On 20/06/2023 09.04, Paul Gevers wrote:
With the upload of 2.0.1-2 to unstable, we noticed that it now also
causes issues on s390x and arm64. On arm64, we loose connection with our
worker if it's testing libthrust for most of it's test time. On s390x,
the test exceeds 10 hours. 10 hours is really *a lot*, particularly for
tests that are NEUTRAL at most.
At least s390x is weird since it should only test the plain CPU variant ...
I'll add 'needs-some-thrusted-big-iron' to the Restrictions in order to
no longer interfere with the CI infrastructure, while still being able
to run the tests manually (yes, that takes awfully long, seems to be
exponential in the version number).
I'll try to investigate test behavior again once we have current CUDA
(12.1.1 last time I checked) and thrust (2.1.x) in sid, in case I ever
manage to catch up with these moving targets ;-)
Andreas
