diff options
| author | tv <tv@krebsco.de> | 2026-11-13 03:00:44 +0000 |
|---|---|---|
| committer | tv <tv@krebsco.de> | 2026-11-13 03:00:44 +0000 |
| commit | 3219e04ce4ff01b5f013e89d6245af9d28d73814 (patch) | |
| tree | a9d29ecf8238e5a58fb8c85b4fab49facf2accf1 /lib/types/populate.nix | |
| parent | 8c78f328bf7a1a4b97f26e5f5bb25f2d8b462f24 (diff) | |
tests: move combinations no code path tells apart to exhaustive tests
The matrices cross every option of a source type with every other one,
also where the code handling one never looks at the other. Scenarios
for such combinations are exhaustive, and run in their own tests, which
nix-build tests doesn't build, but nix-build tests -A exhaustive does:
- file and git: drawing inputs combined with rsync's or Git's options,
since drawing inputs happens after delivery.
- flake: materialize combined with how the flake gets there, except for
flakes in Git work trees and in subdirectories, since materializing
copies what got delivered.
- flake-input: materialize for inputs other than dep.
- deploy: /etc/nixos/flake.nix combined with deploying a flake, since it
only matters without one.
- edge-cases: generated flakes where only delivery matters, and remote
targets where drawing inputs alone matters, which happens on the target
alike for every kind of target.
That's 181 of 707 scenarios.
Diffstat (limited to 'lib/types/populate.nix')
0 files changed, 0 insertions, 0 deletions
