It's a regression, but not one I'm very sure of how to fix sensibly. Some time after 2.3, Nix closed the soundness hole of allowing references for FOD store paths. This makes sense, as they aren't…
This should be mostly fixed now? Dropping the thread will just detach it; and joining it now behaves as intended.
Some question: How do we want to handle e.g. --store
or other nix-global arguments? Do we pass these as specific environment variables, I guess?
channel:
"file" syntax
yeah; I'm not too fussed since it's nixpkgs-specific; it might be a good first example for #342 (aka a nix-shell
but nixpkgs-specific, including easy channel selection) even.
channel:
"file" syntax
I actually use this personally; mostly in the form of NIX_PATH=nixpkgs=channel:nixos-unstable nix-shell -p foo
for fetching something from nixos-unstable
quickly. I don't strongly mind it…
hydraJobs.tests.nixpkgsLibTests.aarch64-darwin
yeah; that's my guess, and hence the issue :) Though I think this might be a bit more complex, because this involves the drv's build dir and seeing as ./.
failed I'm a bit more worried something…