Rust and cargo improvements

[dvc94ch]

Motivation for this change

• Enable cross-compiling std crates
• Prepare for the 1.10.0 release when we can bootstrap from release tarball
• Update cargo and use release tarball
• Add scripts to fetch release tarball hashes for supported platforms

To try out rust on arm (depends on #15867):

with import <nixpkgs> {};

let
  pkgs = import ../../../nixpkgs;
  pkgsNative = pkgs {};
  pkgsCross = pkgs {
    inherit system;
    inherit crossSystem;
  };
  crossSystem = rec {
    config = "arm-linux-gnueabihf";
    bigEndian = false;
    arch = "arm";
    float = "hard";
    fpu = "vfpv3-d16";
    withTLS = true;
    libc = "glibc";
    platform = platforms.armv7l-hf-multiplatform;
    openssl.system = "linux-generic32";
    inherit (platform) gcc;
  };
  rustc = pkgsNative.rustc.override {
    targets = ["armv7-unknown-linux-gnueabihf"];
    targetToolchains = [pkgsCross.gccCrossStageFinal];
  };
  cargo = pkgsNative.cargo.override {
    inherit rustc;
  };
in
stdenv.mkDerivation rec {
    name = "arm-hello";

    buildInputs = [ cargo pkgsCross.gccCrossStageFinal qemu ];

    shellHook = ''
        PS1="rust-arm> "
        alias build="cargo build --target=${rustc.target}"
        alias run-debug="qemu-arm target/${rustc.target}/debug/${name}"
        alias run-release="qemu-arm target/${rustc.target}/release/${name}"
    '';
}

Things done

• Tested using sandboxing
  (nix.useSandbox on NixOS,
  or option build-use-sandbox in nix.conf
  on non-NixOS)
• Built on platform(s)
  • NixOS
  • OS X
  • Linux
• Tested compilation of all pkgs that depend on this change using nix-shell -p nox --run "nox-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• Fits CONTRIBUTING.md.

---

[mention-bot]

By analyzing the blame information on this pull request, we identified @wizeman, @abbradar and @madjar to be potential reviewers

[jagajaga]

cc @wizeman

[dvc94ch]

don't merge this just yet, I want to rework this. rust is moving to using cargo as a build system, so the rustc and cargo packages can share the bootstrapping code. should this be a new package then? or can we put the cargo and rustc code in the same folder so that we can share the bootstrapping code through local imports without polluting all-packages?

[the-kenny]

I'm fine with moving cargo and rustc to the same folder if that simplifies the implementation.

[dvc94ch]

So final building and testing is still pending, but I think it should be ok. If someone wants to test it on mac osx to make sure there are no regressions, that would be good...

[the-kenny]

Seems like there is an issue with evaluating the package set:

% nix-env -f . -qa
error: getting status of ‘/home/moritz/dev/nixpkgs/pkgs/development/tools/build-managers/cargo’: No such file or directory

[the-kenny]

Also, personally, I'd like an alias for rustc in all-packages.nix just like for cargo.

[dvc94ch]

mmh that is right there is no build-managers cargo anymore. have you added any custom patches to nixpkgs? I'm not having that problem... I rebased on master and added a sha256 fix for the exa package, and added a rustc alias, can you try again?

[the-kenny]

I didn't at the time. I re-applied everything after your recent rebase and now it's evaluating fine.

[the-kenny]

I'll check if everything builds fine with sandbox . Otherwise 👍

[the-kenny]

@dvc94ch

nix-build . -A rustc 
...
output path ‘/nix/store/dx95c8mp9g1gqvy5qin10xzwngw2yr97-rust-e4e8b66’ has r:sha256 hash ‘1pz4qx70mqv78fxm4w1mq7csk5pssq4qmr2vwwb5v8hyx03caff8’ when ‘167rh7hs77grn895h54s7np7f0k7b6i8z4wdfinncg4chy08hxq1’ was expected

[dvc94ch]

I get these spurious hash changes sometimes too. My rust src is currently cached, so I didn't notice the change. I think fetchgit isn't 100% deterministic.

[dvc94ch]

Pushed the hash change

[the-kenny]

Ugh, non-determinism in fetchgit isn't good. It pulls quite a bunch of submodules (I think), maybe it's related to that?

[dvc94ch]

Maybe. If I would make a wild guess I'd say this happens when a submodule
gets a new branch. But I haven't looked into this, since once it's
committed the package is in the binary cache, so it isn't an issue until
you make a change...
On Jun 15, 2016 1:13 PM, "Moritz Ulrich" [email protected] wrote:

Ugh, non-determinism in fetchgit isn't good. It pulls quite a bunch of
submodules (I think), maybe it's related to that?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#16052 (comment), or mute
the thread
https://github.com/notifications/unsubscribe/AAtRr6861CK49DlWtNivyJUDE2nav8z7ks5qL95SgaJpZM4IwQmw
.

[dvc94ch]

@the-kenny I know I'm inpatient, but I'll ask anyway. Are we good to go?

[the-kenny]

Sorry! I was busy at work yesterday (while my machine built all of Rust in Nix). To me it looks good. I think it's fine to merge it, but we should try to investigate the sha256 issue further.