILD/votca-2023-rc.1/xtp/src/tests/test_diabatization.cc(110): �[1;31;49merror: in "diabatization_test/ER_coupling_test": difference{0.000874308} between diabatic_H_ref(1, 0){0.00046877322591794279} and diabatic_H(1, 0){0.00046918307805322856} exceeds 1e-06%�[0;39;49m
/builddir/build/BUILD/votca-2023-rc.1/xtp/src/tests/test_diabatization.cc(111): �[1;31;49merror: in "diabatization_test/ER_coupling_test": difference{1.18658e-05} between diabatic_H_ref(1, 1){0.22050562862974854} and diabatic_H(1, 1){0.2205030121948944} exceeds 1e-06%�[0;39;49m
... ... 2023-2-3 18:16:7 Dipole Moment excited state 1 0.0214404 -0.239573 -2.74726e-05
... ... 2023-2-3 18:16:7 Dipole Moment excited state 2 0.154532 -0.885145 -1.99557e-05
... ... 2023-2-3 18:16:7 Data was created with basis set def2-svp
... ... 2023-2-3 18:16:7 Transition Dipole Moment 0.754728 0.131574 9.25604e-06
... ... 2023-2-3 18:16:10 Fragment 1 ground state 9.60042e-07
... ... 2023-2-3 18:16:10 Fragment 2 ground state -4.64864e-06
... ... 2023-2-3 18:16:10 Fragment 1 poulation electron -0.4994
... ... 2023-2-3 18:16:10 Fragment 1 poulation hole 0.4994
... ... 2023-2-3 18:16:10 Fragment 2 poulation electron -0.500599
... ... 2023-2-3 18:16:10 Fragment 2 poulation hole 0.5006
... ... 2023-2-3 18:16:10 dQ11 -1.41042e-06
... ... 2023-2-3 18:16:15 Fragment 1 poulation electron -0.500138
... ... 2023-2-3 18:16:15 Fragment 1 poulation hole 0.500138
... ... 2023-2-3 18:16:15 dQ22 -2.05877e-06
... ... 2023-2-3 18:16:15 Data was created with basis set def2-svp
... ... 2023-2-3 18:16:15 Fragment 1 transition -1.13604e-07
... ... 2023-2-3 18:16:15 Fragment 2 transition -9.55362e-08
... ... 2023-2-3 18:16:15 dQ12 -1.80674e-08/builddir/build/BUILD/votca-2023-rc.1/xtp/src/tests/test_diabatization.cc(240): �[1;31;49merror: in "diabatization_test/FCD_coupling_test": difference{2.14641e-06} between J_ref{0.00071879817182406039} and J{0.00071879971466092994} exceeds 0.0001%�[0;39;49m
�[1;31;49m*** 6 failures are detected in the test module "diabatization_test"
Should we increase the tolerance?
I think the problem is that there are already 3 zero digit before any real number.
I guess that's best. Is there a way to reproduce this somehow? Guess no one has that architecture...
https://www.votca.org/DEVELOPERS_GUIDE.html#failed-release-builds
ppc64le is Power8 etc, so not super uncommon.