wip: let's see what breaks with boot.initrd.systemd.enable

[Mic92]

don't merge

[lilyinstarlight]

I know you're not looking to merge this, but as a quick note: systemd-initrd does not quite yet have correct initrd fs/disk dependencies, because systemd does not respect /sysroot in the fstab as needing the related root fs targets and nixos does not yet set root= on the kernel cmdline because it breaks other things (until the next systemd release where they have been fixed)

I don't know that the failures are really related to that, but it does make the occasional spooky problem with initrd ordering happen and so we should wait on at least NixOS/nixpkgs#208269 before turning it on for any real tests

(I'm sorry if you were already aware, since I know you're just trying to do a quick how-much-might-break test here and it's not the most relevant bit of info)

[ElvishJerricco]

@lilyinstarlight while I broadly agree, I will also note that I don't see anything that should cause failures like we're seeing here. So we've definitely got some bugs to squash.

[jackinloadup]

NixOS/nixpkgs/pull/208269 was merged :-)

[lilyinstarlight]

NixOS/nixpkgs/pull/208269 was merged :-)

With minor modifications, it looks like most disko tests work with systemd-initrd, but the mdadm tests still do not because of NixOS/nixpkgs#210210. So we'll still need that fixed before we can do a test matrix or something with systemd-initrd both on and off for all tests. But we could enable it for a handful of tests now I suppose, if we really wanted to

[Mic92]

rebased.