Unstable chroot production, dependent on system.

[Mats Erik Andersson]

Hello there,

three jobs at Hydra are using chroots when running tests
on GNU Inetutils:

   coverage           [http://hydra.nixos.org/build/21862798]

   build.i686-linux   [http://hydra.nixos.org/build/21862800]

   build.x86_64-linux [http://hydra.nixos.org/build/21862801]

Two of the included tests requires the file "/etc/protocols".
The presence of this file in the chroot was arranged using
"__impureHostDeps" some three weeks ago. In spite of that
the success rate has been irregular!

Using the three jobs stated above with their logging pages,
it turns out that "coverage" and "build.i686-linux" do have
"/etc/protocols" available in the chroot, whereas
"build.x86_64-linux" is explicitly skipping the test "inetd.sh"
because the protocols file is missing.

Question: How come that "hydra-recipes/inetutils/release.nix"
  is producing differently populated chroot environments for
  the to jobs "build.i686-linux" and "build.x86_64-linux"?
  Is Hydra using a non-deterministic and non-predictable
  configuration routine?

Best regards,
  Mats Erik Andersson, on behalf of GNU Inetutils.