aboutsummaryrefslogtreecommitdiffstats
path: root/pkgs/populate
diff options
context:
space:
mode:
authortv <tv@krebsco.de>2026-11-14 04:29:02 +0000
committertv <tv@krebsco.de>2026-11-14 06:26:16 +0100
commit87e5304d39ddf3a0c052cc13fcfdc2b36c3dad68 (patch)
tree5b7ccfcc44527024a2c9a091eb8d84fdc7e5f786 /pkgs/populate
parent41833dc11d75718c279e682593580a46dcec0c73 (diff)
tests: pure evaluation
pure-lib evaluates krops' lib from a flake whose inputs are krops' sources and nixpkgs' lib, checking that mkTarget leaves the user and the host null, and evaluating sources of each type. pure-deploy, which is exhaustive, deploys with writeDeploy from a flake whose inputs are krops' and nixpkgs' sources, evaluating nixpkgs, with a target string. The inputs are the sources in the store, without .git, copied there during the tests' evaluation, which the VM shares with the host. The flakes lock them as tarballs, by their hashes, so Nix uses the store paths, without fetching them, or reading them, which a path: input would, all of nixpkgs on every evaluation. The VM can't build what writeDeploy makes, which needs building populate, so pure-deploy runs what the host built of the same arguments, evaluated alike, failing if pure evaluation makes another derivation. nixpkgs is a flake with lib and legacyPackages, without nixpkgs' configuration and overlays, which pure evaluation can't read, and the host's evaluation could. pure-lib is in the smoke test.
Diffstat (limited to 'pkgs/populate')
0 files changed, 0 insertions, 0 deletions