I hit the same problem, on x86_64-linux
(not i386-linux
). There is a report of the same problem to Gentoo: https://bugs.gentoo.org/show_bug.cgi?id=927454
I rebuilt the failing derivation (by referencing it directly – nix build /nix/store/...-coreutils-....drv
) on the same build cluster and it built correctly. The failing and successful builds were on different machines, but the whole buildfarm is identical hardware running identical software. The kernel in use is Linux 6.1.62 but I don’t think that matters.
I use a much, much larger max-jobs cores
value than the 2
that Hydra uses. I think that probably matters.
The long-sloop.sh
test might be flaky.