icu: fix cross compilation

[uninsane]

Description of changes

recent icu package refactoring caused pkgsCross.aarch64-multiplatform.icu to fail with

error: function 'anonymous lambda' called with unexpected argument 'buildRootOnly'

so lets refactor it further :) and expose the build roots via passthru and thereby pass arguments in such a way that external users can use .override as if these were all ordinary packages.

Things done

• deployed a cross-compiled aarch64-multiplatform sway system with this patch
• deployed a native x86_64 system with this patch (it shouldn't trigger any rebuilds)
• Built on platform(s)
  • x86_64-linux (and pkgsCross.aarch64-multiplatform.icu)
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 24.05 Release Notes (or backporting 23.05 and 23.11 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[7c6f434c]

Oops sorry it looked reasonably close to the original passing of buildroot things…

But now this version breaks some native overrides?

[uninsane]

Oops sorry it looked reasonably close to the original passing of buildroot things…

But now this version breaks some native overrides?

oh, the ofborg output?

       at /ofborg/checkout/2/repo/38dca4e3aa6bca43ea96d2fcc04e8229/mr-est/ofborg-evaluator-2/pkgs/development/python-modules/tensorflow/default.nix:241:8:

          240|       # Necessary to fix the "`GLIBCXX_3.4.30' not found" error
          241|       (icu.override { inherit stdenv; })
             |        ^
          242|       jsoncpp

       … while calling anonymous lambda

weird, i would think that should work. maybe i need to use makeOverridable instead of callPackage. i'll poke around later/tomorrow.

[afh]

Firstly, apologies, @uninsane, for introducing a regression.
Secondly, thanks for jumping on this. Unforutately I do not have time to have a closer look today, but will in the coming days/as soon as you've poked around with this a more as you mentioned in your previous comment 🙂

[Mindavi]

I also took a quick look, but still a bit confused about why it's actually broken...

[anmonteiro]

I also noticed failures downstream in my workflows, see e.g. this GH actions run.

I tested the rev from this PR in my repo and it fixed the issue too, but I haven't attempted to reproduce the issue reported in #290761 (comment).

[uninsane]

i've updated this PR to fix the eval error, built tensorflow to verify.

top-level/all-packages.nix calls icu using callPackages (plural):

icu-versions = callPackages ../development/libraries/icu { };

and then inherits the icu-versions members into the top-level scope. callPackages seems to mean that icu/default.nix can return ordinary derivations, and an override applied to any of the icu packages works by overriding the arguments received by icu/default.nix -- not icu/make-icu.nix (which is how it would have worked if the caller was using lib.recurseIntoAttrs).

[kirillrdy]

happy to merge as it is, since a few build issues have been reported, i've just added one small suggestion

[kirillrdy]

at least this fixes pkgsCross.aarch64-multiplatform.icu we can make further re-factoring, feel free to ping me

@uninsane thank you for PR

[afh]

Thank you @uninsane and @kirillrdy for jumping on this and handling it so well, very much appreciated! 🙏
Glad to see that this got merged already. Apologies for not getting to this earlier.