darwin.libiconv: switch to libiconvReal and add to darwin-aliases

[reckenrode]

Description of changes

With #238993 and #299360, it is now possible to deprecate darwin.libiconv and switch Darwin to using upstream libiconv with a compatible ABI. This PR does that and updates packages that were using darwin.libiconv to use just libiconv.

Since this affects the stdenv, I’m targeting staging. I tested by building the stdenv on both platforms and GHC on aarch64-darwin. GHC uses install_name_tool to swap libiconv from the system one to the one in nixpkgs, so that it built successfully is a good test of the ABI compatibility implemented. I also confirmed eval locally using ofborg’s outpaths.nix.

Things done

• Built on platform(s)
  • x86_64-linux
  • 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.

[ghost]

looks like this is the first use of the darwin compatibility abi in libiconv. I think that the substituteInPlace should be changed to replace-fail . looking over the headers it seems that defining LIBICONV_PLUG almost has the same effect except that the extensions are not defined.

nixpkgs/pkgs/development/libraries/libiconv/default.nix

Lines 38 to 45 in bc06191

	for iconv_h_in in iconv.h.in iconv.h.build.in; do
	substituteInPlace "include/$iconv_h_in" \
	--replace "#define iconv libiconv" "" \
	--replace "#define iconv_close libiconv_close" "" \
	--replace "#define iconv_open libiconv_open" "" \
	--replace "#define iconv_open_into libiconv_open_into" "" \
	--replace "#define iconvctl libiconvctl" "" \
	--replace "#define iconvlist libiconvlist" ""

looks good otherwise.

[reckenrode]

looks like this is the first use of the darwin compatibility abi in libiconv.

I’m not surprised. I added it in anticipation of doing this PR. I originally had it build the compatible ABI unconditionally on Darwin, but it was desired to keep the upstream one the default so that libiconvReal would still correspond to it, so the feature flag was added.

I think that the substituteInPlace should be changed to replace-fail .

I can add it as clean up when I get home.

looking over the headers it seems that defining LIBICONV_PLUG almost has the same effect except that the extensions are not defined.

I think I looked at that, but it didn’t do what we need. Darwin needs a stand alone build with system symbol names (for compatibility with what Apple ships). It also needs to reexport libcharset.dylib, which is a platform peculiarity due to a historical accident.

[reckenrode]

I didn’t do a full bootstrap, but I did confirm libiconv built after changing it to use --replace-fail.

[nixos-discourse]

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/darwin-updates-news/42249/6