[Nix-dev] A few questions about ARM support and NixOS on a Chromebook

James Haigh james.r.haigh at gmail.com
Fri Jan 30 19:54:57 CET 2015

On 28/01/15 01:49, Luke Clifton wrote:
> [...]
> My problem has always been lack of RAM in my ARM devices. I work
> mostly with embedded Linux platforms where RAM is < 256MB. To provide
> perspective, my Laptop (core i7) with 2GB of RAM fails to build some
> nixpkgs packages. Small CPU just means longer build time, small RAM
> means no build.
> [...]
Both of my ThinkPads X60 Tablets have 2GiB; I can upgrade them to about
3.2GB which I plan to do but I generally find 2GiB to be comfortable
amount. On the other hand, my Sony Xperia Z1 has also 2GiB, so for me,
my ARM hardware is at least as specified as my x86 hardware. I've seen a
couple of applications in the F-Droid repository that allow you to
chroot a GNU+Linux system onto Android. If we could do that with NixOS
then I could use my quad-core 2.2GHz Xperia Z1 (my highest specified
device) to build NixOS for itself and my other devices.
    It's an interesting point though about memory causing a build to
fail. I wasn't aware that builds needed so much RAM. If it's failing
even though packages are being built sequentially (rather than
concurrently) then that surely means that a single package build is
using all of that RAM. Does it really need that much?!! It seems strange
that the build is using more RAM than the space on disk that the
resultant package requires. I wonder if that's do to with the complex
optimisation algorithms. Do lesser optimisation levels require
significantly less space in memory?
    The other thing is, how well does compiling work with swap? Is a lot
of the memory idle during the build or does it require the potential to
read any part of that memory at any point? If the former, then swapping
could make the build succeed with less available RAM. It would take
longer, the question is how much longer. But at least it's better than


More information about the nix-dev mailing list