On Mon, 29 Aug 2016 20:20:30 +0200 David Craven <da...@craven.ch> wrote:
> > Note: It should infer the correct u-boot to use on its own - at least > > that's the intention with the whole make-u-boot-package thing. > > Did you see the updated u-boot patch? It includes support for > cross-compilation, Nice! > so make-u-boot-package takes a xgcc package and a > target triplet as arguments. So it probably broke the inference. No, I think I didn't make it work yet. I've modeled along the lines of what was already there for Grub. But I've got no idea where gnu/build/install.scm's install-grub finds "grub-install" - so the u-boot part doesn't work either since I don't know where to put the package reference in the first place (or how in the world it works for Grub to begin with). Also, there's a few mysterious things with install-grub. For example the documentation in the function header there says "Install Grub with GRUB.CFG on DEVICE, which is assumed to be mounted on MOUNT-POINT." I really doubt that it actually does what it says since usually grub is installed on the drive MBR and not inside a partition. And MOUNT-POINT is a mounted filesystem on a partition, right? So the whole drive is certainly not mounted at MOUNT-POINT. MOUNT-POINT originally comes from guix/scripts/system.scm - process-action. And that is called by process-command. And that is called by guix-system which is used when invoking "guix system" "reconfigure". But I've never specified a mountpoint when invoking that...
