lalsimulation 3.1.0 FAILED test_prec_avg_evol.py test, Debian Buster, arm64
As already reported in https://git.ligo.org/sccb/requests/-/issues/804#note_504128, the build of lalsimulation
3.1.0 on arm64 fails repeatedly for Debian Buster. amd64 is fine, as is arm64 Bullseye.
The failing test is
test_prec_avg_evol.py::test_inf[1.040655531768932e+32-2.257653018316939e+31-0.4672064855320761-0.740004232964156-1.7469211240711668-2.493332081990885-3.538962301482434-5.538310812576537-1.7689674276391143-2.4005245969349387-v1] FAILED [ 7%]
the build log is here: https://hypatia.aei.mpg.de/lsc-arm64-buster/lalsimulation_3.1.0-1+deb10u0_arm64.--pbuilderlog
To me, this looks like a precision issue caused by different initialization settings of libm
, and a small adjustment to the tolerance might already fix this?
Edited by Steffen Grunewald