Hi, Nikita Karetnikov <nik...@karetnikov.org> skribis:
>> Most likely the problem is that (%nixpkgs-directory) is #f. To fix >> that, you need to either run configure --with-nixpkgs=/path/to/nixpkgs, >> or set the ‘NIXPKGS’ environment variable. > > You're right. I forgot to specify this option. > > How can I tell Guix that 'nix-worker --daemon' listens on > '/nix/var/nix/daemon-socket/socket', not > '/usr/local/var/nix/daemon-socket/socket'? '--with-nix-prefix=/nix' > doesn't work. With --localstatedir=/nix/var/nix (see store.scm and config.scm). >>> It's possible that I've forgotten to install something. I ran >>> 'nix-env -i guile' to check that Nix-related things work fine, but I >>> interrupted the process when it decided to fetch the Linux kernel from >>> <kernel.org>. > >>> (Actually, it's a bit disappointing because I thought that >>> 'export HYDRA_DISALLOW_UNFREE=1' can handle this. But it's a different >>> issue.) > > Well, it turned out that this is a related issue. > > './pre-inst-env guix-build -K hello' also tries to fetch the kernel from > <http://www.all.kernel.org>. How to instruct Guix to use Linux-libre > instead? Remember: to bootstrap things here, you’re using Nixpkgs. Here it’s Nixpkgs’s glibc that uses kernel.org Linux headers. At this point, you don’t want to build Hello. ;-) Instead, you should build the bootstrap binaries, as described in HACKING, that will allow our distro to be self-contained. Make sure to use ‘core-updates’ as a starting point. HTH, Ludo’.
