Use nix daemon in the test suite

[cole-h]

Motivation

As part of the CLI stabilization effort (#7701), the last remaining checkbox (at the moment) for nix daemon is that it "needs testing". This implements the proposal of using nix daemon in place of nix-daemon in the test suite.

Context

The box for nix daemon is unchecked, and this is supposedly the only thing blocking it, so I figured I'd propose the most minimal diff to address it.

I grepped for nix-daemon in the tests/ subdirectory -- hopefully I didn't miss anything.

Priorities and Process

Add 👍 to pull requests you find important.

The Nix maintainer team uses a GitHub project board to schedule and track reviews.

[thufschmitt]

Thanks!

I don't think we can just replace nix-daemon by nix daemon everywhere, because we still want to test the old nix-daemon. That being said, I'm not sure why we need two different – but semantically equivalent – entry points for both: this one for nix-daemon and that one for nix daemon.

@edolstra, is there a reason why these are separate, or can we have nix-daemon piggyback on CmdDaemon?

[edolstra]

I think they're separate because originally we didn't have a single binary, so nix-daemon really was a separate executable. But I guess we can unify them now.

[thufschmitt]

But I guess we can unify them now.

Yay!

How do we do that? (technically I mean)

[Ericson2314]

I don't get what is so surprising? I don't think we've ever tried to make a legacy command new-style before. Maybe it's fine in this case, but it does mean we have to be careful about modifying nix daemon if it will modify nix-daemon too.

[thufschmitt]

@Ericson2314 : Nothing surprising, just a potential for simplification

[edolstra]

@cole-h Maybe we can arbitrarily keep some of the calls to nix-daemon, so that both get tested.

[cole-h]

I opted to keep nix daemon as the "overall" daemon for the tests and changed back a couple call sites to use nix-daemon again.

I wonder if it would be better to have another set of hydraJobs -- one for nix-daemon and one for nix daemon as the "overall" test daemon. This would essentially mean running the tests twice, of course, so I wonder if that might OK in the "near" term.

[roberth]

This would essentially mean running the tests twice

As you hinted, this strategy is quite expensive.
We could easily run a single tests/functional-style test twice, but we don't have one. All we have is tests that involve the daemon by coincidence, and a NixOS test about authorization, where it's not possible to pick a daemon command.

The right solution is to write a proper tests for the daemon command(s). If we don't merge their implementations, we can have two tiny tests that source the same test script.

[nixos-discourse]

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/2021-01-11-nix-team-meeting-minutes-115/38277/1